Vol A

Marc
2011
Volume A: Theory and User Information
Corporate
MSC.Software Corporation
2 MacArthur Place
Santa Ana, CA 92707
Telephone: (800) 345-2078
FAX: (714) 784-4056
Europe
MSC.Software GmbH
Am Moosfeld 13
81829 Munich
GERMANY
Telephone: (49) (89) 43 19 87 0
Fax: (49) (89) 43 61 71 6
Asia Pacific
MSC.Software Japan Ltd.
Shinjuku First West 8F
23-7 Nishi Shinjuku
1-Chome, Shinjuku-Ku
Tokyo 160-0023, JAPAN
Telephone: (81) (3)-6911-1200
Fax: (81) (3)-6911-1201
Worldwide Web
www.mscsoftware.com
User Documentation: Copyright
2011 MSC.Software Corporation. Printed in U.S.A. All Rights Reserved.
This document, and the software described in it, are furnished under license and may be used or copied only in accordance with the
terms of such license. Any reproduction or distribution of this document, in whole or in part, without the prior written authorization of
MSC.Software Corporation is strictly prohibited.
MSC.Software Corporation reserves the right to make changes in specifications and other information contained in this document
without prior notice. The concepts, methods, and examples presented in this document are for illustrative and educational purposes
only and are not intended to be exhaustive or to apply to any particular engineering problem or design. THIS DOCUMENT IS
PROVIDED ON AN AS-IS BASIS AND ALL EXPRESS AND IMPLIED CONDITIONS, REPRESENTATIONS AND WARRANTIES,
INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, ARE
DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD TO BE LEGALLY INVALID.
MSC.Software logo, MSC, MSC., MD Nastran, Adams, Dytran, Marc, Mentat, and Patran are trademarks or registered trademarks of
MSC.Software Corporation or its subsidiaries in the United States and/or other countries.
NASTRAN is a registered trademark of NASA. Python is a trademark of the Python Software Foundation. LS-DYNA is a trademark of
Livermore Software Technology Corporation. All other trademarks are the property of their respective owners.
This software may contain certain third-party software that is protected by copyright and licensed from MSC.Software suppliers.
METIS is copyrighted by the regents of the University of Minnesota. HP MPI is developed by Hewlett-Packard Development Company,
L.P. MS MPI is developed by Microsoft Corporation. PCGLSS 6.0, copyright 1992-2005 Computational Applications and System
Integration Inc. MPICH Copyright 1993, University of Chicago and Mississippi State University. MPICH2 copyright 2002, University
of Chicago.
Use, duplication, or disclosure by the U.S. Government is subject to restrictions as set forth in FAR 12.212 (Commercial Computer
Software) and DFARS 227.7202 (Commercial Computer Software and Commercial Computer Software Documentation),
as applicable.
MA*V2011*Z*Z*Z*DC-VOL-A
Cont ent s
Marc Volume A: Theory and User Information

Contents
Preface
About This Manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Purpose of Volume A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Contents of Volume A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
How to Use This Manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1 The Marc System
Marc Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Marc for Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Mentat or Patran for GUI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Structure of Marc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Procedure Library. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Material Library. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Element Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Program Function Library. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Features and Benefits of Marc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2 Program Initiation
Marc Host Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Workspace Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Marc Workspace Requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
File Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Program Initiation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Examples of Running Marc Jobs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4
3 Data Entry
Input Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Input of List of Items. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Examples of Lists. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Table Driven Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Table Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Model Definition Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
History Definition Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
REZONE Option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4 Introduction to Mesh Definition
Direct Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Element Connectivity Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Nodal Coordinate Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Activate/Deactivate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
User Subroutine Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
MESH2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Block Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Merging of Nodes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Block Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Symmetry, Weighting, and Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Additional Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Mentat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
FXORD Option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Major Classes of the FXORD Option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Recommendations on Use of the FXORD Option. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Incremental Mesh Generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Bandwidth Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Rezoning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Substructure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Technical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Scaling Element Stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5 Contents
BEAM SECT Parameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Orientation of the Section in Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Definition of the Section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Error Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Local Adaptivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Number of Elements Created. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Location of New Nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Adaptive Criteria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Automatic Global Remeshing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Remeshing Criteria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Remeshing Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Patran Tetrahedral Mesher. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5 Structural Procedure Library
Linear Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Accuracy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Error Estimates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Adaptive Meshing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Fourier Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Nonlinear Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Geometric Nonlinearities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Eulerian Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Arbitrary Eulerian-Lagrangian (AEL) Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Nonlinear Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Buckling Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Perturbation Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Computational Procedures for Elastic-Plastic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
Creep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Viscoelasticity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Viscoplasticity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Fracture Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
Linear Fracture Mechanics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
Nonlinear Fracture Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
Numerical Evaluation of the J-integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Numerical Evaluation of the Energy Release Rate with the VCCT Method . . . . . . . . . . . . . . . . . . . . 146

6
Automatic Crack Propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Dynamic Fracture Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
Dynamic Crack Propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Mesh Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
Mesh Splitting Along Edges or Faces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
Mesh Cutting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Modal (Eigenvalue) Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Harmonic Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Spectrum Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Transient Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Inertia Relief . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Rigid Body Mode Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Rigid-Plastic Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Steady State Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
Superplasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Soil Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
Technical Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
Mechanical Wear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Design Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Theoretical Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
Design Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Approximation of Response Functions Over the Design Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Improvement of the Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
The Optimization Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
Marc User Interface for Sensitivity Analysis and Optimization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Defined Initial State with Result Data from Previous Analysis (including AXITO3D) . . . . . . . . . 203
Load and Displacement Boundary Conditions Transfer for PRE STATE Option. . . . . . . . . . . . . . . . . 204
Steady State Rolling Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Inertia Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Rolling Contact. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
Steady State Rolling with Marc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
7 Contents
Structural Zooming Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
Element Types Supported . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
Uncoupled Thermal Stress Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
Cure-Thermal-Mechanically Coupled Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Cure Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Cure Shrinkage Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6 Nonstructural Procedure Library
Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
Thermal Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Convergence Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Steady State Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Transient Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Temperature Effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Surface Energy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Thermochemical Ablation and Surface Energy Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Mathematical Presentation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Mechanical Erosion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Mechanical Erosion by Other Actions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Pyrolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Coking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Presentation of the Energy Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Ablation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Welding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
Conrad Gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
Channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
Output. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
Technical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
Hydrodynamic Bearing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
Electrostatic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

8
Magnetostatic Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
Magnetodynamic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Piezoelectric Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
Strain Based Piezoelectric Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
Acoustic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
Rigid Cavity Acoustic Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
Fluid Mechanics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
Finite Element Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
Penalty Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
Steady State Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
Solid Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
Solution of Coupled Problems in Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
Degrees of Freedom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
Element Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
Coupled Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
Thermal Mechanically Coupled Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
Coupled Acoustic-Structural Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
Fluid/Solid Interaction Added Mass Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
Coupled Electrostatic-Structural Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
Coupled Magnetostatic-Structural Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
Coupled Thermal-Electrical Analysis (Joule Heating) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
Coupled Electrical-Thermal-Mechanical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
Coupled Magnetostatic-Thermal Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
Coupled Magnetodynamic-Thermal Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
7 Material Library
Linear Elastic Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
Composite Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Layered Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
Classical Lamination Theory for Multi-Layered Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
Material Preferred Direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
Material Dependent Failure Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
9 Contents
Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D Composite Brick Elements . . . . . . . 376
Interlaminar Stresses for Continuum Composite Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
Progressive Composite Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
Mixture Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
Gasket . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
Constitutive Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
Nonlinear Hypoelastic Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
Thermo-Mechanical Shape Memory Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404
Transformation Induced Deformation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406
Constitutive Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
Phase Transformation Strains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
Experimental Data Fitting for Thermo-mechanical Shape Memory Alloy. . . . . . . . . . . . . . . . . . . . . . 409
Mechanical Shape Memory Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
Experimental Data Fitting for Mechanical Shape Memory Alloy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416
Conversion from Thermo-Mechanical to Mechanical SMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417
Elastomer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418
Updated Lagrange Formulation for Nonlinear Elasticity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
Time-independent Inelastic Behavior. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
Yield Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
Mohr-Coulomb Material (Hydrostatic Stress Dependence) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
Buyukozturk Criterion (Hydrostatic Stress Dependence) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
Powder Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
Obtaining Crush Curve and Shear Failure Parameters by Curve Fitting in Marc. . . . . . . . . . . . . . . . 448
Work or Strain Hardening. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452
Flow Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
Constitutive Relations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459
Time-independent Cyclic Plasticity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461
Time-dependent Inelastic Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
Creep (Maxwell Model). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468
Oak Ridge National Laboratory Laws. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
Swelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474
Viscoplasticity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
Time-dependent Cyclic Plasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
Anand Solder Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
Viscoelastic Material. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
Narayanaswamy Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 488

10
Temperature Effects and Coefficient of Thermal Expansion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
Piecewise Linear Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
Temperature-Dependent Creep. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494
Coefficient of Thermal Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495
Time-Temperature-Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495
Low Tension Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 498
Uniaxial Cracking Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 498
Low Tension Cracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
Tension Softening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
Crack Closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
Crushing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
Soil Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
Elastic Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
Cam-Clay Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
Evaluation of Soil Parameters for the Critical State Soil Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
Damage Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
Ductile Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
Elastomers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514
Cohesive Zone Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
Nonstructural Materials. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524
Heat Transfer Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524
Piezoelectric Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
Thermo-Electrical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
Coupled Electrical-Thermal-Mechanical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
Hydrodynamic Bearing Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
Fluid/Solid Interaction Analysis Added Mass Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
Electrostatic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
Magnetostatic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
Magnetodynamic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
Coupled Electrostatic-Structural . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
Acoustic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
Fluid Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
8 Contact
Definition of Contact Bodies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 530
Numbering of Contact Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532
11 Contents
Motion of Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 536
Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
Detection of Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
Shell Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 540
Neighbor Relations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 541
Implementation of Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542
Friction Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544
Coulomb Friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544
Shear Friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 553
Anisotropic Friction Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554
Glue Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
Deact Glue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
Breaking Glue. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
Separation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
Release . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
Glue Contact Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
Coupled Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 561
Thermal Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563
Joule Heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563
Contact in an Electrostatic or Piezoelectric Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
Contact in Magnetostatic Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
Contact in Magnetodynamic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
Contact in Soil Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
Contact in an Acoustic Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
Element Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
2-D Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 566
3-D Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 566
Shell Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568
Dynamic Impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568
Global Adaptive Meshing and Rezoning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 569
Adaptive Meshing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 569

12
Result Evaluation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 570
Tolerance Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571
Numerical and Mathematical Aspects of Contact. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572
Lagrange Multipliers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572
Penalty Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572
Hybrid and Mixed Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572
Direct Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572
Lagrange Multiplier Procedure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573
Direct Constraint Procedure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 574
Solution Strategy for Deformable Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 580
Iterative Penetration Checking. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 582
Instabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 582
Segment-to-Segment Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
Basic Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
Contact Body Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587
Contact Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587
Force Vector and Stiffness Matrix Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 590
Iterative Penetration Checking and Contact Stress Update. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592
Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593
Finite Sliding for Deformable Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594
Postprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595
9 Boundary Conditions
Loading. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 598
Loading Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 598
Face ID for Distributed Loads, Fluxes, Charge, Current, Source, Films, and Foundations . . . . . . . . . 607
Mechanical Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 611
Fluid Drag and Wave Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 614
Cavity Pressure Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615
Cyclic Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 619
Thermal Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
Initial Stress and Initial Plastic Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
Heat Fluxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
Mass Fluxes and Restrictors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 622
Electrical Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
Electrostatic Charges. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
Acoustic Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
Piezoelectric Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
Electrostatic-Structural Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
13 Contents
Magnetostatic Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
Magnetodynamic Currents and Charges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
Kinematic Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625
Transformation of Degree of Freedom. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627
Shell Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627
Tying Constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 630
Rigid Link Constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638
Shell-to-Solid Tying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
Rigid Tying to a Surface Patch. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
Overclosure Tying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
Insert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
AUTOMSET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
Support Conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
Bushings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646
Cyclic Symmetry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
Nastran RBE2 and RBE3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 652
Beam - Shell Offsets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657
Pin Code for Beam Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 660
Mesh Independent Connection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 660
CWELD patch-to-patch connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663
Parameters for the Projection Process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 667
Internal Constraints to Connect Two Surfaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 669
Multiple Surface Connections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 670
Coupled Thermo-mechanical Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 670
Rules for usage of the GS-node or the GA-, GB-node Pair in the Projection Process. . . . . . . . . . . . 671
CWELD Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 672
CWELD Input Styles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 675
CWELD Error and Warning Messages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676
CWELD Output. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 678
Two Dimensional Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 678
CFAST Connections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 679
CFAST Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 679
SWLDPRM, Special CWELD/CFAST Parameters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 680
10 Element Library
Truss Elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 687
Membrane Elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 688
Continuum Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 688

14
Beam Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 689
Plate Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 690
Shell Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 690
Heat Transfer Elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 690
Coupled Electrostatic-Structural . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 691
Fluid/Solid Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 691
Magnetodynamic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 692
Soil Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 692
Fluid Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 692
Piezoelectric Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 692
Special Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 692
Gap-and-Friction Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 692
Pipe-bend Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
Curved-pipe Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
Shear Panel Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
Cable Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
Rebar Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
Interface Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
Incompressible Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694
Large Strain Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694
Large Strain Plasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694
Rigid-Plastic Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694
Constant Dilatation Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695
Reduced Integration Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695
Continuum Composite Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695
Fourier Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696
Semi-infinite Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696
Cavity Surface Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696
Assumed Strain Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696
15 Contents
Follow Force Stiffness Contribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696
Explicit Dynamics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 697
Adaptive Mesh Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 697
11 Solution Procedures for Nonlinear Systems
Considerations for Nonlinear Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 699
Behavior of Nonlinear Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 700
Scaling the Elastic Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 700
Load Incrementation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 700
Selecting Load Increment Size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 702
Fixed Load Incrementation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 702
Arc Length Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 702
Residual Load Correction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 710
Restarting the Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 711
Newton-Raphson Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 712
Modified Newton-Raphson Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 713
Strain Correction Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 714
Direct Substitution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715
Arc-length Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715
Convergence Controls. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 722
Singularity Ratio. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724
AutoSPC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726
Solution of Linear Equations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 726
Direct Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 727
Iterative Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 727
Mixed Direct-Iterative Solver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 728
Preconditioners. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 728
Storage Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 729
Nonsymmetric Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 729
Complex Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 729
Iterative Solvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 729
Parallel Direct Solvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 730
Basic Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 730
Flow Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 731

16
Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 732
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733
12 Output Results
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735
Workspace Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735
Increment Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 738
Summary of Loads. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739
Timing Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739
Singularity Ratio. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739
Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739
Selective Printout. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 740
Options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 740
Grid Force Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 741
User Subroutines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 742
Restart. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 742
Element Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 742
Mentat Computed Element Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743
Solid (Continuum) Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 744
Solid-shell Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 744
Shell Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 744
Beam Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 745
Gap Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746
Linear and Nonlinear Springs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746
Joule Heating Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746
Piezoelectric or Electrostatic-Structural Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747
Magnetodynamics Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747
Nodal Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747
Stress Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747
Reaction Forces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
Residual Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
Dynamic Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
17 Contents
Joule Heating Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
Rigid-Plastic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
Hydrodynamic Bearing Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
Electrostatic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 748
Magnetostatic Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749
Magnetodynamic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749
Piezoelectric or Electrostatic-Structural Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749
Acoustic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749
Supplementary Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749
Contact Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749
Magnetodynamic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749
Post File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749
Forming Limit Parameter (FLP). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 750
Program Messages. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 753
Marc HyperMesh Results Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754
Marc SDRC I-DEAS Results Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754
Marc - ADAMS Results Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 755
Status File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 757
Element Group Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 757
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 758
13 Parallel Processing
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 760
Different Types of Machines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 760
Supported and Unsupported Features. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 760
Matrix Solvers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 762
Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 763
Domain Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 763
Running a Parallel Job . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 764
Domain Decomposition Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 766

18
14 Code Coupling Interfaces
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 771
Code Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 771
Coupling Regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773
Time Step Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 775
Shell Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 775
Parallel Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 775
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 775
A Finite Element Technology in Marc
Governing Equations of Various Structural Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 777
System and Element Stiffness Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 779
Load Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 780
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 781
B Finite Element Analysis of NC Machining Processes
General Description. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 783
NC Files (Cutter Shape and Cutter Path Definition) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 783
Intersection Between Finite Element Mesh and Cutter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785
Deactivation of Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785
Adaptive Remeshing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785
Typical Features for Machining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 786
Input of Initial Stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 786
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 786
Index
Preface
Preface
About This Manual 20
Purpose of Volume A 20
Contents of Volume A 20
How to Use This Manual 21
20 Marc Volume A: Theory and User Information

About This Manual
This manual is Marc Volume A, the first in a series of five volumes documenting the Marc Finite Element program.
The documentation of Marc is summarized below. You will find references to these documents throughout
this manual.
Purpose of Volume A
The purpose of this volume is:
1. To help you define your finite element problem by describing Marcs capabilities to model physical problems.
2. To identify and describe complex engineering problems and introduce Marcs scope and capabilities for
solving these problems.
3. To assist you in accessing Marc features that are applicable to your particular problems and to provide you with
references to the rest of the Marc literature.
4. To provide you with the theoretical basis of the computational techniques used to solve the problem.
Contents of Volume A
This volume describes how to use Marc. It explains the capabilities of Marc and gives pertinent background
information. The principal categories of information are found under the following titles:
Marc Documentation
TITLE VOLUME
Theory and User Information Volume A
Element Library Volume B
Program Input Volume C
User Subroutines and Special Routines Volume D
Demonstration Problems Volume E
Chapter 1 The Marc System
Chapter 2 Program Initiation
Chapter 3 Data Entry
Chapter 4 Introduction to Mesh Definition
Chapter 5 Structural Procedure Library
Chapter 6 Nonstructural Procedure Library
Chapter 7 Material Library
Chapter 8 Contact
Chapter 9 Boundary Conditions
Chapter 10 Element Library
21
Preface
The information in this manual is both descriptive and theoretical. You will find engineering mechanics discussed in
some detail. You will also find specific instructions for operating the various options offered by Marc.
How to Use This Manual
Volume A organizes the features and operations of the Marc program sequentially. This organization represents a
logical approach to problem solving using Finite Element Analysis. First, the database is entered into the system, as
described in Chapter 3. Next, a physical problem is defined in terms of a mesh overlay. Techniques for mesh definition
are described in Chapter 4. Chapters 5 and 6 describe the various structural analyses that can be performed by Marc,
while Chapter 7 describes the material models that are available in Marc. Chapter 8 describes the contact capabilities.
Chapter 9 discusses constraints, in the form of boundary conditions. Chapter 10 explains the type of elements that can
be used to represent the physical problem. Chapter 11 describes the numerical procedures for solving nonlinear
equations. Chapter 12 describes the results of the analysis in the form of outputs. Chapter 13 describes the use of
multiple processors when performing an analysis. Finally, Chapter 14 describes the coupling interface to external
solvers that is available through user subroutine programming.
This volume is also designed as a reference source. This means that all users will not need to refer to each section of
the manual with the same frequency or in the same sequence.
Chapter 11 Solution Procedures for Nonlinear Systems
Chapter 12 Output Results
Chapter 13 Parallel Processing
Chapter 14 Code Coupling Interfaces
Appendix A Finite Element Technology in Marc
Appendix B Finite Element Analysis of NC Machining Processes
Chapter 1 The Marc System
1
The Marc System
Marc for Analysis 23
Mentat or Patran for GUI 24
Procedure Library 24
Material Library 24
Element Library 25
Program Function Library 25

23
CHAPTER 1
The Marc System
The Marc system contains a series of integrated programs that facilitate analysis of engineering problems in the fields
of structural mechanics, heat transfer, and electromagnetics. The Marc system consists of the following programs:
Marc for Analysis
Mentat or Patran for GUI (For a detailed description of the supported functionalities by Patran, refer to the
Patran Marc Preference Guide.)
These programs work together to:
Generate geometric information that defines your structure (Marc and Mentat or Patran)
Analyze your structure (Marc)
Graphically depict the results (Marc and Mentat or Patran)
Figure 1-1 shows the interrelationships among these programs. Marc Programs discusses the Marc
component programs.
Figure 1-1 The Marc System
Marc Programs
Marc for Analysis
You can use Marc to perform linear or nonlinear stress analysis in the static and dynamic regimes, to perform heat
transfer analysis and electromagnetic analysis. The nonlinearities may be due to either material behavior, large
deformation, or boundary conditions. An accurate representation accounts for these nonlinearities.
Physical problems in one, two, or three dimensions can be modeled using a variety of elements. These elements
include trusses, beams, shells, and solids. Mesh generators, graphics, and postprocessing capabilities, which assist you
in the preparation of input and the interpretation of results, are all available in Marc. The equations governing
mechanics and implementation of these equations in the finite element method are discussed in Chapters 5, 6, 7, 8,
and 11.
Marc
Mentat
Mentat
Preprocessing
Analysis
Postprocessing
Patran
Patran
MARC AFEA
or
or

Mentat or Patran for GUI
Mentat is an interactive computer program that prepares and processes data for use with the finite element method.
Interactive computing can significantly reduce the human effort needed for analysis by the finite element method.
Graphical presentation of data further reduces this effort by providing an effective way to review the large quantity of
data typically associated with finite element analysis.
An important aspect of Mentat is that you can interact directly with the program. Mentat verifies keyboard input and
returns recommendations or warnings when it detects questionable input. Mentat checks the contents of input files and
generates warnings about its interpretation of the data if the program suspects that it may not be processing the data in
the manner in which you, the user, have assumed. Mentat allows you to graphically verify any changes the input
generates.
Mentat can process both two- and three-dimensional meshes to do the following:
Generate and display a mesh
Generate and display boundary conditions and loadings
Perform postprocessing to generate contour, deformed shape, and time history plots
The data that is processed includes:
Nodal coordinates
Element connectivity
Nodal boundary conditions
Nodal coordinate systems
Element material properties
Element geometric properties
Element loads
Nodal loads/nonzero boundary conditions
Element and nodal sets
Structure of Marc
Marc has four comprehensive libraries, making the program applicable to a wide range of uses. These libraries contain
structural procedures, materials, elements, and program functions. The contents of each library are described below.
Procedure Library
The structural procedure library contains procedures such as static, dynamic, creep, buckling, heat transfer, fluid
mechanics, and electromagnetic analysis.
The procedure library conveniently relates these various structural procedures to physical phenomena while guiding
you through modules that allow, for example, nonlinear dynamic and heat-transfer analyses.
Material Library
The material library includes many material models that represent most engineering materials. Examples are the
inelastic behavior of metals, soils, and rubber material. Many models exhibit nonlinear properties such as plasticity,
viscoelasticity, and hypoelasticity. Linear elasticity is also included. All properties may depend on temperature.
25
CHAPTER 1
The Marc System
Element Library
The element library contains over 200 elements. This library lets you describe any geometry under any linear or
nonlinear loading conditions.
Program Function Library
The program functions such as selective assembly, user-supplied subroutines, and restart, are tailored for user-
friendliness and are designed to speed up and simplify analysis work. Marc allows you to combine any number of
components from each of the four libraries and, in doing so, puts at your disposal the tools to solve almost any
structural mechanics problem.
Features and Benefits of Marc
Since the mid-1970s, Marc has been recognized as the premier general purpose program for nonlinear finite element
analysis. The programs modularity leads to its broad applicability. All components of the structural procedure,
material, and element libraries are available for use, allowing virtually unlimited flexibility and adaptability.
Marc has helped analyze and influence final design decisions on
Marcs clients gained the following benefits not attainable through other numerical or experimental techniques. These
benefits include:
Accurate results for both linear and nonlinear analysis
Better designs, which result in improved performance and reliability
The ability to model complex structures and to incorporate geometric and material nonlinear behavior
Documentation, technical support, consulting, and education provided by MSC.Software Corporation
Availability of Marc on most computers
Efficient operation
Automotive parts Space vehicles
Nuclear reactor housings Electronic components
Biomedical equipment Steam-piping systems
Offshore platform components Engine pistons
Coated fiberglass fabric roof structures Tires
Rocket motor casings Jet engine rotors
Ship hulls Welding, casting, and quenching processes
Elastomeric motor mounts Large strain metal extrusions
Chapter 2 Program Initiation
2
Program Initiation
Marc Host Systems 27
Workspace Requirements 28
File Units 30
Program Initiation 33
Examples of Running Marc Jobs 35

27
CHAPTER 2
Program Initiation
Chapter 2 explains how to execute Marc on your computer. Marc runs on many types of machines. All Marc
capabilities are available on each type of machine; however, program execution can vary among machine types. The
allocation of computer memory depends on the hardware restrictions of the machine you are using.
Marc Host Systems
Marc runs on most computers. Table 2-1 summarizes the types of machines and operating systems on which Marc
currently runs.
Table 2-1 Marc Machines and Operating Systems
Type OS Hardware
Fortran
Version
C
Version Default MPI
Also
Works On
HP (64-bit)
1
HPUX 11.23 Itanium 2 f90 3.4 A.06.05 HP MPI 2.2
IBM (64-bit)
1
AIX 5.3 Power 6 xlf 11.1 cc 9.0.0 MPICH
2
Linux (32-bit) RedHat RHEL5 Intel
Pentium or
equiv.
Intel 12.0 Intel 12.0
5
HP MPI 2.3
3
RHEL6
5
Linux (64 bit) RedHat RHEL5 Intel EM64T Intel 12.0 Intel 12.0
5
HP MPI 2.3
3
AMD Opteron, SuSE 10,
SuSE 11, RHEL6
5
Linux (64-bit) RedHat RHEL5 Itanium 2 Intel 10.1 Intel 10.1 HP MPI 2.2.51
3
SGI Altix
Windows (32-bit) Windows 7 Intel
Pentium or
equiv.
Intel 12.0 MSVS 10 Intel MPI 3.1 Windows XP, Vista 32, Intel
Fortran 10.1, Intel Fortran 11
Windows (64-bit) Windows 7 Intel EM64T Intel 12.0 MSVS 10 Intel MPI 3.1
4
AMD Opteron, Windows XP
64, Vista 64, Server 2008,
Intel Fortran 10.1,
Intel Fortran 11
1
Does not support the Pardiso and MUMPS solvers.
2
Hardware MPI version (POE) also available.
3
Supports the Intel MPI (version 3.1 for Itanium; 4.0 for the other).
4
Supports the Microsoft MPI 1.0 (SP1).
5
For user subroutines, use Intel 12.0; no previous compiler versions will work.

Workspace Requirements
Computing the amount of workspace required by Marc is a complex function of many variables. The most efficient
method is to use the default values for the allocation. The program dynamically acquires memory if necessary and if
available. In some situations, it is advantageous to initially allocate an amount of memory as described below. The
following sections discuss workspace requirements for Marc.
Marc Workspace Requirements
The workspace used by Marc is allocated in separate parts. One part is referred to as general memory and contains
items such as element stiffness matrices, assembled global stiffness, and mass matrices and decomposed operator
matrix for certain matrix solvers. The initial amount of memory for this part can be entered by the user and the program
dynamically allocates more memory if necessary. If no initial memory is specified, it is automatically allocated as
needed. Other parts of the dynamically allocated workspace can not be influenced by the user. This includes data for
elements, vectors, tables, sets, contact bodies, kinematic boundary conditions, transformations, tying, the so-called
incremental backup, solver workspace for certain solvers among other things. These are all allocated separately. The
incremental backup is an extra copy of stress tensors and similar quantities and is used for the Newton-Raphson
iterations in a non-linear analysis. If the cut-back feature is activated, more data is stored in this part to allow for
redoing the increment with a modified time step if a failure occurs.
1. Element Data memory allocation is reduced if ELSTO is used. This also reduces Incremental Backup memory.
2. Assembled stiffness matrix memory allocation is reduced if out-of-core solver is used.
3. Decomposed Stiffness matrix is for solver type 0, 4, and if Cholesky preconditioner is used, solver type 2.
4. Incremental Backup memory allocation is reduced if IBOOC is used.
5. Decomposed Stiffness matrix is for solver type 6 and 8.
During the analysis, some of the blocks may grow because of changes in the model. In particular because of local or
global adaptive meshing, the Element Data, Vector Data, Assembled Stiffness Matrix, and Decomposed Stiffness
Matrix may grow in size because of the addition of elements and nodes. The Assembled Stiffness and Decomposed
Stiffness matrices also expand due to changes in the bandwidth of the system. Changes in the bandwidth occur when
deformable-deformable or self-contact occur. These changes can have a dramatic effect of the amount of memory
required. The program dynamically requests additional memory. If this memory is not available, it activates one or
more of the out-of-core options.
Basic Data Assembled Stiffness Matrix (2) Decomposed Stiffness Matrix (3)
General Memory stored in common/space/
Second Group of Memory stored in individual dynamically allocated vectors
Contact Data Boundary Conditions Transformation Tying Attach
Incremental Backup (4) Decomposed Stiffness Matrix (5)
Tables Sets Vectors Element Data (1)
29
CHAPTER 2
Program Initiation
The amount of memory that can be used by an analysis is limited by the hardware employed in the operating system
and by internal restrictions within Marc. On 32 bit systems, it is, in general, not possible to allocate more than two GB
of memory for a process. The exception is Windows 2003 Server (which allows three GB). Even on this platform, the
size of the largest vector is two GB. If this limit is reached, the memory allocation request by Marc fails. Also, a failure
to allocate memory can occur if other processes are using large amounts of memory. A typical output when a memory
request fails is
memory request of 250000000 words failed
This occurs when Marc sends a memory allocation request (using the C function malloc) and the system refuses the
request. There is also an internal restriction in the standard version of Marc. It uses standard Fortran 32-bit integers as
pointers in vectors, and the largest amount of memory that can be addressed with this is 8 GB on a 64-bit system. This
limit applies to each part of the separately allocated memory. For example, in a job where the general memory requires
4 GB, the solver needs 6 GB and the incremental backup needs 4 GB. This job can be run provided that the machine
has at least 14 GB of available memory. For a parallel analysis, the 8 GB restriction, as well as the 2 GB limit for 32
bit system, is for each domain of the job since each domain corresponds to a separate process. In a separate set of
versions of Marc utilizing 64-bit integers, this 8 GB restriction is removed.
The memory allocation for the general memory and matrix solver can be affected by the user. The ALLOCATE
parameter specifies the amount of memory that is initially allocated for this part. This amount is allocated regardless
of whether it is used or not. The default is set to a small value and the memory grows as needed.
For large problems, you may want to get an estimate of the workspace requirements for running a job without actually
executing the analysis. To do this, insert the STOP parameter to exit the program normally after the workspace is
allocated. Marc prints out a summary of the memory needed. It should be noted that the reported workspace with this
option is only an estimate. Some parts of the memory will grow during the analysis and there is no way to predict this
without actually running the analysis. For setting the appropriate initial allocation to avoid memory growth of the
general memory (which can be inefficient), one should look at the entry on general memory of the summary printout.
Please note that the ALLOCATE parameter value in a parallel run refer to the complete model; the specified amount is
divided equally among the domains. Chapter 12 Output Results describes the Marc output related to memory in
more detail.
By default, the data is stored in-core. There are three out-of-core storage options in Marc.
Out-of-core element data storage (the ELSTO parameter also implies IBOOC)
Out-of-core storage of incremental backup (the IBOOC parameter)
Out-of-core matrix solution
The out-of-core element data option stores element arrays (strains, stresses, temperatures, etc.) on a file (Fortran unit
3). Data connected with storage of all element quantities occupy a large amount of space for the more complex shell
or three-dimensional elements. Putting this data out-of-core leads to a slowdown of the execution (disk access is, in
general, slower than memory access) but the effect of this is usually not severe. The ELSTO parameter is used to set
this option. The out-of-core option for incremental backup stores this data in a file (Fortran unit 29). If the element
data is out-of-core, the incremental backup is automatically out-of-core as well. The slow-down related to incremental
backup out-of-core is much less than the one related to element data out-of-core.
The out-of-core matrix solution has a more severe effect on the execution speed. Not all matrix solvers support this
option. It is not supported by the iterative sparse solver (Marc solver number 2) and the hardware vendor provided
solver by HP (solver 6). The multifrontal direct solver (solver 8) and the PARDISO solver (solver 11) does support
out-of-core matrix solution.

In the case that a memory allocation request fails, Marc automatically switches to out-of-core storage unless it is too
late to do this at the point where the allocation failure occurs. If the memory allocation for element data or incremental
backup fails, this part is put out-of-core. The incremental backup part can also automatically be put out-of-core to free
up memory for the matrix solver. If the default solver is used (multifrontal, solver 8), there is a special way for
checking if out-of-core is needed. The solver workspace is allocated in two parts. One in general memory for the
assembly and one in separate memory. If the in-core assembly part takes more than 25% of the available memory, the
matrix assembly is done out-of-core. This ensures that this decision is done early enough so there is memory left for
other parts of the job. The actual LDU decomposition of the matrix is done out-of-core if the in-core decomposition
requires more than the available memory to avoid the usage of swap memory during the decomposition. The available
memory is set in the file tools/include (for Windows, tools\include.bat) in the installation directory as the
variable MEMLIMIT. For most platforms (Unix/Linux except HP PA-RISC), it is set to the amount of physical memory
on the machine using a shell command. This can be set explicitly at the installation or specified with the start-up script
run_marc with the option -ml new_memlimit. For Windows and HP PA-RISC, the automatic setting of the amount
of physical memory is done in the program.
File Units
Marc uses auxiliary files for data storage in various ways. Particular FORTRAN unit numbers are used for certain
program functions (for example, ELSTO, RESTART). Table 2-2 lists these file unit numbers.
Several of these files are necessary for solving most problems. The program input file and program output file are
always required.
Note: On most systems, the files are referenced by file names as well as by the file
Table 2-2 FORTRAN File Units Used by Marc
File name Unit Description File Type
jidname.log 0 Analysis sequence log file sequential access, formatted
jidname.t01 1 Usually contains mesh data random access, formatted
jidname.t02 2 OOC* solver scratch file random access, binary
jidname.t03 3 Element data storage (see ELSTO parameter) random access, binary
jidname.dat 5 Data input file sequential access, formatted
jidname.out 6 Output file sequential access, formatted
jidname.t08 8 Restart file, written out sequential access, binary
ridname.t08 9 Restart file to be read in from a previous job sequential access, binary
jidname.t11 11 OOC* solver scratch file sequential access, binary
jidname.t14 14 OOC* solver scratch file random access, binary
jidname.t16 16 Post file, written out sequential access, binary
ridname.t16 17 Post file to be read in from a previous job sequential access, binary
*OOC denotes Out-Of-Core solution.
31
CHAPTER 2
Program Initiation
jidname.t18 18 Mesh optimization correspondence table sequential access, formatted
jidname.fem 18 From Marc to external mesher sequential access, formatted
jidname.t19 19 Post file, written out sequential access, formatted
ridname.t19 20 Post file to be read in from a previous job sequential access, formatted
jidname_j_.dat 21 Temporary input file when cut-back is used. sequential access, formatted
jidname.t22 22 Subspace iteration scratch file random access, binary
jidname.t23 23 Fluid-solid interaction file sequential access, binary
pidname.t19 24 Temperature post file for CHANGE STATE or post
file from previous analysis for PRE STATE or
GLOBALLOCAL or MAP TEMP.
sequential access, formatted
pidname.t16 25 Temperature post file for CHANGE STATE or post
file from previous analysis for PRE STATE or
GLOBALLOCAL or MAP TEMP.
sequential access, binary
jidname.t29 29 Incremental backup file when ELSTO, IBOOC is
used, or insufficient memory exists.
sequential access, binary
sidname.t31 31 Substructure results file sequential access, binary
jidname.t32 32 Secant method file sequential access, binary
jidname.t33 33 Lanczos scratch file sequential access, binary
sidname.t35 35 Substructure results file sequential access, binary
material.mat 38 Material data base file sequential access, formatted
jidname.g 39 Intergraph post file sequential access, formatted
jidname.unv 40 I-DEAS Universal post file sequential access, formatted
jidname.t41 41 Post file Domain Decomposition sequential access, binary
ridname.t42 42 Post file Domain Decomposition sequential access, formatted
jidname.opt 45 Duplicate load case data file during design
optimization run
jidname.t46 46 Design optimization scratch file sequential access, binary
jidname.trk 47 New particle tracking file sequential access, formatted
ridname.trk 48 Old particle tracking file sequential access, formatted
userspecified 49 User default file (see Marc Volume C: Program Input,
Appendix C: Default File)
jidname.vfs or
jidname_CXX.vfs
50 Viewfactors for cavity number xx sequential access, formatted
jidname.lck 51 Locking of post file sequential access, formatted
jidname.cnt 52 Dynamic control file sequential access, formatted
jidname.mfd 52 rebar - Mentat interface sequential access, formatted
jidname-bbc.mfd 52 beam-beam contact - Mentat interface sequential access, formatted
jidname.seq 53 Sequence option sequential access, formatted
jidname.rst 54 Load case data sequential access, formatted
jidname.mesh 55 User supplied mesh sequential access, formatted
Table 2-2 FORTRAN File Units Used by Marc (continued)

jidname.feb 55 From 3-D mesher to Marc sequential access, formatted
jidname.pass 56 Auto restart command line sequential access, formatted
jidname.rms 57 2-D outline file for remeshing sequential access, formatted
jidname.domesh
jidname.donemesh
59 Lock files indicating meshing status
do mesh and done mesh
jidname.doview
jidname.doneview
59 Lock file indicating viewfactor calculation status sequential access, formatted
jidname.sltrk 60 New streamline tracking file sequential access, formatted
ridname.sltrk 61 Old streamline tracking file sequential access, formatted
jidname.sts 67 Analysis progress reporting file sequential access, formatted
bbctch.noconv 80 beam-beam contact information sequential access, formatted
jidname.t81 81 Multifrontal OOC scratch file random access, binary
jidname.t82 82 Multifrontal OOC scratch file random access, binary
jidname.t83 83 Multifrontal OOC scratch file sequential access, binary
jidname.t84 84 Multifrontal OOC scratch file sequential access, binary
jidname.t85 85 Multifrontal DDM scratch file sequential access, binary
jidname.fld 91 Forming Limit input file sequential access, formatted
jidnamd.stm 91 Output at streamline integration points -
PYROLYSIS parameter.
jidname.rec 96 output file of recession surface -
ABLATION parameter.
filename.apt 94 APT file - machining option sequential access, file
filename.ccl 95 CL file - machining option sequential access, file
EXITMSG 97 Exit messages sequential access, formatted
USRDEF 98 User global default file (see Marc Volume C: Program
Input, Appendix C: Default File)
jidname.t08 99 Base restart file for DDM sequential access, formatted
jidname.grd 103 Grid Force Balance Output sequential access, formatted
user specified 110 - 119 Include files for input sequential access, formatted
jidnam-dmig* 120 - 130 DMIG output files. sequential access, formatted
jidname.hmr N/A Hypermesh results file sequential access, binary C file
jidname.dump N/A Scratch file used during memory reallocation on
Windows if in-core reallocation fails.
sequential access, binary C file
Table 2-2 FORTRAN File Units Used by Marc (continued)
33
CHAPTER 2
Program Initiation
Program Initiation
Procedures (shell script) are set up that facilitate the execution of Marc on most computers. These procedures invoke
machine-dependent control or command statements. These statements control files associated with a job.
This shell script submits a job and automatically takes care of all file assignments. This command must be executed
at the directory where all input and output files concerning this Marc job are available. To use this shell script, every
Marc job should have a unique name qualifier and all Marc output files connected to that job uses this same qualifier.
For restart, post, change state, all default Marc Fortran units should be used.
To actually submit a Marc job, the following command should be used:
run_marc -prog prog_name -jid job_name -rid rid_name -pid pid_name \
-sid sid_name -queue queue_name -user user_name -back back_value \
-ver verify_value -save save_value -vf view_name -def def_name \
-nprocd number_of_processors (for Single Input file runs, use -nps
number_of_processors) -nthread number_of_threads \
-dir directory_where_job_is_processed -itree message_passing_type \
-host hostfile (for running over the network) -pq queue_priority \
-at date_time -comp compatible_machines_on_network -cpu time_limit \
-nsolver number of processors (MUMPS) -ml memory_limit -mode i4 -mpi intel-mpi
where the \ provides for continuation of the command line.
Table 2-3 Keyword Descriptions*
Keyword Options Description
-jid (-j) job_name Input file (job) name identification.
-prog (-pr) progname Run saved executable progname.marc from a previous job
(see -user and -save).
-user (-u) user_name User subroutine user_name.f is used to generate a new executable
program called user_name.marc (see -save and -prog).
-save (-sa) no
yes
Do not save the new executable program user_name.marc.
Save the executable program user_name.marc for a future time (see -
prog and -user).
-rid -(r) restart_name Identification of previous job that created RESTART file.
-pid (-pi) post_name Identification of previous job that created the post file.
-sid (-si) substructure Identify the job that contains the solution to the external nodes of the
superelement.
-back (-b) yes
no
Run Marc in the background.
Run Marc in the foreground.
-ver (-v) yes
no
Ask for confirmation of these input options before starting the job.
Start the job immediately.
*Default options are shown in bold.

-def (-de) default_file File name containing user defined default data.
-nprocd (-np) number Number of domains for parallel processing.
-nprods (-nps) number Number of domains for parallel processing using a Single Input file.
-nthread number Number of threads per parallel matrix solver (solver 6, 8, and 11) on
hardware that supports this capability.
-nsolver number Number of processes to use for the MUMPS parallel matrix solver (solver
12).
-dir directory_name Pathname to directory where the job I/O should take place.
Defaults to current directory.
-sdir directory_name Directory where scratch files are placed. Defaults to -dir.
-host (-ho) hostfile Specify the name of the host file for running over a network (default is
execution on one machine only in which case this option is not needed).
-ci yes
no
Copy input files automatically to remote hosts for a network run,
if necessary.
-cr yes
no
Copy post files automatically from remote hosts used for a network run,
if necessary.
-vf viewfactor_name Name of file containing viewfactors for radiation from previous analysis or
from Mentat using either the Monte Carlo or Hemicube method.
-mo i4
i8
Indicate if the i4 or i8 version is used if both are installed; if the default
unless the run_marc_defaults file is used.
Table 2-3 Keyword Descriptions* (continued)
35
CHAPTER 2
Program Initiation
Examples of Running Marc Jobs
Example 1:
run_marc -jid e2x1
This runs the job e2x1 in the background using a single processor.
The input file is e2x1.dat in the current working directory.
Example 2:
run_marc -jid e2x14 -user u2x14 -sav y -nproc 4
This runs the job e2x14 in the background with four processors. The user subroutine is linked with the Marc library
and a new execu module is created as u2x14.marc and saved in the current working directory after completion of the
job.
-mpi intel-mpi
ms-mpi
hpmpi
hardware
Select which version of mpi will be used for Domain Decomposition or the
MUMPS solver. If not specified, the appropriate default shown below will
be used:
Platform Default MPI
Alternative MPI
Windows 32 intel-mpi none
Windows 64 intel-mpi ms-mpi
Linux 32 hpmpi intelmpi
Linux 64 hpmpi intelmpi
Linux Itanium hpmpi intelmpi
HP UX hpmpi none
HP UX Itanium hpmpi none
IBM mpich hardware
SGI Altrix hardware none
SGI/Irix mpich hardware
SUN/Solaris/SPARC mpich hardware
SUN/Solaris/8664 hardware none
-ml memlimit Memory limit for deciding if the solver should go out-of-core. Specified in
Mbyte. Defaults to the physical amount of memory on the machine.
Table 2-3 Keyword Descriptions* (continued)

Example 3:
run_marc -jid e2x14a -prog u2x14 -nproc 4
Use the above saved module u2x14.marc to run the job e2x14a in the background with four processors.
Example 4:
run_marc -jid e3x2a -v no -b no -nproc 2
Run the job e3x2a in the foreground with two processors. The job runs immediately without verifying any arguments
interactively. If there are any input errors in the arguments, the job does not run and the error message is sent to the
screen.
Example 5:
run_marc -jid e3x2b -rid e3x2a
Run the job e3x2b in the background using a single processor. The job uses e3x2a.t08, which is created from
Example 4, as restart file.
Example 6:
run_marc -jid e2x1 -nproc 2
Runs a two processor job on a single parallel machine.
Example 7:
run_marc -jid e2x1 -nproc 2 -host hostfile
Runs a two-processor job over a network. The hosts are specified in the file hostfile.
Example 8:
run_marc -jid e2x1 -nps 2
Runs a two-processor job on a single parallel machine using a single input file.
Example 9:
run_marc -jid e2x1 -nps 2 -host hostfile
Runs a two-processor job over a network using a single input file. The hosts are specified in the file hostfile.
Example 10:
run_marc -jid e2x10 -mo i8
Runs a job in large integer mode.
Example 11:
run_marc -jid e8x102d -nthread 4
Runs a job using four threads on a machine that supports parallel processing. The Pardiso solver should also be
activated.
Chapter 3 Data Entry
3
Data Entry
Input Conventions 38
Table Driven Input 41
Parameters 46
Model Definition Options 46
History Definition Options 46
REZONE Option 47

The input data structure is made up of three logically distinct sections:
1. Parameters describe the problem type and size.
2. Model definition options give a detailed problem description.
3. History definition options define the load history.
Input data is organized in (optional) blocks. Key words identify the data for each optional block. This form of input
enables you to specify only the data for the optional blocks that you need to define your problem. The various blocks
of input are optional in the sense that many have built-in default values which can be used by Marc in the absence
of any explicit input from you.
Input Conventions
Marc performs all data conversion internally so that the system does not abort because of data errors made by you.
The program reads all input data options alphanumerically and converts them to integer, floating point, or keywords,
as necessary. Marc issues error messages and displays the illegal option image if it cannot interpret the option data
field according to the specifications given in the manual. When such errors occur, the program attempts to scan the
remainder of the data file and ends the run with an exit error message at the END OPTION option or at the end of the
input file.
Two input format conventions can be used: fixed and free format. You can mix fixed and free format options within a
file, but you can only enter one type of format on a single option.
The syntax rules for fixed fields are as follows:
You must right-justify integers in their fields. (The right blanks are filled with zeroes).
Give floating point numbers with or without an exponent. If you give an exponent, it must be preceded by the
character E or D and must be right-justified.
The syntax rules for free fields are as follows:
Check that each option contains the same number of data items that it would contain under standard fixed-
format control. This syntax rule allows you to mix fixed-field and free-field options in the data file because the
number of options you need to input any data list are the same in both cases.
Separate data items on a option with a comma. The comma can be surrounded by any number of blanks.
Within the data item itself, no embedded blanks can appear.
Give floating point numbers with or without an exponent. If you use an exponent, it must be preceded by the
character E or D and must immediately follow the mantissa (no embedded blanks).
Give keywords exactly as they are written in the manual. Embedded blanks do not count as separators here (for
example, BEAM SECT is one word only).
If a option contains only one free-field data item, follow that item with a comma. For example, the number 1
must be entered as 1, if it is the only data item on a option. If the comma is omitted, the entry is treated as
fixed format and may not be properly right-justified.
39
CHAPTER 3
Data Entry
If the EXTENDED parameter is used, integer data is given using 10 fields as opposed to 5 fields. This allows
very large models to be included in Marc. Additionally, real numbers are entered using 20 or 30 fields as
opposed to 10 or 15. This allows increased accuracy when reading in data.
All data can be entered as uppercase or lowercase text.
Input of List of Items
Marc often requests that you enter a list of items in association with certain program functions. As an example, these
items can be a set of elements as in the ISOTROPIC option, or a set of nodes as in the POINT LOAD option. Fourteen
types of items can be requested:
This list can be entered using either the OLD format (compatible with the G, H, and J versions of Marc) or the NEW
format (the K version).
Using the OLD format, you can specify the list of items in three different forms. You can specify:
1. A range of items as:
m n p
which implies items m through n by p. If p is not specified, the program assumes it is 1. Note that the range
can either increase or decrease.
2. A list of items as:
-n a
1
a
2
a
3
... a
n
which implies that you should give n items, and they are a ,...a.
3. A set name as:
MYSET
which implies that all items previously specified in the set MYSET are used. Specify the items in a set using the
DEFINE model definition option.
Using the NEW format, you can express the list of items as a combination of one or more sublists. These sublists can
be specified in three different forms. The following operations can be performed between sublists:
AND
INTERSECT
EXCEPT
Element numbers Curves IDs
Node numbers Surfaces IDs
Degree of freedom numbers Body numbers
Integration point numbers Edges pairs
Layer numbers Faces pairs
Increment numbers Oriented curves
Points IDs Oriented surfaces

When you form a list, subsets are combined in binary operations (from left to right). The following lists are examples.
1. SUBLIST1 and SUBLIST2
This list implies all items in subsets SUBLIST1 and SUBLIST2. Duplicate items are eliminated and the
resulting list is sorted.
2. SUBLIST1 INTERSECT SUBLIST2
This list implies only those items occurring both in subsets SUBLIST1 and SUBLIST2; the resulting list is
sorted.
3. SUBLIST1 EXCEPT SUBLIST2
This list implies all items in subset SUBLIST1 except those which occur in subset SUBLIST2; the resulting
list is sorted.
4. SUBLIST1 AND SUBLIST2 EXCEPT SUBLIST3 INTERSECT SUBLIST4
This list implies the items in subsets SUBLIST1 and SUBLIST2 minus those items that occur in subset
SUBLIST3. Then, if the remaining items also occur in subset SUBLIST4, they are included in the list.
Sublists can have several forms. You can specify:
1. A range of items as:
m TO n BY p
or
m THROUGH n BY p
which implies items m through n BY p. If BY p is not included, the program assumes BY 1. Note that
the range can either increase or decrease.
2. A string of items as:
a1 a2 a3 ... an
which implies that n items are to be included. If continuation options are necessary, then either a C or
CONTINUE should be the last item on the option.
3. A set name as:
MYSET
which implies that all items you previously specified to be in the set MYSET are used. You specify the items in
a set using the DEFINE option.
In a list, edges, faces, oriented curves, and oriented surfaces are entered as pairs (i:j) where i is the user element id
and j is the edge id or face id. The edge id/face id for the different element classes is given in Marc Volume C: Program
Input, Chapter 1.
An oriented curve or an oriented surface is used in conjunction with shell elements to indicate if the top surface or the
bottom surface of the shell is to be used.
Note:
INTERSECT or EXCEPT cannot be used when defining lists of degrees of freedom.
41
CHAPTER 3
Data Entry
There are two types of edge and face sets; those expressed in Marc or Mentat convention. The edge/face ID in the Marc
convention is one greater than the Mentat convention.
For example, to specify edge 1 on elements 1 to 20, one would use:
1:1 TO 20:1
Examples of Lists
This section presents some examples of lists and entry formats.
Use the DEFINE model definition option to associate a list of items with a set name with items. Three sets are defined
below: FLOOR, NWALL, and WWALL.
DEFINE NODE SET FLOOR contains:
1 TO 15 (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15)
DEFINE NODE SET NWALL contains:
5 TO 15 BY 5 AND 20 TO 22 (5,10,15,20,21,22)
DEFINE NODE SET WWALL contains:
11 TO 20 (11,12,13,14,15,16,17,18,19,20)
Some possible lists are:
NWALL AND WWALL, which would contain nodes:
5 10 11 12 13 14 15 16 17 18 19 20 21 22
NWALL INTERSECT WWALL, which would contain nodes:
15 20
NWALL AND WWALL EXCEPT FLOOR, which would contain nodes:
16 17 19 20 21 22
Table Driven Input
An alternative approach to defining the data which is based upon defining nonlinear material properties and variations
in the boundary conditions using tables. This input procedure is activated by use of the TABLE parameter. There are
six aspects to this input procedure:
1. The data associated with material properties, boundary conditions, and contact may reference a function of up
to four variables defined in the TABLE model definition option.
2. Boundary conditions and initial conditions are input in a consistent manner and are associated with a name.
They are not applied unless they have been activated by the LOADCASE model and history definition option.
3. Based upon the use of tables and loadcases, all boundary conditions may be defined in the model definition
section.
4. Mechanical boundary conditions are, in general, entered as total values at the end of the loadcase.

5. Boundary conditions can be applied to either finite element entities (such as nodes, elements, edges or faces)
or they may be applied to geometric entities (such as points, curves, or surfaces). The boundary conditions are
applied to the finite element based upon the current attach information. This is updated for local adaptive
meshing (2-D or 3-D) or global adaptive meshing in 2-D.
6. Each history definition section provides information on the type of analysis to be performed, the duration, and
the active boundary conditions. The LOADCASE history definition option is used to perform the latter.
Table Input
Many physical quantities cannot be represented by constants. This includes boundary conditions that are dependent
upon spatial location and/or time, material properties dependent upon temperature or material damage, contact
parameters such as the coefficient of friction and many more. In previous versions, this data could be input through a
variety of options including TEMPERATURE EFFECTS, WORK HARD, STRAIN RATE, by repetitive input of
boundary conditions, or, in the most general case, by user subroutines.
1 time
28 contact force
55 normalized arc distance
2 normalized time
29
contact body
56 distance to other contact surface
(near contact only)
3 increment number 30

n
(normal stress)
57 term of series
4 normalized increment 31 voltage 58 hydrostatic stress
5 x coordinate 32 current 59 hydrostatic strain
6 y coordinate
33
(see THROAT)
60
.
7 z coordinate
34 (pyrolysis damage). 61 .
8 s = 35 (water vapor fraction).
62 2nd state variable
9
angle
36
(coking damage).
63 3rd state variable
10 mode number 37 gasket closure distance 64 4th state variable
11 frequency 38 displacement magnitude 65 5th state variable
12 temperature 39 stress rate 66 loadcase number
13 function 40 experimental data 67 degree of cure
14 fourier 41 porosity 68 magnetic field intensity
15
(equivalent plastic strain)
42 void ratio 69 equivalent mechanical strain
16
(equivalent strain rate)
43
(equivalent creep strain rate)
70 1st strain invariant
17
(normalized
mass flow rate).
44 minor principal total strain 71 2nd strain invariant
18 arc length 45 distance from neutral axis
(-t/2, +t/2)
72 3rd strain invariant
19 relative density
(not available for shells)
46 normalized distance from neutral
axis (-1, +1)
73 any strain component
20
(equivalent stress)
47 local x-coordinate of layer point for
open or closed section beam
74 damage
F
M
current radius
radius of throat
------------------------------------
\ .
| |
2
B
g p ,
m
g p ,
m =
p
B
g w ,
m
g w ,
m =
x
2
y
2
z
2
+ +

w
c
B
g
m
m =
43
CHAPTER 3
Data Entry
Using the table driven input procedure, virtually all physical data can reference a table/function. Whenever necessary
in the analysis, the table is evaluated based upon the current value of the independent variables and multiplied with
the reference value. A table can have as many as four independent variables, as long as that independent variable is
physically meaningful, For example, requesting that the Youngs modulus is a function of the equivalent plastic strain
is considered a data error because elasticity does not support this behavior. The list of available independent variable
types is given in the following table.
The function can be input through the TABLE model definition option using either the piecewise linear mode or the
equation mode. In the piecewise linear mode, the result is interpolated between the given data values. If the
independent variable is outside of the range of data entered, the result will be either extrapolated or the last value
will be used.
In the second mode, the function may be entered as a mathematical equation.
A mathematical formula may be either 80 characters or 160 characters long if extended input format is used. The
formula is defined in terms of independent variables , , , and/or , where the meaning of those variables is
based on the variable type defined in the 3rd data block.
The evaluation is based upon usual mathematical standards moving from left to right with the conventional rules of
the use of parentheses. The following mathematical symbols/operations are available.
21 magnetic induction 48 local y-coordinate of layer point for
open or closed section beam
75 accumulated crack growth
22 velocity 49 1st isoparametric coordinate
(not available in this release)
76 relative sliding velocity
23 particle diameter 50 2nd isoparametric coordinate
(not available in this release)
-1 to
-100
parametric variable 1 to 100
24 x
0
coordinate 51 wavelength (used in spectral
radiation)
25 y
0
coordinate
52
creep strain
26 z
0
coordinate 53 pressure or primary quantity in
diffusion
27
s
0
=
54 equivalent strain rate for
nonNewtonian viscosity
+ addition
- subtraction
* multiplication
/ division
^ exponential
! factor
% mod
er
x
0
2
y
0
2
z
0
2
+ +
v
1
v
2
v
3
v
4

In addition to , , , and , the following constants may be used in the equation:
The following mathematical functions may be used in an equation:
pi
e exponent
tz offset temperature entered via the PARAMETERS model definition option
q Activation energy entered via MATERIAL DATA model definition option
r Universal gas constant entered via the PARAMETERS model definition option
sb Stefan Boltzman constant entered via the PARAMETERS model definition option
cos cosine (x) x in radians
sin sine (x) x in radians
tan tangent (x) x in radians
dcos cosine (x) x in degrees
x
1
x
2
x
3
x
4
v
1
v
1
Function
Allowing Extrapolation No Extrapolation
F
1
Data Entered
x
1
f
1
,
x
2
f
2
,
x
3
f
3
,
x
4
f
4
,
Given x
0
with extrapolation
f
0
f
1
f
1
f
2
x
1
x
2
------------------
\ .
| |
x
0
x
1
( ) + =
without extrapolation
f
0
f
1
=
v
1
v
2
v
3
v
4
45
CHAPTER 3
Data Entry
The equation is evaluated based upon the current value of the independent variable. It is your responsibility to make
sure that the equation may be evaluated for the potential values of the independent values. For example, if the function
is , then if , an error will occur.
dsin sine (x) x in degrees
dtan tangent (x) x in degrees
acos inverse cosine (x) f in radians
asin inverse sine (x) f in radians
atan inverse tangent (x) f in radians
atan2 inverse tangent (x,y) f in radians
dacos inverse cosine (x) f in degrees
dasin inverse sine (x) f in degrees
datan inverse tangent (x) f in degrees
datan2 inverse tangent (x,y) f in degrees
log log based 10
ln natural log
exp exponent
cosh hyperbolic cosine
sinh hyperbolic sine
tanh hyperbolic tangent
acosh inverse hyperbolic cosine
asinh inverse hyperbolic sine
atanh inverse hyperbolic tangent
sqrt square root
rad convert degrees to radians
deg convert radians to degrees
abs obtain absolute value
int truncates the value to whole
frac take the fractional value
max takes the maximal value
min takes the minimal value
mod return the remainder of x, based on y
mod(x,y) = x - y * int (x/y)
1 v
1
v
1
0.0 =

Parameters
This group of parameters allocates the necessary working space for the problem and sets up initial switches to control
the flow of the program through the desired analysis, This set of input must be terminated with an END parameter. The
input format for these parameters is described in Marc Volume C: Program Input.
Model Definition Options
This set of data options enters the initial loading, geometry, and material data of the model and provides nodal point
data, such as boundary conditions. Model definition options are also used to govern the error control and restart
capability. Model definition options can also specify print-out and postprocessing options. The data you enter on
model definition options provides the program with the necessary information for determining an initial elastic
solution (zero increment solution). When boundary conditions reference time dependent tables, the transient nonlinear
behavior can be defined. The transient period is defined in the history definition section. This group of options must
be terminated with the END OPTION option. The input format for these options is described in Marc Volume C:
Program Input.
History Definition Options
This group of options provides the load incrementation and controls the program after the initial elastic analysis.
History definition options also include blocks which allow changes in the initial model specifications. Each set of load
sets must be terminated with a CONTINUE option. This option requests that the program perform another increment
or series of increments if you request the auto-incrementation features. The input format for these options is described
in Marc Volume C: Program Input.
A typical input file setup for the Marc program is shown below.
Marc Parameter
Terminated by an END parameter
Marc Model Definition Options
(Zero Increment)
Terminated by an END OPTION option
Marc History Definition
Data for the First Increment
Terminated by a CONTINUE option
(Additional History Definition
Option for the second, third, ..., Increments)
Figure 3-1 is a dimensional representation of the Marc input data file.
47
CHAPTER 3
Data Entry
Figure 3-1 The Marc Input Data File
REZONE Option
When the REZONE option is inserted into the input file and the manual procedure is used, the program reads additional
data options to control the rezoning steps. These options must immediately follow the END OPTION option or a
CONTINUE history definition option.
You can select as many rezoning steps in one increment as you need. Every rezoning step is defined by the data,
starting with the REZONE option and ending with the CONTINUE option. The END REZONE option terminates the
complete set of rezoning steps that form a complete rezoning increment. Follow the rezoning input with normal history
definition data, or again by rezoning data. The input format for these options is described in Marc Volume C: Program
Input.
Load
Incrementation
Model
Definition
Parameter
Proportional
Increment
Auto Load
Etc.
Connectivity
Coordinates
Fixed Displacements
Etc.
Title
Sizing
Etc.
L
i
n
e
a
r

a
n
d

N
o
n
l
i
n
e
a
r

A
n
a
l
y
s
i
s
R
e
q
u
i
r
i
n
g

I
n
c
r
e
m
e
n
t
a
t
i
o
n
L
i
n
e
a
r

A
n
a
l
y
s
i
s
Chapter 4 Introduction to Mesh Definition
4
Introduction to Mesh Definition
Direct Input 49
User Subroutine Input 54
MESH2D 54
Mentat 59
FXORD Option 59
Incremental Mesh Generators 64
Bandwidth Optimization 65
Rezoning 66
Substructure 66
BEAM SECT Parameter 69
Error Analysis 73
Local Adaptivity 74
Automatic Global Remeshing 80

49
CHAPTER 4
This chapter describes the techniques for mesh definition available internally in Marc. Mesh definition is the process
of converting a physical problem into discrete geometric entities for the purpose of analysis. Before a body can
undergo finite element analysis, it must be modeled into discrete physical elements. An example of mesh definition is
shown in Figure 4-1.
Figure 4-1 Finite Element Mesh of a Brake Rotor (click figure to activate 3D content)
Mesh definition encompasses the placement of geometric coordinates and the grouping of nodes into elements. For
Marc to have a valid mesh definition, the nodes must have geometric coordinates and must be connected to an element.
First, describe the element by entering the element number, the element type, and the node numbers that make up the
element.
Next, enter the physical coordinates of the nodal points.
Direct Input
You must enter two types of data into Marc for direct mesh definition: connectivity data, which describes the nodal
points for each element, and coordinate data which gives the spatial coordinates of each nodal point. This section
describes how to enter this data.
Element Connectivity Data
You can enter connectivity data from either the input option file (FORTRAN unit 5) or from an auxiliary file. Several
blocks of connectivity can be input. For example, the program can read one block from tape and subsequently read a
block from the input option file. Each block must begin with the word CONNECTIVITY. In the case of duplicate
specification, Marc always uses the data that was input last for a particular element.
Enter the nodal points of two-dimensional elements in a counterclockwise order. Figure 4-2 illustrates correct and
incorrect numbering of element connectivity data.
Note: You do not need to enter element numbers and node numbers sequentially or consecutively.

Figure 4-2 Correct/Incorrect Numbering of Two-Dimensional Element Connectivity of 4-Node Elements
When there are eight nodal points on a two-dimensional element, number the corner nodes 1 through 4 in
counterclockwise order. The midside nodes 5 through 8 are subsequently numbered in counterclockwise order.
Figure 4-3 illustrates the correct numbering of element connectivity of 8-node elements.
Figure 4-3 Numbering of Two-Dimensional Element Connectivity for 8-Node Quadrilateral Elements
Lower-order triangular elements are numbered using the counterclockwise rule.
Figure 4-4 Numbering of 3-Node Triangular Element
Note that quadrilateral elements can be collapsed into triangular elements by repeating the last node.
The higher order triangular elements have six nodes, the corner nodes are numbered first in a counterclockwise
direction. The midside nodes 4 through 6 are subsequently numbered as shown in Figure 4-5.
Y,R
X,Z
Correct Numbering Incorrect Numbering
1
2
3
4
1
2 3
4
Y,R
X,Z
1 2
3 4
5
8 6
7
1 2
3 X,R
X,Z
51
CHAPTER 4
Figure 4-5 Numbering of 6-Node Triangular Element
Number three-dimensional elements in the same order as two-dimensional elements for each plane. Enter nodes for
an 8-node brick in counterclockwise order as viewed from inside the element. First, enter nodes comprising the base;
then enter ceiling nodes as shown in Figure 4-6.
Figure 4-6 Numbering of Element Connectivity for 8-Node Brick
A 20-node brick contains two 8-node planes and four nodes at the midpoints between the two planes. Nodes 1 through
4 are the corner nodes of one face, given in counterclockwise order as viewed from within the element. Nodes 5
through 8 are on the opposing face; nodes 9 through 12 are midside nodes on the first face, while nodes 13 through 16
are their opposing midside nodes. Finally, nodes 17 through 20 lie between the faces with node 17 between 1 and 5.
Figure 4-7 illustrates the numbering of element connectivity for a 20-node brick.
1 2
3
6 5
4
8 7
6 5
4 3
2 1
Z
Y
X

Figure 4-7 Numbering of Three-Dimensional Element Connectivity for 20-Node Brick
The four node tetrahedral is shown in Figure 4-8.
Figure 4-8 Numbering of Four-Node Tetrahedral
The ten-node tetrahedral is shown in Figure 4-9. The corner nodes 1-4 are numbered first. The first three midside
nodes occur on the first face. Nodes 8, 9, and 10 are between nodes 1 and 4, 2 and 4, and 3 and 4, respectively.
Figure 4-9 Numbering of 10-Node Tetrahedral
The 6-node pentahedral element is shown in Figure 4-10.
8 7
6 5
4 3
2 1
Z
Y
X
9
12
10
17 18
11
13
20 19
16
14
15
4
3
2 1
4
3
2 1
9
5
6
8
7
10
53
CHAPTER 4
Figure 4-10 6-Node Pentahedral
The 15-node pentahedral element is shown in Figure 4-11.
Figure 4-11 15-Node Pentahedral
Nodal Coordinate Data
You can enter nodal coordinates directly from the input option file (FORTRAN unit 5) or from an auxiliary file. You
can enter several blocks of nodal coordinate data in a file. In the case of duplicate specifications, the program uses data
entered last for a particular nodal point in the mesh definition.
Direct nodal input can be used to input local corrections to a previously generated set of coordinates. These options
give the modified nodal coordinates.
The CYLINDRICAL option can be used to transform coordinates given in a cylindrical system to a Cartesian system.
6
2
3
5
4
1
8
12
2 7
11
1
9
3
15
14
5
10
13
4
6

Activate/Deactivate
You have the ability to turn on and off elements using this option, which is useful when modeling ablation or
excavation. When you enter the mesh connectivity, the program assumes that all elements are to be included in the
analysis unless they are deactivated. This effectively removes this material from the model. These elements can be
reinstated later by using the ACTIVATE option. If the element is activated, one can select if the level of stress is to be
reinstated or set to zero. The use of these options results in nonlinear behavior and have an effect upon convergence.
User Subroutine Input
User subroutines can be used to generate or modify the data for mesh definition. User subroutine UFCONN generates
or modifies element connectivity data. The UFCONN model definition option activates this subroutine. The user
subroutine is called once for each element requested. Refer to Volume D: User Subroutines and Special Routines for
a description of the UFCONN user subroutine and instructions for its use.
The UFXORD user subroutine generates or modifies the nodal coordinates. The UFXORD model definition option
activates this subroutine. The user subroutine is called once for each node requested. Refer to Volume D: User
Subroutines and Special Routines for a description of the UFXORD user subroutine and instructions for its use.
MESH2D
MESH2D generates a mesh of quadrilateral or triangular elements for a two-dimensional body of any shape. The
generated mesh is written to a separate file and must be read with the CONNECTIVITY, COORDINATES, and FIXED
DISP, etc., options.
Block Definition
In MESH2D, a physical object or domain is divided into quadrilateral and/or triangular parts, called blocks.
Quadrilateral blocks are created by Marc by mapping with polynomials of the third order from a unit square. These
blocks can, therefore, be used to approximate curved boundaries. The geometry of a quadrilateral block is defined by
the coordinates on 12 nodes shown in Figure 4-12. If the interior nodes on an edge of the block are equal to zero or
are not specified, the edge of the block is straight.
Triangular blocks have straight edges. The geometry of a triangular block is defined by the coordinates of the three
vertices.
Note:
Requires the final coordinate data in terms of a single Cartesian system. Refer to Marc Volume B: Element
Library to determine the required coordinate data for a particular element type.
55
CHAPTER 4
Figure 4-12 Typical Quadrilateral Block Mapping
Merging of Nodes
Marc creates each block with a unique numbering scheme. The MERGE option fuses all nodes that lie within a small
circle, renumbers the nodes in sequence, and then removes all gaps in the numbering system. You can select which
blocks are to be merged together, or you can request that all blocks be merged. You must give the closeness distance
for which nodes will be merged.
Block Types
MESH2D generates two types of quadrilateral blocks. Block Type 1 is a quadrilateral block that is covered by a regular
grid. The program obtains this grid by dividing the block edge into M by N intervals. Figure 4-13 illustrates the
division of block edges into intervals with M = 4, N = 3.
The P
1
P
4
face of the block is the 1-4 face of triangular elements and the P
1
P
2
face of the block becomes the 1-2 face
of quadrilateral elements.
x , ( ) x P
i
( )
i
i 1 =
12
= , ( )
y , ( ) y P
i
( )
i
i 1 =
12
= , ( )
1 2
3 4
5 6
7
8
9 10
11
12
2
2
P
1
P
2
P
3
P
4
P
5
P
6
P
7
P
8
P
9
P
10
P
11
P
12
x
y

Figure 4-13 Block Type 1
Block Type 2 is a quadrilateral block that allows the transition of a coarse mesh to a finer one. In one direction, the
block is divided into M, 2M, 4M...; while in the other direction, the block is divided into N intervals. Figure 4-14
illustrates the division of Block Type 2 edges into intervals.
The P
1
P
2
face of the block becomes the 1-2 face of quadrilateral elements, and the P
2
P
3
face of the block is the 2-3
face of triangular elements.
Block Type 3 is a triangular block. The program obtains the mesh for this block by dividing each side into N equal
intervals. Figure 4-15 illustrates Block Type 3 for triangular and quadrilateral elements.
Block Type 4 is a refine operation about a single node of a block (Figure 4-16). The values of N and M are not used.
If quadrilateral elements are used in a triangular block, the element near the P
2
P
3
face of the block is collapsed by
MESH2D in every row. The P
1
P
2
face of the block is the 1-2 face of the generated elements.
17
24
9
2
1 3 6 7
4 5 8
16
7 8 9
1 2 3 4 5
10
15
6
11
16 17 18 19 20
12 13 14
P
2
P
3
P
1
P
4
2/3
2/3
2/3
N = 3
1/2 1/2 1/2 1/2
M = 4
9 12
2/3
2/3
2/3
N = 3
1/2 1/2 1/2 1/2
M = 4
16 17 18 19 20
11 12 13 14 15
10 6 7 8 9
1 2 3 4 5
11 10
5 6 7 8
1 2 3 4
Triangular Quadrilateral
57
CHAPTER 4
8
7
6
5
4 3
2
1
1 2 3
4
5 6 7
8
9
10 11 12 13 14 15 16
17
18
19
34
2/7
4/7
3/7
M = 2
N = 3
P
4
x
P
3
P
2
P
1
1 1
Triangular
1 1
Quadrilateral
6
5
4 3
2
1
1/2
1 2 3
4 5 6
7 8 9 10 11
2
x
1/2 1/2 1/2
6
8
7
4 3
2
1
16
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
P
1
N = 4
Triangular Quadrilateral
P
2
P
3
5
4
P
2
10
P
3
1
P
1
2 3
5
6
7
8 9
4
3 2 1
1 2
3 4
1 2
7 5
4
6
3

Symmetry, Weighting, and Constraints
MESH2D contains several features that facilitate the generation of a mesh: use of symmetries, generation of weighted
meshes, and constraints. These features are discussed below.
MESH2D can use symmetries in physical bodies during block generation. An axis of symmetry is defined by the
coordinates of one nodal point, and the component of a vector on the axis. One block can be reflected across many
axes to form the domain. Figure 4-17 illustrates the symmetry features of MESH2D.
Figure 4-17 Symmetry Option Example
A weighted mesh is generated by the program by spacing the two intermediate points along the length of a boundary.
This technique biases the mesh in a way that is similar to the weighting of the boundary points. This is performed
according to the third order isoparametric mapping function.
The CONSTRAINT option generates boundary condition restraints for a particular degree of freedom for all nodes on
one side of a block. The option then writes the constraints into the file after it writes the coordinate data. The FIXED
DISP, etc., option must be used to read the boundary conditions generated from the file.
Note:
If a weighted mesh is to be generated, be cautious not to move the interior boundary points excessively. If
the points are moved more than 1/6 of the block length from the 1/3 positions, the generated elements can
turn inside-out.
1
Original Block
1
One Symmetry Axis
2
1
Two Symmetry Axes
2
4
3
1
Three Symmetry Axes
2
4
3
8
7
5
6
59
CHAPTER 4
Additional Options
Occasionally, you might want to position nodes at specific locations. The coordinates of these nodes are entered
explicitly and substituted for the coordinates calculated by the program. This is performed using the SPECIFIED
NODES option. Some additional options in MESH2D are:
MESH2D can be used several times within one input file.
The START NUMBER option gives starting node and element numbers.
The CONNECT option allows forced connections and/or disconnections with other blocks. This option is
useful when the final mesh has cracks, tying, or gaps between two parts.
The MANY TYPES option specifies different element types.
Mentat
Mentat is an interactive program which facilitates mesh definition by generating element connectivity and nodal
coordinates. Some of the Mentat capabilities relevant to mesh generation are listed below.
Prompts you for connectivity information and nodal coordinates. Accepts input from a keyboard or mouse.
Accepts coordinates in several coordinate systems (Cartesian, cylindrical, or spherical).
Translates and rotates (partial) meshes.
Combines several pre-formulated meshes.
Duplicates a mesh to a different physical location.
Generates a mirror image of a mesh.
Subdivides a mesh into a finer mesh.
Automatic mesh generation in two- and three-dimensions.
Imports geometric and finite element data from CAD systems.
Smooths nodal point coordinates to form a regular mesh
Converts geometric surfaces to meshes.
Refines a mesh about a point or line.
Expands line mesh into a surface mesh, or a surface mesh into a solid mesh.
Calculates the intersection of meshes.
Maps nodal point coordinates onto prescribed surfaces.
Writes input data file for connectivity and coordinates in Marc format for use in future analyses.
Apply boundary conditions to nodes and elements.
Define material properties.
Submit Marc jobs.
FXORD Option
The FXORD model definition option (Volume C: Program Input) generates doubly curved shell elements of element
type 4, 8, or 24 for the geometries most frequently found in shell analysis. Since the mathematical form of the surface

is well-defined, the program can generate the 11 or 14 nodal coordinates needed by element type 8, 24, or 4 to fit a
doubly curved surface from a reduced set of coordinates. For example, you can generate an axisymmetric shell by
entering only four coordinates per node. The FXORD option automatically generates the complete set of coordinates
required by the elements in the program from the mathematical form of the surface.
A rotation and translation option is available for all components of the surface to give complete generality to the
surface generation. The input to FXORD consists of the reduced set of coordinates given in a local coordinate system
and a set of coordinates which orient the local system with respect to the global system used in the analysis. The
program uses these two sets of coordinates to generate a structure made up of several shell components for analysis.
The FXORD option allows for the generation of several types of geometries. Because you may need to analyze shells
with well-defined surfaces not available in this option, you can use the UFXORD user subroutine to perform your own
coordinate generation (Volume D: User Subroutines and Special Routines). The FXORD option can also be used to
convert cylindrical coordinates or spherical coordinates to Cartesian coordinates for continuum elements.
Major Classes of the FXORD Option
The following cases are considered:
Shallow Shell (Type I)
Axisymmetric Shell (Type 2)
Cylindrical Shell Panel (Type 3)
Circular Cylinder (Type 4)
Plate (Type 5)
Curved Circular Cylinder (Type 6)
Convert Cylindrical to Cartesian (Type 7)
Convert Spherical to Cartesian (Type 8)
Shallow Shell (Type I)
Type 1 is a shallow shell with
(4-1)
The middle surface of Figure 4-18 (Type 1) is defined by an equation of the form
(4-2)
and the surface is determined when the following information is given at each node.
(4-3)
The last coordinate is only necessary for Element Type 4.
1
x
1

2
, x
2
= =
x
3
x
3
x
1
x ,
2
( ) =
x
1
x
2
x
3
x
3
x
1
---------
x
3
x
2
---------

2
x
3
x
1
x
2
------------------ , , , , ,
61
CHAPTER 4
Figure 4-18 Classification of Shells
Axisymmetric Shell (Type 2)
The middle surface symmetric to the axis (Figure 4-18, Type 2) is defined as:
(4-4)
where and are the angles shown in Figure 4-18. In this case, the surface is defined by
(4-5)
The angles and are given in degrees.
X
3
R
f
q
Type 2
X
2
Type 1
X
1
X
2
X
3
X
3
X
1
X
2
X
3
X
1
X
2
X
3
Type 3
Type 4
3
R
x
3
x
1
R ( ) cos cos =
x
2
R ( ) sin cos =
x
3
R ( ) sin =

R
dR
d
------- , , ,


Cylindrical Shell Panel (Type 3)
The middle surface is the cylinder defined by Figure 4-18.
(4-6)
The nodal geometric data required is
(4-7)
Circular Cylinder (Type 4)
This is the particular case of Type 3 where the curve
(4-8)
is the circle given by Figure 4-18 (Type 4).
(4-9)
The only nodal information is now
(4-10)
Note that is given in degrees and, because is constant, it needs to be given for the first nodal point only.
Plate (Type 5)
The shell is degenerated into the plate
(4-11)
The data is reduced to
(4-12)
x
1
x
1
s ( ) =
x
2
x
2
s ( ) =
x
3
x
3
=
s x
3
x
1
x
2
dx
1
ds
---------
dx
2
ds
--------- , , , , ,
x
1
s ( ) x
2
, s ( )
x
1
R cos =
x
2
R sin =
x
3
R , ,
R
x
3
0 =
x
1
x
2
,
63
CHAPTER 4
Curved Circular Cylinder (Type 6)
Figure 4-19 illustrates this type of geometry.
Figure 4-19 Curved Circular Cylinder
The middle surface of the shell is defined by the equations
(4-13)
The Gaussian coordinates on the surface are
(4-14)
and form an orthonormal coordinate system. The nodal point information is
(4-15)
and in degrees. You need to specify the radii and only for the first nodal point.
Shell Middle Surface
Type 6
X
1
R
q
X
2
X
3
q
1
f
q
2
x
1
r cos =
x
2
r R 1 cos ( ) + cos sin =
x
3
R r sin r sin + ( ) sin =
1
r =
2
R =
r R , , ,
r R

Convert Cylindrical to Cartesian (Type 7)
Type 7 allows you to enter the coordinates for continuum elements in cylindrical coordinates, which are converted by
Marc to Cartesian coordinates. In this way, you can enter , , and obtain x, y, z where is given in degrees and
(4-16)
Convert Spherical to Cartesian (Type 8)
Type 8 allows you to enter the coordinates for continuum elements in spherical coordinates, which are converted by
Marc to Cartesian coordinates. In this way, you can enter , , and obtain x, y, z where and are given in
degrees and
(4-17)
Recommendations on Use of the FXORD Option
When a continuous surface has a line of discontinuity, for example, a complete cylinder at , you
must place two nodes at each nodal location on the line to allow the distinct coordinate to be input. You must use tying
element type 100 to join the degrees of freedom. Generally, when different surfaces come together, you must use the
intersecting shell tyings.
The FXORD option cannot precede the COORDINATES option, because it uses input from that option.
Incremental Mesh Generators
Incremental mesh generators are a collection of options available in Marc to assist you in generating the mesh.
Incremental mesh generators generate connectivity lists by repeating patterns and generate nodal coordinates by
interpolation. Use these options directly during the model definition phase of the input.
During the model definition phase, you can often divide the structure into regions, or blocks, for which a particular
mesh pattern can be easily generated. This mesh pattern is established for each region and is associated with a single
element connectivity list. Use the CONNECTIVITY option to input this element connectivity list. The incremental mesh
generators then generate the remainder of the connectivity lists.
Critical nodes define the outline of the regions to be analyzed. Use the COORDINATES option to enter the critical
nodes. The incremental mesh generators complete the rest and join the regions by merging nodes.
A special connectivity interpolator option generates midside nodes for elements where these nodes have not been
specified in the original connectivity. A separate mesh generation run is sometimes required to determine the position
of these nodes. This run can be followed by mesh display plotting.
The incremental mesh generators are listed below:
Element Connectivity Generator The CONN GENER option repeats the pattern of the connectivity data for
previously defined master elements. One element can be removed for each series of elements, allowing the
R Z
x R cos =
y R sin =
z Z =
R
x R sin cos =
y R cos sin =
z R cos =
0 360 = =
65
CHAPTER 4
program to generate a tapered mesh. Two elements can be removed for each series with triangular elements.
Element Connectivity Interpolator The CONN FILL option completes the connectivity list by generating
midside nodes. You first generate the simpler quadrilateral or brick elements without the midside nodes. You
can then fill in the midside nodes with this option.
Coordinate Generator The NODE GENER option creates a new set of nodes by copying the spacing of
another specified set of nodes.
Coordinate Interpolator The NODE FILL generates intermediate nodes on a line defined by two end nodes.
The spaces between the nodes can be varied according to a geometric progression.
Coordinate Generation for Circular Arcs The NODE CIRCLE option generates the coordinates for a series of
nodes which lie on a circular arc.
Bandwidth Optimization
Marc can minimize the nodal bandwidth of a structure in several ways. The amount of storage is directly related to the
size of the bandwidth, and the computation time increases in proportion to the square of the average bandwidth.
The OPTIMIZE option allows you to choose from several bandwidth optimization algorithms. The minimum degree
algorithm should only be used if the direct sparse solver is used.
The four available OPTIMIZE options are listed in Table 4-1.
The nodal correspondence obtained through this process can be saved and then used in subsequent analyses. This
eliminates the need to go through the optimization step in later analyses. The correspondence is used to relate the
user-defined node (external) numbers to the program-optimized (internal) node numbers and vice versa.
Note:
This option creates an internal node numbering that is different from your node numbering. Use your node
numbering for all inputs. All output appears with your node numbering. The occurrence of gap or
Herrmann elements can change the internal node numbers. On occasion, this change can result in a non
optimal node numbering system, but this system is necessary for successful solutions.
Table 4-1 Bandwidth Optimization Options
Option
Number Remarks Solver
2 Cuthill-McKee algorithm Profile (0)
9 Sloan (Recommended) Profile (0)
10 Minimum Degree Algorithm Sparse Direct (4) or Multifrontal (8)
11 METIS (Recommended)* Multifrontal (8)
*Reference A Fast and Highly Quality Multilevel Scheme for Partitioning Irregular
Graphs, George Karypis and Vipin Kumar, SIAM Journal on Scientific Computing, Vol.
20, No. 1, pp. 359-392, 1999. METIS documentation may be downloaded from
http://glaros.dtc.umn.edu/gkhome/views/metis

Rezoning
The REZONING parameter defines a new mesh and transfers the state of the old mesh to the new mesh. Elements or
nodes can be either added to or subtracted from the new mesh. This procedure requires a sub increment to perform the
definition of the new mesh. The rezoning capability can be used for two- and three-dimensional continuum elements
and for shell elements 22, 75, 138, 139, and 140. See Figure 4-20 for an example of rezoning.
Rezoning or remeshing can be carried automatically using the mesh generators. This feature is described in detail in
the Automatic Global Remeshing section.
Figure 4-20 Mesh Rezoning
Substructure
Marc is capable of multilevel substructuring that includes:
Generation of superelements
Use of superelements in subsequent Marc analyses
Recovery of solutions (displacements, stresses, and strains) in the individual substructures
The Marc multilevel procedure allows superelements to be used. One self-descriptive database stores all data needed
during the complete analysis. You only have to ensure this database is saved after every step of the analysis.
The advantages of substructuring are the following:
Separates linear and nonlinear parts of the model
Allows repetition of symmetrical or identical parts of the model for linear elastic analysis
Separates large models into multiple, moderate-size models
Separates fixed model parts from parts of the model that may undergo design changes
A disadvantage of substructuring is the large amount of data that must be stored on the database.
Three steps are involved in a substructuring run.
The superelement generation step is done for every superelement at a certain level.
The use of superelements in subsequent Marc runs is done at the highest level, or is incorporated into Step 1 for
the intermediate levels.
Recovery of solutions within a certain superelement can or cannot be done for every superelement.
Before Rezoning After Rezoning
67
CHAPTER 4
Substructuring in Marc is only possible for static analysis. Nonlinearities are not allowed with a superelement. You,
as a user, must ensure that nonlinearities are not present.
The maximum number of levels in a complete analysis is 26. The maximum number of substructures in the complete
analysis is 676.
Each step can be done in an individual run, or an unlimited number of steps can be combined into a single run.
During superelement generation in Marc, you can generate a complete new superelement or you can copy a previously
defined superelement with identical or newly defined external load conditions. Any number of superelements can be
formed in a generation run.
Marc offers flexibility in the use of superelements by allowing rotation or mirroring of a superelement. If a run is
nonlinear, superelements are treated as linear elastic parts. At every increment, you can perform a detailed analysis of
certain substructures by descending down to the desired superelements. Use the normal Marc control algorithm (AUTO
INCREMENT, AUTO LOAD, PROPORTIONAL INCREMENT) to control the load on the superelements.
Technical Background
The system of equations for a linear static structure is
(4-18)
When local degrees of freedom (subscripted ) and external degrees of freedom (subscripted ) are considered, this
can be rewritten as
(4-19)
To obtain both the stiffness matrix and the load vector of the substructure, it is necessary to eliminate and rewrite
the above system with as the only unknown splitting the above equation.
(4-20)
The first equation can be written as
(4-21)
Substituting this equation into the second
(4-22)
Ku P =
l e
K
l l
K
el
K
l e
K
ee
u
l
u
e \ .
|
|
| |
P
l
P
e \ .
|
|
| |
=
u
1
u
e
K
l l
u
l
K
el
u
e
+ P
l
=
K
l e
u
l
K
ee
u
e
+ P
e
=
and
u
l
K
l l
1
= K
el
u
e
K
l l
1
P
l
+
K
l e
K
l l
1
K
el
u
e
K
l e
K
l l
1
P
l
K
ee
u
e
+ + P
e
=

This can be rewritten as
(4-23)
where
(4-24)
and
(4-25)
and are solved by the triangularization of , the forward and backward substitution of and ,
respectively, and premultiplication with , , and are used in the next part of the analysis with other
substructures or with another element mesh. That analysis results in the calculation of u
e
.
You can now calculate the local degrees of freedom and/or the stresses using the following procedure:
(4-26)
which can be written as
(4-27)
where
(4-28)
The displacement of the substructure is, therefore, known, and stresses and strains can be calculated in the normal way.
Scaling Element Stiffness
Occasionally, it is desirable to perform a scalar multiplication of the stiffness, mass, and load matrix to represent a
selective duplication of the finite element mesh.
The STIFSCALE option can be used to enter the scaling factor for each element. In this case, the global stiffness, mass,
and load matrices are formed as follows:
(4-29)
Note that no transformation of the stiffness matrix occurs and that point loads are not scaled.
K
ee
*
u
e
P
e
*
=
K
ee
*
K
ee
K
l e
K
l l
1
K
el
=
P
e
*
P
e
K
l e
K
l l
1
P
l
=
K
ee
*
P
e
*
K
l l
K
el
P
l
K
l e
K
ee
*
P
e
*
K
l l
u
l
P
l
K
el
u
e
=
u
l
K
l l
1
P
l
*
=
P
l
*
P
e
K
el
u
e
=
K
g
s
i
K
i
el
M
g
, s
i
M
i
el
and F
g
, s
i
f
i
el
= = =
69
CHAPTER 4
4
BEAM SECT Parameter
The BEAM SECT parameter inputs data to define the sectional properties for three-dimensional beam elements.
Include this option if you are using:
element types 13, 77, or 79
element types 14, 25, 76, or 78 with a noncircular section
element types 52 or 98 and torsional and shear stiffness must be defined independently
element types 52 or 98 with numerical integrated solid cross-section.
The convention adopted for the local (beam) coordinate system is: the first and second director (local X and Y) at a
point are normal to the beam axis; the third director (local Z) is tangent to the beam axis and is in the direction of
increasing distances along the beam. The director set forms a right-handed system.
Orientation of the Section in Space
The beam axis in an element is interpolated from the two nodes of the element.
(4-30)
where the last three coordinates are only used for element 13.
The beam section orientation in an element is defined by the direction of the first director (local X) at a point, and this
direction is specified via the coordinates of an additional node or through the GEOMETRY option (see Marc Volume
C: Program Input).
Definition of the Section
You can include any number of different beam sections in any problem. Data options following the BEAM SECT
parameter of the Marc input (see Marc Volume C: Program Input) define each section. The program numbers the
sections in the order they are entered. To use a particular section for a beam element, set EGEOM2 (GEOMETRY
option, 3rd data block, columns 11-20) to the floating-point value of the section number, for example, 1, 2, or 3. The
program uses the default circular section for the closed section beam elements (14, 25, 76, 78) if EGEOM1 is nonzero.
The program uses the default solid rectangular cross section for elements 52 or 98 if EGEOM1 is nonzero. Figure 4-21
shows how the thin-walled section is defined using the input data.
x y z
dx
ds
------
dy
ds
------
dz
ds
------ , , , , ,

Figure 4-21 Beam Section Definition Examples for Thin-walled Sections
The rules and conventions for defining a thin-walled section are listed below:
1. An coordinate system defines the section, with the first director at a point of the beam. The origin
of the system represents the location of the node with respect to the section.
2. Enter the section as a series of branches. Branches can have different geometries, but they must form a complete
traverse of the section in the input sequence so that the endpoint of one branch is the start of the next branch.
It is often necessary for the traverse of the section to double back on itself. To cause the traverse to do this,
specify a branch with zero thickness.
Branch Divisions Thickness
1-2 8 0.5
2-3 4 0.5
3-4 8 0.5
4-5 4 0.5
5-6 8 0.5
1-2 10 1.0
2-3 10 1.0
3-4 10 1.0
Section 1
1 2
3
4 5 6
10
2
0
x
l
y
l
1 2
3 4
8
8
1
6
x
l
y
l
Section 2
1-2 8 1.0
2-3 4 0.0
3-4 8 0.3
4-5 4 0.0
5-6 8 1.0
60
2
3 4
5
1 6
R=5
17.85
1
0
x
l
y
l
Section 3
x
1
y
1
x
1
x
1
y
1
71
CHAPTER 4
3. You must divide each branch into segments. The stress points of the section are the branch division points. The
stress points are the points used for numerical integration of a sections stiffness and for output for stress results.
Branch endpoints are always stress points. There must always be an even number of divisions (nonzero) in any
branch. Not counting branches of zero thickness, you can use a maximum of 31 stress points (30 divisions) in
a complete section.
4. Branch thickness varies linearly between the values given for branch endpoint thickness. The thickness can be
discontinuous between branches. A branch is assumed to be of constant thickness equal to the thickness given
at the beginning of the branch if the thickness at the end of the branch is given as an exact zero.
5. The shape of a branch is interpolated as a cubic based on the values of and and their directions, in
relation to distance along the branch. The data is input at the two ends of the branch. If both and
are given as exact zeros at both ends of the branch, the branch is assumed to be straight. The section
can have a discontinuous slope at the branch ends. The beginning point of one branch must coincide with the
endpoint of the previous branch. As a result, and for the beginning of a branch need to be given only
for the first branch of a section.
6. Stress points are merged into one point if they are separated by a distance less than , where is a thickness
at one of these points.
Figure 4-21 shows three sections of a beam. Notice the use of zero thickness branches in the traverse of the I section.
The program provides the following data: the location of each stress point in the section, the thickness at that point,
the weight associated with each point (for numerical integration of the section stiffness), and the warping function at
each section.
Solid sections can be defined by using one of the standard sections and specifying its typical dimensions or by using
quadrilateral segments and specifying the coordinates of the four corner points of each quadrilateral segment in the
section. The local axes of a standard section are always the symmetry axes. For the standard sections, the symmetry
axes are also the principal axes. The principal axes for each section are shown in Figure 4-22. For the more general
sections, it is not required to enter the coordinates of the corner points of the quadrilateral segments with respect to the
principal axes. The section is automatically realigned and you can make specifications about how this realignment
should be carried out. The details about these orientation methods are described in the Cross-section Orientation and
Location of the Local-section Axis sections in Marc Volume B: Element Library.
Figure 4-22 shows the standard solid sections, the dimensions needed to specify them, and the orientation of their local
axes. If the dimension b is omitted for an elliptical section, the section will be circular. If the dimension b is omitted
for a rectangular section, the section will be square. If the dimension c is omitted for a trapezoidal or a hexagonal
section, the section will degenerate to a triangle or a diamond. For the elliptical section, the center point is the first
integration point. The second is located on the negative y-axis and the points are numbered radially outward and then
counterclockwise. For the rectangular, trapezoidal and hexagonal sections the first integration point is nearest to the
lower left corner and the last integration point is nearest to the upper right corner. They are numbered from left to right
and then from bottom to top. The default integration scheme for all standard solid sections uses 25 integration points.
x
1
y
1
dx
1
ds
dy
1
ds
x
1
y
1
t 10 t

Figure 4-22 Standard Solid Section Types with Typical Dimensions
The rules and conventions for defining a general section solid are listed below:
1. Each quadrilateral segment is defined by entering the coordinates of the four corner points in the local xy-plane.
The corner points must be entered in a counterclockwise sense.
2. The quadrilateral segments must describe a simply connected region; i.e., there must be no holes in the section,
and the section cannot be composed of unconnected islands and should not be connected through a single
point.
3. The quadrilateral segments do not have to match at the corners. Internally, the edges must, at least, partly
match.
4. The segments must not overlap each other (i.e., the sum of the areas of each segment must equal the total area
of the segment).
5. Each section can define its own integration scheme and this scheme applies to all segments in the section.
6. A section cannot have more than 100 integration points in total. This limits each section to, at most, 100
segments that, by necessity, will use single point integration. If less than 100 segments are present in a section,
they may use higher order integration schemes, as long as the total number of points in the section does not
exceed 100.
7. Integration points on matching edges are not merged, should they coincide.
This method allows you to define an arbitrary solid section in a versatile way, but you must make sure that the
conditions of simply connectedness and nonoverlap are met. The assumptions underlying the solid sections are not
accurate enough to model sections that are thin-walled in nature. It is not recommended to use them as an alternative
x
y
a
b
Elliptical Section
Circular (set b = 0)
x
y
a
b
Rectangular Section
Square (set b = 0)
x
y
a
b
Hexagonal Section
x
y
a
c
b
[2(a+c)b]/[3(a+c)]
Trapezoidal Section
c
73
CHAPTER 4
to thin-walled sections. Solid sections do not account for any warping of the section and, therefore, overestimate the
torsion stiffness as it is predicted by the Saint- Venant theory of torsion.
The first integration point in each quadrilateral segment is nearest to its first corner point and the last integration point
is nearest to its third corner point. They are numbered going from first to second corner point and then from second to
third corner point. The segments are numbered consecutively in the order in which they have been entered and the first
integration point number in a new segment simply continues from the highest number in the previous segment. There
is no special order requirement for the quadrilateral segments. They must only meet the previously outlined geometric
requirements. For each section, the program provides the location of the integration points with respect to the principal
axes and the integration weight factors of each point. Furthermore, it computes the coordinates of the center of gravity
and the principal directions with respect to the input coordinate system. It also computes the area and the principal
second moments of area of the cross section. For each solid section, all of this information is written to the output file.
All solid sections can be used in a pre-integrated fashion. In that case, the area A and the principal moments of area
I
xx
and I
yy
are calculated by numerical integration. The torsional stiffness is always computed from the polar moment
of inertia J=I
xx
+I
yy
. The shear areas are set equal to the section area A. The stiffness behavior may be altered by
employing any of the stiffness factors. In a pre-integrated section, no layer information is computed and is, therefore,
not available for output. Pre-integrated sections can only use elastic material behavior and cannot account for any
inelasticity (e.g. plasticity). For pre-integrated sections there is no limit on the number of segments to define the
section.
Figure 4-23 shows a general solid section built up from three quadrilateral segments using a 2 x 2 Gauss integration
scheme. It also shows the numbering conventions adopted in a quadrilateral segment and its mapping onto the
parametric space. For each segment in the example, the first corner point is the lower-left corner and the corner points
have been entered in a counterclockwise sense. The segment numbers and the resulting numbering of the integration
points are shown in the figure. Gauss integration schemes do not provide integration points in the corner points of the
section. If this is desired, you can choose other integration schemes like Simpson or Newton-Cotes schemes. In
general, a 2 x 2 Gauss or a 3 x 3 Simpson scheme in each segment suffices to guarantee exact section integration of
linear elastic behavior.
Figure 4-23 Beam Section Definition for Solid Section plus Numbering Conventions Adopted
Error Analysis
You can determine the quality of the analysis by using the ERROR ESTIMATE option.
The ERROR ESTIMATE option can be used to determine the mesh quality (aspect ratio, and skewness), and how they
change with deformation. While all of the Marc elements satisfy the patch test, the accuracy of the solution
often depends on having regular elements. In analyses where the updated Lagrangian method is used, the mesh often
1
5 6
7
9
11 12
10
8
2
3 4
1
2
3
y
x
(x
2
,y
2
)
(x
3
,y
3
)
(x
4
,y
4
)
(x
1
,y
1
)
4 3
2 1
(1,1) (-1,1)
(1,-1) (-1,-1)

becomes highly distorted during the deformation process. This option tells you when it would be beneficial to perform
a rezoning step.
This option can also be used to examine the stress discontinuity in the analysis. This is a measure of the meshes ability
to represent the stress gradients in the problem. Large stress discontinuities are an indication that the mesh is not of
sufficient quality. This can be resolved by increasing the number of elements or choosing a higher order element or
using local adaptive meshing.
Local Adaptivity
The adaptive mesh generation capability increases the number of elements and nodes to improve the accuracy of the
solution. The capability is applicable for both linear elastic analysis and for nonlinear analysis. The capability can be
used for lower-order elements, 3-node triangular solids and shells, 4-node quadrilateral solids and shells, 4-node
tetrahedrals, 8-node hexahedral elements, and the 8-node solid-shell elements. Local adaptive meshing cannot be used
with the triangular and tetrahedral Herrmann elements, type 155 - 157.
When used in conjunction with the ELASTIC parameter for linear analysis, a steady-state heat transfer, electrostatic,
or magnetostatic, the mesh is adapted and the analysis repeated until the adaptive criteria is satisfied. When used in a
nonlinear analysis, an increment is performed. If necessary, this increment is followed by a mesh adjustment which is
followed by the analysis of the next increment in time. While this can result in some error, as long as the mesh is not
overly coarse, it should be adequate.
Local adaptivity is available in a parallel analysis. In this case, the new elements remain in the same domain as the
parent element.
Number of Elements Created
The adaptive meshing procedure works by dividing an element and internally tying nodes to insure compatibility.
Figure 4-24 shows the process for a single quadrilateral element.
Figure 4-24 Single Quadrilateral Element Process
A similar process occurs for the triangles, tetrahedrons, and hexahedrons elements. You can observe that for
quadrilaterals the number of elements expands by four with each subdivision; similarly, the number of elements
increases by eight for hexahedrals. For the 8-node solid-shell element, the subdivision is similar to 4-node
quadrilaterals; there is no refinement through the thickness. If full refinement occurs, you observe that the number of
elements is for quadrilaterals and for hexahedrons elements. The number of levels also limits the
amount of subdivisions that may occur.
Original Element Level 1 Refinement Level 2 Refinement Level 3 Refinement
2
level x2 ( )
2
level x3 ( )
75
CHAPTER 4
For this reason, it is felt that the number of levels should, in general, be limited to three.
When adaptive meshing occurs, you can observe that discontinuities are created in the mesh as shown below:
To ensure compatibility node B is effectively tied to nodes A and C, and node D is effectively tied to nodes C and E.
All of this occurs internally and does not conflict with other user-defined ties or contact.
Boundary Conditions
When mesh refinement occurs, boundary conditions are automatically adjusted to reflect the change in mesh. The rules
listed below are followed:
1. Fixed Displacement
For both 2-D and 3-D, if both corner nodes on an edge have identical boundary conditions, the new node
created on that edge has the same boundary conditions. For 3-D, if all four nodes on a face have identical
boundary conditions, the new node created in the center of the face has the same boundary conditions. Note
that identical here means the same in the first degree of freedom, second degree of freedom, etc. independently
of one another.
2. Point Loads
The point loads remain unchanged on the original node number.
3. Distributed Loads
Distributed loads are automatically placed on the new elements. Caution should be used when using the
FORCEM user subroutine as the element numbers can be changed due to the new mesh process.
Level
Number of Elements
Quadrilaterals Hexahedrals
0 1 1
1 4 8
2 16 64
3 64 512
4 256 4096
A
B C
D
E

4. Contact
The new nodes generated on the exterior of a body are automatically treated as potential contact nodes. The
elements in a deformable body are expanded to include the new elements created. After the new mesh is
created, the new nodes are checked to determine if they are in contact. It should be noted that new nodes on
shells that are completely tied to the corresponding edge nodes are not checked for contact.
Location of New Nodes
When an element is refined, the default is that the new node on an edge is midside to the two corner nodes. As an
alternative, the POINTS, CURVES, SURFACES, ATTACH EDGE and ATTACH FACE options or the UCOORD user
subroutine can be used. The CURVES and SURFACES options can be used to describe the mathematical form of the
curve or surface. If the corner nodes of an edge are attached to the surface, the new node is placed upon the actual
surface. This option to attach new nodes is available both in a linear as well as a nonlinear analysis.
This is illustrated in Figure 4-25 and Figure 4-26, where initially a single element is used to represent a circle. The
circle is defined with the CURVES option and the original four nodes are placed on it using the ATTACH EDGE option.
Notice that the new nodes are placed on the circle.
Figure 4-25 Original Mesh and Surface
Caution: None of the nodes of an element being subdivided should have a local coordinate system defined
through the TRANSFORMATION or COORD SYSTEM option.
77
CHAPTER 4
Figure 4-26 Levels of Refinement
Adaptive Criteria
The adaptive meshing subdivision occurs when a particular adaptive criterion is satisfied. Multiple adaptive criteria
can be selected using the ADAPTIVE model definition option. These include:
Mean Strain Energy Criterion
The element is refined if the strain energy of the element is greater than the average strain energy in a chosen set of
elements times a given factor, .
(4-31)
Zienkiewicz-Zhu Criterion
The error norm is defined as either
(4-32)
The stress error and strain energy errors are
and (4-33)
Level 1 Refinement Level 2 Refinement
Level 2 and 3 Refinement
f
1
element strain energy
total strain energy
number of elements
----------------------------------------------- * f
1
>
( )
2
dV
}
2
dV
( )
2
dV
}
+
}
------------------------------------------------------------- =
2
E
E ( )
2
dV
}
E
2
dV E
E ( )
2
dV
}
+
}
------------------------------------------------------------- =
X
( )
2
dV
}
= Y E
E ( )
2
dV
}
=

where is the smoothed stress and is the calculated stress. Similarly, is for energy.
An element is subdivided if
and (4-34)
(4-35)
or
and (4-36)
(4-37)
where NUMEL is the number of elements in the mesh. If , , , and are input as zero, then .
Zienkiewicz Zhu Plastic Strain Criterion
The plastic strain error norm is defined as .
The plastic strain error is . The allowable element plastic strain error is
. The element will be subdivided when and
. NUMEL is the number of elements in the mesh.
Zienkiewicz-Zhu Creep Strain Criterion
Zienkiewicz-Zhu creep strain error norm is defined as . The creep strain error is
. The allowable element creep strain error is
.
The element will be subdivided when and . NUMEL is the number of elements in the mesh.
Equivalent Values Criterion
This method is based upon either relative or absolute testing using either the equivalent von Mises stress, the
equivalent strain, equivalent plastic strain or equivalent creep strain. An element is subdivided if the current element
value is a given fraction of the maximum (relative) or above a given absolute value.
* E
f
1
>
X
el
f
2
* X/NUMEL f
3
* X * f
1
NUMEL + >
f
1
>
Y
el
f
4
* Y/NUMEL f
5
* Y * f
1
NUMEL + >
f
2
f
3
f
4
f
5
f
2
1.0 =
2

p
*

p
( )
2
dV
}
p
2
dV
p
*
p
( )
2
dV
}
+
}
--------------------------------------------------------------- =
A
p
*
p
( )
2
dV
}
=
AEPS f
2
* A NUMEL f
3
* A * f
1
NUMEL + = f
1
>
A
el
AEPS >
2

c
*

c
( )
2
dV
}
c
2
dV
c
*
c
( )
2
dV
}
+
}
-------------------------------------------------------------- =
B
c
*

c
( )
2
dV
}
=
AECS f
2 *
B NUMEL f
3
* B * f
1
NUMEL + =
f
1
> B
el
AECS >
79
CHAPTER 4
or
or
Node Within A Box, Cylinder, or Sphere Criterion
An element is subdivided if it falls within the specified box, cylinder, or sphere, respectively. If all of the nodes of the
subdivided elements move outside the box, the elements are optionally merged back together.
The location of this box can be repositioned using the UADAPBOX user subroutine.
The criteria may also be used in conjunction with the WELD FLUX model definition option. In such cases, the motion
of the weld source controls the location of the box.
Nodes In Contact Criterion
An element is subdivided if one of its nodes is associated with a new contact condition. In the case of a deformable-
to-rigid contact, this implies that the node has touched a rigid surface. For deformable-to-deformable contact, the node
can be either a tied or retained node. Note that if chattering occurs, there can be an excessive number of elements
generated. Use the level option to reduce this problem.
Temperature Gradient Criterion
An element is subdivided if the temperature gradient in the element is greater than a given fraction of the maximum
gradient in the solution. This is the recommended method for heat transfer.
Pressure Gradient Criterion
An element is subdivided if the pressure gradient in the element is greater than a given fraction of the maximum
gradient. This is the recommended method for diffusion analysis.
Electrical Potential Criterion
An element is subdivided if the electrical potential in the element is greater than a given fraction of the maximum
gradient. This is the recommended method for electrostatic analysis.
Magnetostatic Potential Criterion
An element is subdivided if the magnetic potential in the element is greater than a given fraction of the maximum
gradient. This is the recommended method for magnetostatic analysis.
User-defined Criterion
The UADAP user subroutine can be used to prescribe a user-defined adaptive criteria. The user subroutine UADAP2
can be used to prescribe a criterion for merging elements back (unsubdivide).
vm
f
1
max
vm
>
vm
f
2
>
vm
f
3
max
vm
>
vm
f
4
>

Previously Refined Mesh Criterion
Use the refined mesh from a previous analysis as the starting point to this analysis. The information from the previous
adapted analysis is read in.
Angle Between Shell Elements
An element is refined if the change in angle between neighboring shell elements is larger than the given value. For
each node, a normal is calculated by averaging shell element normals for elements connected to the node. If an element
is to be subdivided, the angle used for checking is two times the angle between the averaged nodal normal and the
element normal. Only the change in the normals from the undeformed shape is used to avoid that elements originally
connected at an angle get subdivided immediately.
Automatic Global Remeshing
In the analysis of metal or rubber, the materials may be deformed from some initial (maybe simple) shape to a final,
very often, complex shape. During the process, the deformation can be so large that the mesh used to model the
materials become highly distorted, and the analysis cannot go any further without using some special techniques (see
Figure 4-27). Remeshing/rezoning in Marc is a useful feature to overcome the difficulties. Although global remeshing
can be done manually using the REZONE option, the automatic global remeshing procedure is recommended.
Figure 4-27 Mesh too Distorted and Simulation Stops (Left); Using Global Remeshing, the Simulation Completes (click
to play video)
Global remeshing can only be carried out on a contact body. Therefore, contact bodies are expected in the analysis
with global remeshing. The basic steps, which are automatic in global remeshing, are as follows:
1. Analysis checks remeshing criteria at the end of each increment. When one of the remeshing criteria is met,
analysis starts the remeshing procedure.
2. The deformed shape of the contact body is extracted. A new mesh is created by calling a stand-alone mesher
or an internal mesher.
3. The new mesh is checked and corrected to avoid any penetration or contact loss to other contact bodies.
4. A data mapping is performed to transfer necessary data from the old, deformed mesh to the new mesh.
5. The contact tolerance is recalculated (if not specified by you) and the contact conditions are redefined.
81
CHAPTER 4
6. Boundary conditions, if any, are transferred to the new mesh.
7. The analysis completes global remeshing and continues its computation based on the new mesh.
Global remeshing/rezoning can be used in two- or three-dimensional solid body or a three-dimensional shell body.
Most of the linear element types are supported. For shell, element type 75, 138, 139, and 140 are supported.
Figure 4-28 shows a simple 3-D rubber seal remeshing.
When remeshing/rezoning in 2-D, Marc finds the outline of the body to be rezoned and repairs the outline to remove
possible penetration. Marc then calls the mesher to create a new mesh based on the clean outline.
Figure 4-28 3-D Automatic Remeshing and Rezoning of a Rubber Seal
2-D Global Remeshing of Rubber Seal Insertion
3-D Global Remeshing of Connecting Rod Forging
Deep Drawing of a Box with Shell Remeshing

When remeshing in 3-D with tetrahedral elements, Marc extracts and outputs the surface information together with
other contact surface information and the meshing control parameters. A stand-alone 3-D mesher is then called to
recreate 3-D surface mesh and the volume mesh. Contact check and volume checks are used to ensure no penetration
and loss of volume in the new mesh. In 3-D hexahedral remeshing, hexahedral elements are build from the interior
toward the surface. Projections are used to project nodes to the surface.
When remeshing with 3-D shell elements, the 3-D mesher is called and surface mesh is created with triangular or
quadrilateral elements Contact check is also performed to remove possible penetration.
The automatic remeshing/rezoning feature uses the Updated Lagrange formulation by default.
For global remeshing controls, the following commands are required:
1. REZONING parameter to activate global remeshing controls.
2. ADAPT GLOBAL used in both model definition and history definition options to specify global remeshing
criteria, remeshing body, mesh generator, and meshing parameters.
The following are the supported features and limitations:
1. Analysis type:
Mechanical analysis, thermal-mechanical, thermal Joule mechanical, and electrostatic-structural coupled
analysis are supported.
2. Element types:
Low order, continuum element types are supported. In 2-D, these include both lower order quadrilateral and
triangular elements. In 3-D, only lower order tetrahedral element types are supported including Herrmann type
element 157. For 3-D shell elements, triangle element 138 and quadrilateral element 75, 139, and 140 are tested
and supported.
3. Contact analysis:
The remeshing body is a meshed contact body. Contact information, including boundary conditions defined
through contact definitions, is re-determined based on the new mesh. Meshes that are not defined in contact
bodies are not supported.
Support Rigid-Deformable contact
Support Deformable-Deformable contact
Support 2-D Self-contact
4. Boundary conditions:
Boundary conditions are allowed in a remeshing body. These boundary conditions are transferred to the new
mesh after remeshing. The new table style input format is required. These boundary conditions include:
Distributed load, flux, and current
Point load, flux, and current
Fixed displacement, temperature, and potential
The above boundary conditions defined in the interior of a contact body are not supported.
5. Local adaptivity meshing is not supported.
6. Parallel computation is not supported.
83
CHAPTER 4
7. Other analysis options:
PRE STATE model definition option is supported
CYCLIC SYMMETRY model definition option is supported
FLOW LINE and TRACK model definition options are supported
Other supported global remeshing features include:
1. Remeshing procedures:
pause and continue: a standalone mesher is called within the analysis process.
stop and continue: analysis is stopped while a standalone mesher is called. This is to save memory usage
during the remeshing process. Analysis is automatically resumed after remeshing is complete.
2. Forced remeshing control:
When global remeshing is selected with automatic time stepping scheme, remeshing will be forced when
there is an element distortion during the analysis. This is carried out before the end of the increment.
3. Remeshing criteria:
Element distortion check
Increment frequency check
Contact penetration check
Equivalent strain change check
Immediate meshing
4. Meshing control:
Based on input element size
Based on input number of elements
Change element types to triangle (2-D) or tetrahedral elements (3-D)
Using Overlay mesher quadrilateral elements
Using advancing front mesher quadrilateral or triangular elements
Using Delaunay triangulation mesher triangular elements
Using Hybrid Patran mesher 3-D tetrahedral elements
Using Patran MOM mesher - 3-D shell elements
Using 3-D Overlay mesher hexahedral elements (SuperForm only)
Read user defined mesh files (.mesh) also supported with RESTART and REAUTO options
5. Refinement and Coarsening control:
Local refinement based on surface curvatures
Local refinement based on thin section (2-D only)
Local refinement based on refinement boxes
Local refinement based on the USIZEOUTL user subroutine for 2-D advancing front mesher and
Delaunay triangulation mesher
Interior coarsening using 2-D overlay mesher
Interior coarsening using 3-D Tetrahedral mesher or Hexahedral mesher

6. Special Features:
Support trimming load case to remove part of mesh in a contact body
Support 2-D contact body split using the USPLIT user subroutine.
Support user-defined meshing using the UMAKENET user subroutine. (3-D shell remeshing not supported)
Remeshing Criteria
It is possible to choose any remeshing criteria simultaneously. How these criteria work is shown below.
Increment
Remeshing occurs at specified increment frequency.
Element Distortion
The identified body is remeshed when the distortion in the elements becomes large.
For 2-D analysis, the distortion check is based on corner angles. Remeshing is performed if the following conditions
are met:
Any inner angle is greater than 175 or less than 5
Any inner angle change is greater than the user input data
For 3-D analysis, a volume ratio is measured to determine if remeshing is required. A volume ratio is calculated based
on each corner node and its connecting nodes. If is the volume of a tetrahedron formed by nodes 1, 2, 3, and 4 and
is the triangle area of nodes 2, 3, and 4, then the ratio:
or if ,
Where is the height and is the equivalent length of the triangle. They can be calculated respectively by
The default control ratio is 0.01. Any volume ratio of each corresponding corner node smaller than this value forces
the analysis to perform remeshing. Users can change this number to control the remeshing.
Note:
In general, frequent remeshing should be avoided for an effective and computationally efficient analysis.
Also, since each remeshing and subsequent rezoning step involves interpolation and extrapolation of
element variables, a possibility of error accumulation exists as the analysis progresses when remeshing
occurs too frequently.
v
s
r
h
l
--- =
h l > r
l
h
--- =
h l
h
3v
s
------ =
l s =
85
CHAPTER 4
Contact Penetration
The identified body is remeshed when the curvature of the contact body is such that the current mesh cannot accurately
detect penetration.
For 2-D analysis, the penetration remeshing criteria is based upon examining the distance between the edge of an
element and the contacted body.
For 3-D analysis, the penetration is measured from the center of each boundary element face to the contacting surface.
By default, remeshing is carried out if the penetration is greater than twice the contact
tolerance and less than the target element size, where the contact tolerance is 0.05 of the
smallest element length and the target element size is the element size for remeshing.
This check does not apply to the self-contact situation. Also, this penetration limit can
be given by user input to avoid too many remeshing operations.
Remeshing is activated when the penetration distance reaches or exceeds the given
penetration tolerance.
Immediate
The identified body is remeshed before performing any analysis. This control can also be used to change element types
from quadrilateral elements to triangular elements or from hexahedral elements to tetrahedral elements. For example,
it is possible to use this control to change an initially defined mesh using element type 7 (an 8 node-hexahedral
element) to one using element type 157 (5 node-tetrahedral element) in order to use the tetrahedral remeshing
capability.
Strain Change
Equivalent strain measures element deformation. This criterion keeps a record of the strain change after remeshing for
each element. When any element of the body has a strain change greater than the control limit, the remeshing will start.
Remeshing Techniques
For 2-D analysis, the remeshing techniques include outline extraction and repair and the mesh generation. After the
outline is extracted and repaired, the mesh generator is called to create a mesh. For 2-D remeshing, the new mesh is
created either through the built-in mesh generator using the overlay method or through a standalone mesh generator.
When a standalone mesh generator is called, the program pauses while waiting for the mesh generator to create the
new mesh. This can be memory intensive as both program and mesher are using the memory. Alternatively, the
program can be stopped automatically while the mesh is being created, freeing the memory for the mesh generation.
The program automatically resumes after the meshing is complete. In Marc, this is accomplished by using the AUTO
RESTART option, -autorst, through the command line parameter. In Mentat, this is instructed through JOBJOB
PARAMETERSREMESHING CONTROLSTOP AND RESTART.
For 3-D analysis, the outer surface of the contact body is extracted to a data file. This data file also contains remeshing
control information. For tetrahedral meshing, a standalone 3-D mesher is called to create the surface mesh with
triangles and then mesh the body with the tetrahedral elements. Patran mesher is used for the meshing. The new mesh
nodes are adjusted to ensure that there is no penetration on the contact surfaces.

Mesh Generation
There are various 2-D and 3-D mesh generators available in Marc based on Advancing front, Overlay, and Delaunay
Triangulation techniques.
Advancing front mesher
This 2-D mesher creates either triangular, quadrilateral, or mixed triangular and quadrilateral mesh. For a given outline
boundary, it starts by creating the elements along the boundary. The new boundary front is then formed when the layer
of elements is created. This front advances inward until the complete region is meshed. Some smoothing technique is
used to improve the quality of the elements. In general, this mesher works with any enclosed geometry and for
geometry that has holes inside. The element size can be changed gradually from the boundary to the interior allowing
smaller elements near the boundary with no tying constraints necessary (see Figure 4-29).
Figure 4-29 Advancing Front Meshing
Overlay meshing
This is a quadrilateral mesh generator. The 2-D overlay mesher is included within Marc. It creates a quadrilateral mesh
by forming a regular grid covering the center area of a body. A projection is then used to project all boundary nodes
onto the real surface and form the outer layer elements. For the surface that is not in contact with other bodies, a cubic
spline line is used to make the outline points smoother. This mesher also allows up to two level refinements on
boundary where finer edges are needed to capture the geometry detail, and one level of coarsening in the interior where
small elements are not necessary. This refinement and coarsening are performed using the tying constraints (see
Figure 4-30).
Figure 4-30 Local Refinement and Coarsening
87
CHAPTER 4
In general, overlay meshing produces good quality elements. However, it does not take geometry with holes inside. It
may not create a good mesh with geometry that has a very thin region or very irregular shape. Also, because the regular
grid is created based on the global coordinate system, it may create a poor mesh if the geometry is not aligned with
the global coordinate system.
Delaunay Triangulation
This mesher creates only the triangular mesh. All the triangles satisfy the Delaunay triangulation property. It takes all
the seed points on the improved outlines as initial triangulation points. The triangulation is implemented by sequential
insertion of new points into the triangulation until all the triangles satisfy the local density and quality requirement.
Delaunay triangulation algorithm assures the triangular mesh created has the best quality possible for the given set of
points. The mesher also allows geometry to have holes inside the body and a variation of the elements with different
sizes (Figure 4-31).
Figure 4-31 Meshing with Delaunay Triangulation
Patran Tetrahedral Mesher
MOM (Meshing-On-Mesh) Surface Mesher
This is a 3-D surface mesher that creates a surface mesh based on an input mesh. The Meshing-on-mesh technology
allows users to input a mesh, which is not good for the analysis, such as a distorted mesh or a mesh from the STL file.
Tetrahedral Hybrid Mesher
This 3-D mesher is based on Delaunay triangulation and advancing front technology. It generates tetrahedral elements
based on an enclosed triangular surface mesh.
2-D Outline Extraction and Repair
The outline consists of all the boundary edges of the contact body. Once the outline is extracted from the mesh, it is
checked against other contacting bodies. The penetration from other contacting bodies is marked and the outline is
corrected according to the penetration. If the outline is to be used for the advancing front or Delaunay mesher, some
refinement and corrections are required to obtain better outline points. These outline points become boundary nodes
in the new mesh and cannot be altered during the meshing process. The following procedures are taken by the program
to prepare the new outline for the remeshing:

Step 1: Marking the hard points: The hard points are those points that represent important features of the original
outline. Hard points are points that mark the beginning or end of the boundary portion of the mesh that is
in contact or assigned with distributed boundary conditions, and the points that represent a sharp corner
(such as 90 angle) or nodal boundary conditions.
Step 2: Marking the points with target element size and minimum element size: The outline points are placed
based on the target element size. The refinements and the user-defined outline points are allowed to
change this control. However, the minimum element size is used to make sure the outline segment is not
too small for the mesh generation.
Step 3: Marking the points with curvature consideration: The curvature control is used to allow small outline
segments to be used on the boundary where curvature radii are small. Three neighboring outline points
are used to calculate the curvature radius. The associate curvature circle can then be formed by a number
of line segments. With the same number of line segments used to approximate a circle, the curve with a
smaller curvature circle gets a smaller line segment (Figure 4-31). Thus, we have a good variation of mesh
size based on the curvature.
Figure 4-32 Curvature Consideration
Step 4: Marking the points with thin region consideration: In the area where a thin region is formed, small
elements are preferred. This can be done by detecting the thin region and using smaller outline segments
in the area. The segment length used for the thin area is to allow at least three elements to be presented
across the thin area.
Step 5: Smoothing the outline points: Smoothing is required on the outline so that the segment length can
gradually vary.
Step 6: Interpolations: Interpolation is the actual process to create a new outline based on the extracted outline
and the marking of the outline points. Linear interpolation is used on the contact area to prevent
penetration into the other contact bodies. Cubic spline line interpolation is used for the free surfaces.
3-D Surface Extraction and Meshing
Similar to the 2-D outline extraction, 3-D surface faces are extracted before calling the mesh generator. If the original
mesh is a hexahedral mesh, the boundary element faces are converted into triangles. The contact information on each
element face is also extracted and output to a data file. A standalone mesher reads the information and perform the
following steps:
Step 1: Construct surface information for contact projection and volume check.

R
r l
2R
n
----------- =
89
CHAPTER 4
Step 2: Generate triangle surface mesh using Patran surface mesher (MOM). Surface elements are created
observing the basic geometry features and contact conditions. Local refinement will be carried out based
on the surface curvatures and user defined refinement boxes (see Figure 4-33).
Figure 4-33 Local Refinement
Step 3: Check contact penetration of new nodes and adjust coordinates if penetration is found.
Step 4: Generate tetrahedral element mesh with Patran Hybrid mesher. By default, a coarsening factor is adopted
to enlarge interior elements (see Figure 4-34).
Step 5: Output the mesh to a data file.
Figure 4-34 Interior Coarsening
Remeshing Based on the Target Number of Elements
Instead of giving the element size, users can give the target number of elements for the remeshing. A uniform mesh
assumption is used to compute the element size with the target number of elements. If the target number of elements
is not provided, the number of elements in the current mesh is used. For 2-D analysis, the number of elements in the
new mesh can also be controlled by using a percentage tolerance to ensure that the new mesh does not have too many
or too few elements. However, this tolerance control requires remeshing trials and it cannot be used with the AUTO
RESTART option.
Remeshing with Boundary Conditions
There are two types of boundary conditions that global remeshing needs to transfer from old mesh to the new one:
Contact boundary conditions:
This type of boundary conditions includes friction, heat convection, radiation, symmetry or cyclic symmetry
conditions. After a new mesh is created, these contact boundary conditions are automatically updated based on
the new contact detection using the new mesh.
Small Elements
Large Elements

User defined boundary conditions:
If the remeshing body has user defined boundary conditions, such as, point loads, distributed loads or fixed
nodal displacements, the boundary conditions are transferred to the new mesh with two different approaches:
a. Defined Set Approach
This approach allows users to directly apply boundary conditions to the mesh entities, such as, element faces,
edges, or nodes. Each boundary condition is arranged in a defined set with a set name and a set type.
Currently, the set type can be either a node or an element edge or face set. The element edge sets in 2-D and
face sets in 3-D are typically those of distributed boundary conditions. Edges or faces in the set can be
continuous or discontinuous. During the remeshing stage, the boundaries of these edges or faces are marked
and preserved. This allows the new edges or faces to be created coincident with the old edges or faces in the
set. After remeshing, the element edges or faces in the set are replaced with the new element edges or faces
(see Figure 4-35). The node ids in nodal sets are automatically replaced with the new node ids (see
Figure 4-36).
Figure 4-35 Distributed Loads Transfer
Figure 4-36 Point Loads Transfer
The nodal displacement or temperature conditions applying to all the nodes of an element edge or face are
treated as an element edge or face boundary condition. This means the new nodes created on the same
element edge or face will have the same displacement or temperature boundary conditions. In 2-D, an
element edge or face can take up to four boundary condition sets for the remeshing. In 3-D, the number of
sets on the same element face is two. There is no limitation on the node set.
91
CHAPTER 4
b. Geometry Attachment Approach
Boundary conditions can be assigned to geometry entities, such as points, curves and surfaces. In this
approach, these geometry entities are attached to the mesh entities of a remeshing body, such as, a point
attached to a node and a curve attached to a set of element edges. During remeshing, mesh entities associated
with these geometry entities are marked and preserved. After remeshing, the geometry entities are re-
attached to the new mesh entities. By doing so, boundary conditions assigned to these geometry entities are
automatically applied to the new mesh (see Figure 4-37). In 3-D, the attached surface ID number cannot
exceed 99 and only one surface can be attached to an element face. In 3-D, curves can also be attached to
the element edges but only nodal displacement conditions are supported (see Figure 4-38). Geometry
attachments are shown.
Figure 4-37 Distributed Load with Curve Attachment
Figure 4-38 Geometry Attachment Boundary Conditions in 3-D

History Data Mapping Technique
History data such as stress, strain or temperature need to be transferred to the new mesh (dashed red), from the old
mesh (solid black). In general, this is carried out in the following steps (see Figure 4-39):
Figure 4-39 Locate New Nodes in Old Mesh
Store nodal data at the nodal position based on the old mesh
If old mesh is formed of quadrilateral or hexahedral elements, subdivide into triangles or tetrahedrals,
respectively.
Extrapolate data from the integration points to the nodal position
If old mesh is formed of quadrilateral or hexahedral elements, determine values at extra nodes
Compute weighted averaged nodal data based on the contributions from different elements
Locate new nodes in the old triangular/tetrahedral region
Determine the value at the new node based upon linear interpolation within triangular/tetrahedral region
Data in the new integration points can be computed based on interpolating from the nodal data in the new mesh
After data mapping, new equilibrium is achieved at the end of the new increment.
Chapter 5 Structural Procedure Library
5
Structural Procedure Library
Linear Analysis 94
Nonlinear Analysis 99
Fracture Mechanics 140
Dynamics 161
Inertia Relief 180
Rigid-Plastic Flow 183
Superplasticity 185
Soil Analysis 188
Mechanical Wear 192
Design Sensitivity Analysis 194
Design Optimization 197
Defined Initial State with Result Data from Previous Analysis

(including AXITO3D) 203
Compatible element types for PRE STATE 204
Structural Zooming Analysis 208
Cure-Thermal-Mechanically Coupled Analysis 211
References 216

This chapter describes the analysis procedures in Marc applicable to structural problems. These procedures range from
simple linear elastic analysis to complex nonlinear analysis. A large number of options are available, but you need to
consider only those capabilities that are applicable to your physical problem. This chapter provides technical
background information as well as usage information about these capabilities.
Linear Analysis
Linear analysis is the type of stress analysis performed on linear elastic structures. Because linear analysis is simple
and inexpensive to perform and generally gives satisfactory results, it is the most commonly used structural analysis.
Nonlinearities due to material, geometry, or boundary conditions are not included in this type of analysis. The behavior
of an isotropic, linear, elastic material can be defined by two material constants: Youngs modulus , and Poissons
ratio .
Marc allows you to perform linear elastic analysis using any element type in the program. Various kinematic
constraints and loadings can be prescribed to the structure being analyzed; the problem can include both isotropic and
anisotropic elastic materials.
The principle of superposition holds under conditions of linearity. Therefore, several individual solutions can be
superimposed (summed) to obtain a total solution to a problem.
Linear analysis does not require storing as many quantities as does nonlinear analysis; therefore, it uses the core
memory more sparingly. The ELASTIC parameter uses the assembled and decomposed stiffness matrices to arrive at
repeated solutions for different loads.
Linear analysis in Marc requires only the basic input. Table 5-1 shows a subset of the Marc options and parameters
which are often used for linear analysis.
Note: Linear analysis is always the default analysis type in the Marc program.
Table 5-1 Basic Input
Type Name
Parameter TITLE
SIZING
ELEMENTS
ELASTIC
ALL POINTS
CENTROID
ADAPTIVE
FOURIER
END
E
v
95
CHAPTER 5
More complex linear analyses require additional data blocks.
1. The ELASTIC parameter allows solutions for the same structural system with different loadings (multiple
loading analysis). When using the ELASTIC parameter, you must apply total loads, rather than incremental
quantities (for example., total force, total moment, total temperature) in subsequent increments.
2. The RESTART option, used with the ELASTIC parameter and/or CASE COMBIN option, stores individual load
cases in a restart file.
3. The CASE COMBIN model definition option combines the results obtained from different loading cases
previously stored in a restart file.
4. The ADAPTIVE option can be used to improve the accuracy of the analysis.
5. The J-INTEGRAL option allows the study of problems of linear fracture mechanics.
6. The FOURIER option allows the analysis of axisymmetric structures subjected to arbitrary loadings.
7. The ORTHOTROPIC or ANISOTROPIC model definition option activates the anisotropic behavior option. In
addition, the ANELAS, HOOKLW, ANEXP, and ORIENT user subroutines define the mechanical and thermal
anisotropy and the preferred orientations.
8. You can use both the linear SPRINGS and FOUNDATION options in a linear stress analysis.
Accuracy
It is difficult to predict the accuracy of linear elastic analysis without employing special error estimation techniques.
An inaccurate solution usually exhibits itself through one or more of the following phenomena:
Strong discontinuities in stresses between elements
Strong variation in stresses within an element
Stresses that oscillate from element to element
Error Estimates
The ERROR ESTIMATE option can also be used to obtain an indication of the quality of the results. You can have the
program evaluate the geometric quality of the mesh by reporting the aspect ratios and skewness of the elements. In a
large deformation updated Lagrange analysis, you can also observe how these change during the analysis, which
indicates mesh distortion. When the mesh distortion is large, it is a good idea to do a rezoning step.
Model Definition CONNECTIVITY
COORDINATES
GEOMETRY
ISOTROPIC
FIXED DISP
DIST LOADS
POINT LOAD
CASE COMBIN
END OPTION
Table 5-1 Basic Input (continued)
Type Name

The ERROR ESTIMATE option can also be used to evaluate the stress discontinuity between elements. Marc first
calculates a nodal stress based upon the extrapolated integration point values. These nodal values are compared
between adjacent elements and reported. Large discrepancies indicate an inability of the mesh to capture high stress
gradients, in which case you should refine the mesh and rerun the analysis or use local adaptive meshing.
The ERROR ESTIMATE option can be used for either linear or nonlinear analysis.
Adaptive Meshing
The ADAPTIVE option can be used to insure that a certain level of accuracy is achieved. The elastic analysis is repeated
with a new mesh until the level of accuracy requested is achieved.
Fourier Analysis
Through Fourier expansion, Marc analyzes axisymmetric structures that are subjected to arbitrary loading. The
FOURIER option is available only for linear analysis.
During Fourier analysis, a three-dimensional analysis decouples into a series of independent two-dimensional
analyses, where the circumferential distribution of displacements and forces are expressed in terms of the Fourier
series. Both mechanical and thermal loads can vary arbitrarily in the circumferential direction. You can determine the
structures total response from the sum of the Fourier components.
The Fourier formulation is restricted to axisymmetric structures with linear elastic material behavior and small strains
and displacements. Therefore, conditions of linearity are essential and material properties must remain constant in the
circumferential direction.
To use Fourier expansion analysis in Marc, the input must include the following information:
The FOURIER parameter allocates storage for the series expansion.
Fourier model definition blocks for as many series as are needed to describe tractions, thermal loading, and
boundary conditions. Number the series sequentially in the order they occur during the FOURIER model
definition input. Three ways to describe the series are listed below:
Specify coefficients on the Fourier model definition blocks.
Describe (where is the angle in degrees about the circumference) in point-wise fashion with an
arbitrary number of pairs given on the blocks. Marc forms the corresponding series coefficients.
Generate an arbitrary number of pairs using the UFOUR user subroutine and let the program
calculate the series coefficients.
You can obtain the total solution at any position around the circumference by superposing the components already
calculated after completion of all increments required by the analysis. The CASE COMBIN option calculates this total
solution by summing the individual harmonics which are stored in the restart file.
The number of steps or increments needed for analysis depends on the number of harmonics that are chosen. For a full
analysis with symmetric and antisymmetric load cases, the total number of increments equals twice the number of
harmonics. Table 5-2 shows which Fourier coefficients are used for a given increment.
a
0
a
1
b
1
, ,
F ( )
F ( ) , [ ]
F ( ) , [ ]
97
CHAPTER 5
The magnitude of concentrated forces should correspond to the value of the ring load integrated around the
circumference. Therefore, if the Fourier coefficients for a varying ring load are found from the
distribution, where has the units of force per unit length, the force magnitude given in the point load block should
equal the circumference of the loaded ring. If is in units of force per radian, the point load magnitude should be
.
The Fourier series can be found for varying pressure loading from input with expressed in force per unit
area. Marc calculates the equivalent nodal forces and integrates them around the circumference. The distributed load
magnitude in the distributed loads block should be 1.0.
Table 5-3 shows the elements in the program that can be used for Fourier analysis.
Table 5-2 Fourier Coefficients Increment Number
LOAD TERMS
INC. 1st DOF,Z 2nd DOF,R 3rd DOF,
0 0
1 0 0
2
3
. . . .
. . . .
. . . .
Table 5-3 Elements Used for Fourier Analysis
Element Type Description
62 8-node
73 8-node with reduced integration
63 8-node for incompressible behavior
74 8-node for incompressible behavior with reduced integration
90 3-node shell
a
0
a
0
a
0
a
1
a
1
b
1
b
1
b
1
a
1
2n a
n
a
n
b
n
2n 1 + b
n
b
n
a
n
p ( ) p ( ) , [ ]
p ( )
p ( )
2
p ( ) , [ ] p

The general form of the Fourier series expansion of the function is shown in the equation below.
(5-1)
This expression expands the displacement function in terms of sine and cosine terms. A symmetric and an
antisymmetric problem are formulated for each value of .
The displacements for the symmetric case, expressed in terms of their nodal values, are
(5-2)
Nodal forces are
(5-3)
The value is a special case in Fourier analysis. If only the symmetric expansion terms are used, the formulation
defaults to the fully axisymmetric two-dimensional analysis. The antisymmetric case for yields a solution
for the variable that corresponds to loading in the tangential direction. Analyze axisymmetric solids under pure
torsion in this way.
Modal Shapes and Buckling Load Estimations During a Fourier Analysis
During a Fourier analysis, Marc can be asked to estimate both the modal shapes and buckling loads for each harmonic
in the analysis. In either case, the program performs a Fourier analysis first and then estimates the modal
shapes/buckling load at prescribed harmonic numbers. In addition to the input data required for a Fourier analysis
(FOURIER parameter and FOURIER model definition option), the following must also be added: DYNAMIC parameter
and MODAL INCREMENT model definition option; BUCKLE parameter and BUCKLE INCREMENT model definition
option, for Fourier modal shape and Fourier buckling load estimations, respectively.
F ( )
F ( ) a
0
a
n
n cos b
n
n sin + ( )
n 1 =
+ =
n
u
n
N
1
N
2
, , [ ] n u
n
{ }
e
cos =
v
n
N
1
N
2
, , [ ] n v
n
{ }
e
cos =
w
n
N
1
N
2
, , [ ] n sin w
n
{ }
e
=
Z Z
0
Z
n
n cos
1
n
+ =
R R
0
R
n
n cos
1
n
+ =
T T
0
T
n
n sin
1
n
+ =
n 0 =
n 0 =
99
CHAPTER 5
In the Fourier modal analysis, the mass matrix in the eigenvalue equation is a constant matrix. The stiffness matrix in
the eigenvalue equation is the one associated with a prescribed harmonic of the Fourier analysis. The expression of
the eigenvalue equation is:
(5-4)
where is the stiffness matrix associated with the m
th
harmonic of the Fourier analysis and is a constant
matrix. Multiple modes for each harmonic can be extracted.
Similarly, in a Fourier buckling analysis, the stiffness matrices in the eigenvalue equation are (respectively); the linear
elastic stiffness matrix and the geometric stiffness matrix associated with the prescribed harmonic of the Fourier
analysis.
The eigenvalue equation is expressed
(5-5)
where is the linear elastic stiffness matrix and is the geometric stiffness matrix, associated with the m
th

harmonic of the Fourier analysis. The stresses used in the calculation of the geometric stiffness matrix are those
associated with the symmetric load case, . Multiple buckling load estimations for each harmonic are also
available.
Nonlinear Analysis
The finite element method can be used for nonlinear, as well as linear, problems. Early development of nonlinear finite
element technology was mostly influenced by the nuclear and aerospace industries. In the nuclear industry,
nonlinearities are mainly due to the nonlinear, high-temperature behavior of materials. Nonlinearities in the aerospace
industry are mainly geometric in nature and range from simple linear buckling to complicated post-bifurcation
behavior. Nonlinear finite element techniques have become popular in metal forming manufacturing processes, fluid-
solid interaction, and fluid flow. In recent years, the areas of biomechanics and electromagnetics have seen an
increasing use of finite elements.
A problem is nonlinear if the force-displacement relationship depends on the current state (that is, current
displacement, force, and stress-strain relations). Let be a generalized displacement vector, a generalized force
vector, and the stiffness matrix. The expression of the force-displacement relation for a nonlinear problem is
(5-6)
Linear problems form a subset of nonlinear problems. For example, in classical linear elastostatics, this relation can
be written in the form
(5-7)
where the stiffness matrix is independent of both and . If the matrix depends on other state variables that
do not depend on displacement or loads (such as temperature, radiation, moisture content, etc.), the problem is still
linear.
K
m
[ ]
2
M
0
[ ] 0 =
K
m
[ ] M
0
[ ]
K
m
[ ] K
g
m
[ ] 0 =
K
m
[ ] K
g
m
[ ]
m 0 =
u P
K
P K P u , ( )u =
P Ku =
K u P K

Similarly, if the mass matrix is a constant matrix, the following undamped dynamic problem is also linear:
(5-8)
There are three sources of nonlinearity: material, geometric, and nonlinear boundary conditions. Material nonlinearity
results from the nonlinear relationship between stresses and strains. Considerable progress has been made in attempts
to derive the continuum or macroscopic behavior of materials from microscopic backgrounds, but, up to now,
commonly accepted constitutive laws are phenomenological. Difficulty in obtaining experimental data is usually a
stumbling block in mathematical modeling of material behavior. A plethora of models exist for more commonly
available materials like elastomers and metals. Other material model of considerable practical importance are:
composites, viscoplastics, creep, soils, concrete, powder, and foams. Figure 5-1 shows the elastoplastic, elasto-
viscoplasticity, and creep. Although the situation of strain hardening is more commonly encountered, strain softening
and localization has gained considerable importance in recent times.
Figure 5-1 Material Nonlinearity
Geometric nonlinearity results from the nonlinear relationship between strains and displacements on the one hand and
the nonlinear relation between stresses and forces on the other hand. If the stress measure is conjugate to the strain
measure, both sources of nonlinearity have the same form. This type of nonlinearity is mathematically well defined,
but often difficult to treat numerically. Two important types of geometric nonlinearity occur:
a. The analysis of buckling and snap-through problems (see Figure 5-2 and Figure 5-3).
P Mu
Ku + =
Elasto-Plastic Behavior
Elasto-Viscoplastic Behavior
c
t
Creep Behavior
101
CHAPTER 5
Figure 5-2 Buckling
Figure 5-3 Snap-Through
b. Large strain problems such as manufacturing, crash, and impact problems. In such problems, due to large
strain kinematics, the mathematical separation into geometric and material nonlinearity is nonunique.
Boundary conditions and/or loads can also cause nonlinearity. Contact and friction problems lead to nonlinear
boundary conditions. This type of nonlinearity manifests itself in several real life situations; for example, metal
forming, gears, interference of mechanical components, pneumatic tire contact, and crash (see Figure 5-4). Loads on
a structure cause nonlinearity if they vary with the displacements of the structure. These loads can be conservative, as
in the case of a centrifugal force field (see Figure 5-5); they can also be nonconservative, as in the case of a follower
force on a cantilever beam (see Figure 5-6). Also, such a follower force can be locally nonconservative, but represent
a conservative loading system when integrated over the structure. A pressurized cylinder (see Figure 5-7) is an
example of this.
P
u
Linear
S
Neutral
Uns
P
P
c
u
P
u
P
u

Figure 5-4 Contact and Friction Problem
Figure 5-5 Centrifugal Load Problem (Conservative)
Figure 5-6 Follower Force Problem (Nonconservative)
P
P
103
CHAPTER 5
Figure 5-7 Pressurized Cylinder (Globally Conservative)
The three types of nonlinearities are described in detail in the following sections.
Geometric Nonlinearities
Geometric nonlinearity leads to two types of phenomena: change in structural behavior and loss of structural stability.
There are two natural classes of large deformation problems: the large displacement, small strain problem and the large
displacement, large strain problem. For the large displacement, small strain problem, changes in the stress-strain law
can be neglected, but the contributions from the nonlinear terms in the strain displacement relations cannot be
neglected. For the large displacement, large strain problem, the constitutive relation must be defined in the correct
frame of reference and is transformed from this frame of reference to the one in which the equilibrium equations are
written.
The collapse load of a structure can be predicted by performing an eigenvalue analysis. If performed after the linear
solution (increment zero), the Euler buckling estimate is obtained. An eigenvalue problem can be formulated after each
increment of load; this procedure can be considered a nonlinear buckling analysis even though a linearized eigenvalue
analysis is used at each stage.
The kinematics of deformation can be described by the following approaches:
A. Lagrangian Formulation
B. Eulerian Formulation
C. Arbitrary Eulerian-Lagrangian (AEL) Formulation
The choice of one over another can be dictated by the convenience of modeling physics of the problem, rezoning
requirements, and integration of constitutive equations.
Lagrangian Formulation
In the Lagrangian method, the finite element mesh is attached to the material and moves through space along with the
material. In this case, there is no difficulty in establishing stress or strain histories at a particular material point and the
treatment of free surfaces is natural and straightforward.

The Lagrangian approach also naturally describes the deformation of structural elements; that is, shells and beams,
and transient problems, such as the indentation problem shown in Figure 5-8.
Figure 5-8 Indentation Problem with Pressure Distribution on Tool
This method can also analyze steady-state processes such as extrusion and rolling. Shortcomings of the Lagrangian
method are that flow problems are difficult to model and that the mesh distortion is as severe as the deformation of the
object. Severe mesh degeneration is shown in Figure 5-9b. However, recent advances in adaptive meshing and
rezoning have alleviated the problems of premature termination of the analysis due to mesh distortions as shown in
Figure 5-9c.
The Lagrangian approach can be classified in two categories: the total Lagrangian method and the updated Lagrangian
method. In the total Lagrangian approach, the equilibrium is expressed with the original undeformed state as the
reference; in the updated Lagrangian approach, the current configuration acts as the reference state. The kinematics of
deformation and the description of motion is given in Figure 5-10 and Table 5-4.
Figure 5-9 Rezoning Example
s
z
u
(a) Original (b) Deformed Mesh
(Undeformed Mesh) Before Rezoning
(c) Deformed Mesh
After Rezoning
105
CHAPTER 5
Figure 5-10 Description of Motion
Depending on which option you use, the stress and strain results are given in different form as discussed below. If the
LARGE DISP or LARGE STRAIN parameters are not used, the program uses and prints engineering stress and strain
measures. These measures are suitable only for analyses without large incremental or total rotation or large
incremental or total strains.
LARGE DISP
Using the LARGE DISP parameter, Marc uses the total Lagrangian method. The program uses and prints second Piola-
Kirchhoff stress and Green-Lagrange strain. These measures are suitable for analysis with large incremental rotations
and large incremental strains.
LARGE STRAIN
With the LARGE STRAIN parameter, Marc offers a convenient option to flag the appropriate large displacement / large
strain formulation for a particular element / material combination. The default formulation preference is Updated
Table 5-4 Kinematics and Stress-Strain Measures in Large Deformation
Configuration Measures Reference (t = 0 or n) Current (t = n + 1)
Coordinates
Deformation Tensor
Strain Measure
Stress Measure
X
C (Right Cauchy-Green)
E (Green-Lagrange)
F (Deformation Gradient)
S (second Piola-Kirchhoff)
P (first Piola-Kirchhoff)
x
b (Left Cauchy-Green)
e (Logarithmic)
(Cauchy)
Reference
t = 0
Current
t = n + 1
Previous
t = n
u
n + 1
F
f
u
F
n
u
n
F
n+1
= fF
n

Lagrange which uses Cauchy stresses and true strains. This is suitable for analyses with large elastic and plastic
strains. Stress and strain components are printed with respect to the current state. The LARGE STRAIN parameter is
normally specified as LARGE STRA, N where N can take three different values 1, 2 or 4 (default = 1):
1. LARGE STRA,1: this favors the Additive Decomposition Updated Lagrange procedure at the global level. For
specific element and material combinations, it may override the global parameters and replace them with more
suitable local parameters. Also, for certain cases (for example, analysis with mechanical shape memory alloy,
analysis with Herrmann elements, analysis with hyperelastic materials), it may change the global parameters
from LARGE STRA,1 to LARGE STRA,2. The user has no flexibility in controlling the global parameters and
limited flexibility in controlling the local parameters. More information on this is provided in the next section.
2. LARGE STRA,2: this favors the Multiplicative Decomposition Updated Lagrange procedure at the global level.
For specific element and material combinations, it may override the global parameters and replace them with
more suitable local parameters. The user has no flexibility in changing the global or local parameters.
3. LARGE STRA,4: this does not favor any particular procedure at the global level. For each element and material
combination, the optimal choice of formulation flags is automatically determined by the program. The user has
complete flexibility in changing these local parameters. In general, the preference at the local level is for the
updated Lagrange procedure.
For more details on the criteria followed in the selection of a procedure, the user is referred to the LARGE STRAIN
parameter in Marc Volume C: Program Input.
Theoretically and numerically, if formulated mathematically correct, the Total Lagrange and Updated Lagrange
formulations yield exactly the same results. However, integration of constitutive equations for certain types of material
behavior (for example, plasticity) make the implementation of the total Lagrange formulation inconvenient. If the
constitutive equations are convected back to the original configuration and proper transformations are applied, then
both formulations are equivalent. However, for deformations involving excess distortions, ease of rezoning favors the
updated Lagrangian formulation. This is reflected in the fact that a rezoned mesh in the current state is mapped back
to excessively distorted mesh leading to negative Jacobian in the total Lagrangian formulation.
The terminology total and updated Lagrangian has been used with some vagueness [1, 2]. In this document, for a
sequence of incremental motions at and , the total Lagrangian formulation entails the use of
configuration as reference; while in the updated Lagrangian configuration, the (unequilibriated)
configuration is the reference.
Total Lagrangian Procedure
The total Lagrangian procedure can be used for linear or nonlinear materials, in conjunction with static or dynamic
analysis. Although this formulation is based on the initial element geometry, the incremental stiffness matrices are
formed to account for previously developed stress and changes in geometry.
This method is suitable for the analysis of nonlinear elastic problems (for instance, with the Mooney, Ogden, or
NLELAST material behavior or the HYPELA2 user subroutine). The total Lagrangian approach is also useful for
problems in plasticity and creep, where moderately large rotations but small strains occur. A case typical in problems
of beam or shell bending. However, this is only due to the approximations involved.
To activate the large displacement (total Lagrangian approach) option in Marc, use the LARGE DISP parameter.
Include the FOLLOW FOR parameter for follower force (for example, centrifugal or pressure load) problems. This
parameter forms all distributed loads on the basis of the current geometry. Do not use the CENTROID parameter with
t 0 1 2 n , , , = n 1 +
t 0 = t n 1 + =
107
CHAPTER 5
this parameter. Always use residual load corrections with this parameter. To input control tolerances for large
displacement analysis, use CONTROL model definition option.
In the total Lagrangian approach, the equilibrium can be expressed by the principle of virtual work as:
(5-9)
Here is the symmetric second Piola-Kirchhoff stress tensor, , is the Green-Lagrange strain, is the body force
in the reference configuration, is the traction vector in the reference configuration, and is the virtual
displacements. Integrations are carried out in the original configuration at . The strains are decomposed in total
strains for equilibrated configurations and the incremental strains between and as:
(5-10)
while the incremental strains are further decomposed into linear, and nonlinear, parts as:
where is the linear part of the incremental strain expressed as:
(5-11)
The second term in the bracket in Equation (5-11) is the initial displacement effect.

is the nonlinear part of the
incremental strain expressed as:
(5-12)
Linearization of equilibrium of Equation (5-9) yields:
(5-13)
where is the small displacement stiffness matrix defined as
S
i j
E
i j
V d
V
0
}
b
i
0
i
V t
i
0
i
A d
A
0
}
+ d
V
0
}
=
S
i j
E
i j
b
i
0
t
i
0
i
t 0 =
t n = t n 1 + =
E
i j
n 1 +
E
i j
n
E
i j
+ =
E
i j
E
i j
n
E
i j
E
i j
E
n
i j
+ =
E
E
1
2
---
u
i
X
j
------------
u
j
X
i
------------ +
1
2
---
u
k
n
X
i
---------
\ .
|
| | u
k
X
j
-------------
\ .
| |
u
k
n
X
j
---------
\ .
|
| | u
k
X
i
-------------
\ .
| |
+ + =
E
n
E
n 1
2
---
u
k
X
i
-------------
\ .
| |
u
k
X
j
-------------
\ .
| |
=
K
0
K
1
K
2
+ + { }u F R =
K
0
K
0
( )
i j

i mn
0
D
mnpq
pqj
0
V d
V
0
}
=

is the initial displacement stiffness matrix defined as
in the above equations, and are the constant and displacement dependent symmetric shape function
gradient matrices, respectively, and is the material tangent,
and is the initial stress stiffness matrix
in which is the second Piola-Kirchhoff stresses and is the shape function gradient matrix.
Also, is the correction displacement vector. Refer to Chapter 11 in this manual for more details on the solution
procedures. and are the external and internal forces, respectively.
This Lagrangian formulation can be applied to problems if the undeformed configuration is known so that integrals
can be evaluated, and if the second Piola-Kirchhoff stress is a known function of the strain. The first condition is not
usually met for fluids, because the deformation history is usually unknown. For solids, however, each analysis usually
starts in the stress-free undeformed state, and the integrations can be carried out without any difficulty.
For viscoelastic fluids and elastic-plastic and viscoplastic solids, the constitutive equations usually supply an
expression for the rate of stress in terms of deformation rate, stress, deformation, and sometimes other (internal)
material parameters. The relevant quantity for the constitutive equations is the rate of stress at a given material point.
It, therefore, seems most obvious to differentiate the Lagrangian virtual work equation with respect to time. The rate
of virtual work is readily found as
(5-14)
This formulation is adequate for most materials, because the rate of the second Piola-Kirchhoff stress can be written as
(5-15)
For many materials, the stress rate is even a linear function of the strain rate
(5-16)
Equation (5-14) supplies a set of linear relations in terms of the velocity field. The velocity field can be solved non
iteratively and the displacement can be obtained by time integration of the velocities.
K
1
K
1
( )
i j

i mn
u
D
mnpq
pqj
u
i mn
u
D
mnpq
pqj
0
i mn
u
D
mnpq
pqj
u
+ + { } v d
V
0
}
=
i mn
0
i mn
u
D
mnpq
K
2
K
2
( )
i j
N
i k ,
N
j l ,
S
kl
V d
V
0
}
=
S
kl
N
i k ,
u
F R
S
i j
E
i j
S
i j
v
k
X
i
---------
k
X
j
------------- + dV
V
0
}
b
i
dV t
i
dA
A
0
}
+
V
0
}
=
S
i j
S
i j
E
kl
S
mn
E
pq
, , ( ) =
S
i j
D
i j kl
S
mn
E
pq
, ( )E
kl
=
109
CHAPTER 5
The second Piola-Kirchhoff stress for elastic and hyperelastic materials is a function of the Green-Lagrange strain
defined below:
(5-17)
If the stress is a linear function of the strain (linear elasticity)
(5-18)
the resulting set of equations is still nonlinear because the strain is a nonlinear function of displacement.
Updated Lagrangian Procedure
The Updated Lagrange formulation takes the reference configuration at . True or Cauchy stress and an
energetically conjugate strain measure, namely the true strain, are used in the constitutive relationship.
The updated Lagrangian approach is useful in:
a. analysis of shell and beam structures in which rotations are large so that the nonlinear terms in the curvature
expressions may no longer be neglected, and
b. large strain elasticity and plasticity analysis.
In general, this approach can be used to analyze structures where inelastic behavior (for example, plasticity,
viscoplasticity, or creep) causes the large deformations. The (initial) Lagrangian coordinate frame has little physical
significance in these analyses since the inelastic deformations are, by definition, permanent.
The LARGE STRAIN parameter generally specifies large strain analysis within the framework of Updated Lagrange
formulation. It signals Marc to calculate a geometric stiffness matrix and the initial stress stiffness matrix based on the
current deformed configuration.
For large strain analysis for rubber-like materials with incompressibility (such as materials defined with MOONEY,
OGDEN, GENT, and ARRUDBOYCE model definition options), Marc uses a mixed formulation, in which both the
displacement and the hydrostatic pressure are independent variables, to overcome the numerical difficulties resulting
from the volumetric constraints. For compressible hyperelastic materials defined with FOAM model definition option,
Marc uses conventional displacement formulation.
For large strain elastic-plastic analysis, the default procedure in Marc uses a procedure based on an additive
decomposition of incremental strain into an elastic part and a plastic part, together with a mean normal return-mapping
algorithm. In this case, volumetric strain in a lower-order plane strain, axisymmetric or 3-D brick element is assumed
to be constant for von Mises plasticity to overcome volumetric locking because of the possible large and
incompressible plastic deformation.
Marc can also use a procedure (LARGE STRAIN parameter with option 2) based on a multiplicative decomposition of
deformation gradient into an elastic part and a plastic part together with a radial return-mapping algorithm for large
strain elastic-plastic analysis. A mixed formulation is used to deal with the problem associated with volume
constraints. This procedure is only available for continuum elements.
Note:
Because large strain analysis involves nonlinearity, the CENTROID parameter must not be used with
this option.
S
i j
S
i j
E
kl
( ) =
S
i j
D
i j kl
E
kl
=
t n 1 + =

Because Herrmann elements do not support additive plasticity, Marc internally switches to the multiplicative
procedure. Herrmann elements have additional pressure degrees of freedom which increase numerical costs; hence, it
is generally more efficient to use displacement-based elements.
Marc uses Cauchy stress (true stress) and logarithmic strain with Updated Lagrange formulation.
It is instructive to derive the stiffness matrices for the updated Lagrangian formulation starting from the virtual work
principle in Equation (5-9).
Direct linearization of the left-hand side of Equation (5-9) yields:
(5-19)
where u and are actual incremental and virtual displacements respectively, and is Cauchy stress tensor.
(5-20)
denotes the symmetric part of , which represents the gradient operator in the current configuration. Also, in
Equation (5-19) and Equation (5-20), three identities are used:
(5-21)
in which represents the material moduli tensor in the reference configuration which is convected to the current
configuration, . This yields:
(5-22)
where is the material stiffness matrix written as
in which is the symmetric gradient operator-evaluated in the current configuration and is the Cauchy
stresses
S
i j
E
i j
( ) d ( ) V d
V
0
}

i k
kj
u
i j
v d
V
n 1 +
}
=
kj
S
i j
E
i j
d V d
V
0
}

s
i j
L
i j kl
s
u
kl
( ) v d
V
n 1 +
}
=
i j
1
J
--- F
i m
S
mn
F
j n
=
E
i j
F
mi
mn
F
nj
=
L
i j kl
1
J
--- F
i m
F
j n
F
kp
F
l q
D
mnpq
=
and
D
mnpq
L
i j k
K
1
K
2
+ { }u F R =
K
1
K
1
( )
i j

i mn
L
mnpq
pqj
V
n 1 +
}
=
i mn

kl
111
CHAPTER 5
and is the geometric stiffness matrix written as
while and are the external and internal forces, respectively.
Keeping in view that the reference state is the current state, a rate formulation analogous to Equation (5-14) can be
obtained by setting:
(5-23)
where is the deformation tensor, and is the rate of deformation. Hence,
(5-24)
in which and is the body force and surface traction, respectively, in the current configuration
In this equation, is the Truesdell rate of Cauchy stress which is essentially a Lie derivative of Cauchy stress
obtained as:
(5-25)
The Truesdell rate of Cauchy stress is materially objective implying that if a rigid rotation is imposed on the material,
the Truesdell rate vanishes, whereas the usual material rate does not vanish. This fact has important consequences in
the large deformation problems where large rotations are involved. The constitutive equations can be formulated in
terms of the Truesdell rate of Cauchy stress as:
(5-26)
Eulerian Formulation
In analysis of fluid flow processes, the Lagrangian approach results in highly distorted meshes since the mesh convects
with the material. Hence, an alternative formulation, namely Eulerian, is used to describe the motion of the body. In
this method, the finite element mesh is fixed in space and the material flows through the mesh. This approach is
particularly suitable for the analysis of steady-state processes, such as the steady-state extrusion or rolling processes
shown in Figure 5-11.
K
2
K
2
( )
i j

kl
N
i k ,
N
j l ,
v d
V
n 1 +
}
=
F R
F
i j

i j
E
i j
, d
i j

X
i
--------- ,

x
i
-------- S
i j
,
i j
= = = =
F d
i j
d
i j

i j
v
k
x
i
---------
k
x
j
------------- + dv
V
n 1 +
}
b
i
dv
V
n 1 +
}
t
i
da
A
n 1 +
}
+ =
b
i
t
i
i j
i j
F
i n
JF
nk
1
kl
F
ml
1
( )
= F
mj
i j L
i j k
d
k
=

Figure 5-11 Rolling Analysis
The governing differential equations of equilibrium for fluid flow through an enclosed volume are now written as:
(5-27)
where, is the material time derivative of a quantity and is the velocity of the particle flowing through the mesh.
For an incompressible fluid, Equation (5-27) along with continuity equation (mass conservation) yields:
(5-28)
The left-hand side of Equation (5-28) represents the local rate of change augmented by the convection effects. The
same principle can be called to physically explain the material time derivative of Cauchy stress; that is, Truesdell rate
of Cauchy stress. It can be seen from Equation (5-25) that:
(5-29)
The second and third terms on the right-hand side represent the convection effects. The last term vanishes for a
completely incompressible material; a condition enforced in the rigid-plastic flow of solids.
D v
i
( )
Dt
------------------ b
i
i j
x
j
---------- + =
D
Dt
------ v
v
i
t
------- - v
j

v
i
x
j
-------- + b
i
i j
x
j
---------- + =
i j

i j
v
i
x
k
---------
kj

i k

v
j
x
k
---------
i j

v
k
x
k
--------- + =
113
CHAPTER 5
5
Nonlinear Boundary Conditions
There are three types of problems associated with nonlinear boundary conditions: contact, nonlinear support, and
nonlinear loading. The contact problem might be solved through the use of special gap elements or the CONTACT
option. Nonlinear support might involve nonlinear springs and/or foundations. Sometimes nonlinearities due to rigid
links that become activated or deactivated during an analysis can be modeled through adaptive linear constraints.
Nonlinear loading is present if the loading system is nonconservative, as is the case with follower forces or frictional
slip effects.
Discontinuities are inherent in the nature of many of these nonlinearities, making the solution by means of incremental
linear approximations difficult. Some of the most severe nonlinearities in mechanics are introduced by nonlinear
boundary conditions. It is, therefore, very important to be aware of potential problem areas and to have a good
understanding of the underlying principles. This awareness and understanding enables you to validate numerical
answers and to take alternative approaches if an initial attempt fails.
Arbitrary Eulerian-Lagrangian (AEL) Formulation
In the AEL formulation referential system, the grid moves independently from the material, yet in a way that is spans
the material at any time. Hence, a relationship between derivative with respect to the material and grid derivative is
expressed as:
(5-30)
where is the relative velocity between the material particle, and the mesh velocity, ; for example,
(5-31)
The second or latter term represents the convective effect between the grid and the material. Note that for ,
a purely Eulerian formulation is obtained. The equation of momentum, for instance, can be represented as:
(5-32)
Due to its strong resemblance to the pure Eulerian formulation, AEL is also called quasi-Eulerian formulation.
Nonlinear Boundary Conditions
There are three types of problems associated with nonlinear boundary conditions: contact, nonlinear support, and
nonlinear loading. The contact problem might be solved through the use of special gap elements of the CONTACT
option. Nonlinear support might involve nonlinear springs and/or foundations. Sometimes nonlinearities due to rigid
links that become activated or deactivated during an analysis can be modeled through adaptive linear constraints.
Nonlinear loading is present if the loading system is nonconservative, as is the case with follower forces or frictional
slip effects.
( )
*
( ) c
i
( ) i , + =
c
i
v
i
p
v
i
m
c
i
v
i
p
v
i
m
=
v
i
p
v
i
m
=
v
i
*
v
j
p
v
j
m
( )
v
i
x
j
-------- +

i j
x
j
---------- b
i
+ =

Discontinuities are inherent in the nature of many of these nonlinearities, making the solution by means of incremental
linear approximations difficult. Some of the most severe nonlinearities in mechanics are introduced by nonlinear
boundary conditions. It is, therefore, very important to be aware of potential problem areas and to have a good
understanding of the underlying principles. This awareness and understanding enables you to validate numerical
answers and to take alternative approaches if an initial attempt fails.
Contact Problems
Contact problems are commonly encountered in physical systems. Some examples of contact problems are the
interface between the metal workpiece and the die in metal forming processes, pipe whip in piping systems, and crash
simulation in automobile designs.
Contact problems are characterized by two important phenomena: gap opening and closing and friction. As shown in
Figure 5-12, the gap describes the contact (gap closed) and separation (gap open) conditions of two objects
(structures). Friction influences the interface relations of the objects after they are in contact. The gap condition is
dependent on the movement (displacement) of the objects, and friction is dependent on the contact force as well as the
coefficient of (Coulomb) friction at contact surfaces. The analysis involving gap and friction must be carried out
incrementally. Iterations can also be required in each (load/time) increment to stabilize the gap-friction behavior.
Figure 5-12 Normal Gap Between Potentially Contacting Bodies
Two options are available in Marc for the simulation of a contact problem. A detailed description of these options (gap-
friction element and the CONTACT option) is given in Chapter 8 Contact of this manual.
Nonlinear Support
Marc provides two options for the modeling of support conditions: springs and elastic foundations. Both linear and
nonlinear springs can be specified in the input. In a nonlinear problem, the spring stiffness and the equivalent spring
stiffness of the elastic foundation can also be modified through a user subroutine. In the nonlinear spring option, the
incremental force in the spring is
(5-33)
where is the spring stiffness, is the displacement increment of the degree of freedom at the second end of the
spring, and is the displacement increment of the degree of freedom at the first end of the spring.
A
B
n
F K u
2
u
1
( ) =
K u
2
u
1
115
CHAPTER 5
Use the SPRINGS model definition option for the input of linear and nonlinear spring data. The USPRNG user
subroutine may also be used to specify the value of based on the amount of previous deformation for nonlinear
springs. In dynamic analysis, the SPRINGS option can also be used to define a dashpot. In thermal analysis or
electrical analysis (heat transfer, Joule heating, heat transfer pass of a coupled analysis), the SPRINGS option can be
used to define a thermal or electrical link.
In the elastic nonlinear FOUNDATION option, the elements in Marc can be specified as being supported on a
frictionless (nonlinear) foundation. The foundation supports the structure with an increment force per unit area given
by
(5-34)
where is the equivalent spring stiffness of the foundation (per unit surface area), and is the incremental
displacement of the surface at a point in the same direction as .
To input nonlinear foundation data, use the FOUNDATION model definition option.
To specify the value of for the nonlinear equivalent spring stiffness based on the amount of previous deformation
of the foundation, use the USPRNG user subroutine.
Nonlinear Loading
When the structure is deformed, the directions and the areas of the surface loads are changed. For most deformed
structures, such changes are so small that the effect on the equilibrium equation can be ignored. But for some structures
such as flexible shell structure with large pressure loads, the effects on the results can be quite significant so that the
surface load effects have to be included in the finite element equations.
Marc forms both pressure stiffness and pressure terms based on current deformed configuration with the FOLLOW
FOR parameter. The FOLLOW FOR parameter should be used with the LARGE DISP or LARGE STRAIN parameters.
The CENTROID parameter should not be included due to the use of the residual load correction.
Follower force point loads may be applied by either specifying this in the POINT LOAD option or by applying a
transformation which rotates with the displacements of the nodes specified in the COORD SYSTEM option.
A special case of nonlinear loading is the resultant pressure due to a gas in an enclosed cavity. In such problems, the
pressure changes as the volume changes based upon the ideal gas law. This is discussed in Chapter 9, in the Cavity
Pressure Loading section of this manual.
Buckling Analysis
Buckling analysis allows you to determine at what load the structure will collapse. You can detect the buckling of a
structure when the structures stiffness matrix approaches a singular value. You can extract the eigenvalue in a linear
analyses to obtain the linear buckling load. You can also perform eigenvalue analysis for buckling load in a nonlinear
problem based on the incremental stiffness matrices.
The buckling option estimates the maximum load that can be applied to a geometrically nonlinear structure before
instability sets in. To activate the buckling option in the program, use the BUCKLE parameter. If a nonlinear buckling
analysis is performed, also use the LARGE DISP parameter.
K
P
n
K u
n
( )u
n
=
K u
n
P
n
K

Use the BUCKLE history definition option to input control tolerances for buckling load estimation (eigenvalue
extraction by a power sweep or Lanczos method). You can estimate the buckling load after every load increment. The
BUCKLE INCREMENT option can be used if a collapse load calculation is required at multiple increments.
The linear buckling load analysis is correct when you take a very small load step in increment zero, or make sure the
solution has converged before buckling load analysis (if multiple increments are taken). Linear buckling (after
increment zero) can be done without using the LARGE DISP parameter, in which case the restriction on the load step
size no longer applies. This value should be used with caution, as it is not conservative in predicting the actual collapse
of structures.
In a buckling problem that involves material nonlinearity (for example, plasticity), the nonlinear problem must be
solved incrementally. During the analysis, a failure to converge in the iteration process or nonpositive definite stiffness
signals the plastic collapse.
For extremely nonlinear problems, the BUCKLE option cannot produce accurate results. In that case, the AUTO
INCREMENT history definition option allows automatic load stepping in a quasi-static fashion for both geometric large
displacement and material (elastic-plastic) nonlinear problems. The option can handle elastic-plastic snap-through
phenomena. Therefore, the post-buckling behavior of structures can be analyzed.
The buckling option solves the following eigenvalue problem by the inverse power sweep method:
(5-35)
where is assumed to be a linear function of the load increment to cause buckling.
The geometric stiffness used for the buckling load calculation is based on the stress and displacement state
change at the start of the last increment. However, the stress and strain states are not updated during the buckling
analysis. The buckling load is therefore estimated by:
(5-36)
where for increments greater than 1, is the load applied at the beginning of the increment prior to the
buckling analyses, and is the value obtained by the power sweep or Lanczos method.
The control tolerances for the inverse power sweep method are the maximum number of iterations in the power sweep
and the convergence tolerance. The power sweep terminates when the difference between the eigenvalues in two
consecutive sweeps divided by the eigenvalue is less than the tolerance. The Lanczos method concludes when the
normalized difference between all eigenvalues satisfies the tolerance. The maximum number of iterations and the
tolerance are specified through the BUCKLE history definition option.
K K
G
u u , , ( ) + [ ] 0 =
K
G
P
K
G
P beginni ng ( ) P +
P beginning ( )
117
CHAPTER 5
Perturbation Analysis
The buckling mode can be used to perform a perturbation analysis of the structure. In the manual mode, a buckling
increment is performed upon request and the coordinates are perturbed by a fraction of the buckling mode or
eigenvector. You can enter an individual eigenvector number and the fraction or can request that a combination of
modes be used. In the subsequent increments, the coordinates are:
or (5-37)
The manual mode can be activated by using the BUCKLE INCREMENT model definition, or BUCKLE load
incrementation option. In the automatic mode, the program checks for a nonpositive definite system during the
solution phase. When this occurs, it automatically performs a buckle analysis during the next increment and updates
the coordinates. The automatic mode can be activated by using the BUCKLE INCREMENT option. Also, be sure to
force the solution of the nonpositive definite system through the CONTROL option or PRINT parameter.
Material Nonlinearities
In a large strain analysis, it is usually difficult to separate the kinematics from the material description. Table 5-5 lists
the characteristics of some common materials.
Table 5-5 Common Material Characteristics
Material Characteristics Examples Marc Models
Composites Anisotropic:
1) layered,
21 constants
2) Fiber reinforced,
one-dimensional strain in fibers
Bearings, aircraft panels
Tires, glass/epoxy
Composite continuum
elements
Rebars
Creep Strains increasing with time under
constant load.
Stresses decreasing with time under
constant deformations.
Creep strains are non-instantaneous.
Metals at high temperatures,
polymide films,
semiconductor materials
ORNL
Norton
Maxwell
Elastic Stress functions of instantaneous strain
only. Linear load-displacement relation.
Small deformation (below
yield) for most materials:
metals, glass, wood
Hookes Law
Elasto-plasticity Yield condition flow rule and hardening
rule necessary to calculate stress, plastic
strain. Permanent deformation
upon unloading.
Metals
Soils
Snow
Wood
von Mises Isotropic
Cam -Clay
Hills Anisotropic
Generalized Mohr-
Coulomb
X X f

------ + = X X f
i
i
--------
+ =
ds
i j
C
i j k
d
k
=
S
E
2
--- T
t
CT 1 ( ) =

A complete description of the material types mentioned is given in Chapter 7 of this manual. However, some no
characteristics and procedural considerations of some commonly encountered materials behavior are listed next.
Inaccuracies in experimental data, misinterpretation of material model parameters and errors in user-defined material
law are some common sources of error in the analysis from the materials viewpoint. It is useful to check the material
behavior by running a single element test with prescribed displacement and load boundary conditions in uniaxial
tension and shear.
Large Strain Elasticity
Structures composed of elastomers, such as tires and bushings, are typically subjected to large deformation and large
strain. An elastomer is a polymer, such as rubber, which shows a nonlinear elastic stress-strain behavior. The large
strain elasticity capability in Marc deals primarily with elastomeric materials. These materials are characterized by the
form of their elastic strain energy function. For a more detailed description of elastomeric material, see Elastomer in
Chapter 7 on this manual.
For the finite element analysis of elastomers, there are some special considerations which do not apply for linear elastic
analysis. These considerations, discussed below, include:
Large Deformations
Incompressible Behavior
Instabilities
Existence of Multiple Solutions
Large Deformations
The formulation is complete for arbitrarily large displacements and strains.
Hyperelastic Stress function of instantaneous strain.
Nonlinear load-displacement relation.
Unloading path same as loading.
Rubber Mooney
Ogden
Arruda-Boyce
Gent
NLELAST
Hypoelastic Rate form of stress-strain law Concrete Buyukozturk
Viscoelastic Time dependence of stresses in elastic
material under loads. Full recovery after
unloading.
Rubber, Glass, industrial
plastics
Simo Model
Narayanaswamy
Viscoplastic Combined plasticity and
creep phenomenon
Metals
Powder
Power law
Shima Model
Shape Memory Superelastic and shape memory effect
with phase transformations.
Biomedical stents, Satellite
antennae
Aurrichio, Thermo-
mechanical
Table 5-5 Common Material Characteristics (continued)
Material Characteristics Examples Marc Models
119
CHAPTER 5
When extremely large deformations occur, the element mesh should be designed so that it can follow these
deformations without complete degeneration of elements. For problems involving extreme distortions, rezoning must
be done. Rezoning can be used with the formulation in the updated Lagrangian framework using conventional
displacement based elements.
Incompressible Behavior
One of the most frequent causes of problems analyzing elastomers is the incompressible material behavior. Lagrangian
multipliers (pressure variables) are used to apply the incompressibility constraint. The result is that the volume is kept
constant in a generalized sense, over an element.
Both the total, as well as updated Lagrange formulations, are implemented with appropriate constraint ratios for lower-
and higher-order elements in 2-D and 3-D. For many practical analysis, the LBB (Ladyszhenskaya-Babuska-Brezzi)
condition does not have to be satisfied in the strictest sense; for example, four node quadrilateral based on Herrmann
principle.
For elements that satisfy the LBB condition, error estimates of the following form can be established
(5-38)
where and are the orders of displacements and pressure interpolations, respectively. If , the
rate of convergence is said to be optimal, and elements satisfying the LBB condition would not lock.
The large strain elasticity formulation may also be used with conventional plane stress, membrane, and shell elements.
Because of the plane stress conditions, the incompressibility constraint can be satisfied without the use of Lagrange
multipliers.
Instabilities
Under some circumstances, materials can become unstable. This instability can be real or can be due to the
mathematical formulation used in calculation.
Instability can also result from the approximate satisfaction of incompressibility constraints. If the number of
Lagrangian multipliers is insufficient, local volume changes can occur. Under some circumstances, these volume
changes can be associated with a decrease in total energy. This type of instability usually occurs only if there is a large
tensile hydrostatic stress. Similarly, overconstraints give rise to mesh locking and inordinate increase in total energy
under large compressive stresses.
Existence of Multiple Solutions
It is possible that more than one stable solution exists (due to nonlinearity) for a given set of boundary conditions. An
example of such multiple solutions is a hollow hemisphere with zero prescribed loads. Two equilibrium solutions
exist: the undeformed stress-free state and the inverted self-equilibrating state. An example of these solutions is shown in
Figure 5-13 and Figure 5-14. If the equilibrium solution remains stable, no problems should occur; however, if the
equilibrium becomes unstable at some point in the analysis, problems can occur.
u
h
u
1
p
h
p
0
+ O h
mi n k 1 + , { }
( ) =
k K min k 1 + , { } =

Figure 5-13 Rubber Hemisphere
Figure 5-14 Inverted Rubber Hemisphere
When incompressible material is being modeled, the basic linearized incremental procedure is used in conjunction
with mixed variational principles similar in form to the Herrmann incompressible elastic formulation. These
formulations are incorporated in plane strain, axisymmetric, generalized plane strain, and three-dimensional elements.
These mixed elements may be used in combination with other elements in the library (suitable tying may be necessary)
and with each other. Where different materials are joined, the pressure variable at the corner nodes must be uncoupled
to allow for mean pressure discontinuity. Tying must be used to couple the displacements only.
Large Strain Plasticity
In recent years there has been a tremendous growth in the analysis of metal forming problems by the finite element
method. Although an Eulerian flow-type approach has been used for steady-state and transient problems, the updated
Lagrangian procedure, pioneered by McMeeking and Rice, is most suitable for analysis of large strain plasticity
problems. The main reasons for this are: (a) its ability to trace free boundaries, and (b) the flexibility of taking elasticity
and history effects into account. Also, residual stresses can be accurately calculated.
y
x
y
x
121
CHAPTER 5
The large strain plasticity capability in Marc allows you to analyze problems of large-strain, elastic-plastic material
behavior. These problems can include manufacturing processes such as forging, upsetting, extension or deep drawing,
and/or large deformation of structures that occur during plastic collapse. The analysis involves both material,
geometric and boundary nonlinearities.
In addition to the options required for plasticity analysis, the LARGE STRAIN parameter is needed for large strain
plasticity analysis.
In performing finite deformation elastic-plastic analysis, there are some special considerations which do not apply for
linear elastic analysis include:
Choice of Finite Element Types
Nearly Incompressible Behavior
Treatment of Boundary Conditions
Severe Mesh Distortion
Instabilities
Choice of Finite Element Types
Accurate calculation of large strain plasticity problems depends on the selection of adequate finite element types. In
addition to the usual criteria for selection, two aspects need to be given special consideration: the element types
selected need to be insensitive to (strong) distortion; for plane strain, axisymmetric, and three-dimensional problems,
the element mesh must be able to represent nondilatational (incompressible) deformation modes.
Nearly Incompressible Behavior
Most finite element types tend to lock during fully plastic (incompressible) material behavior. A remedy is to introduce
a modified variational principle which effectively reduces the number of independent dilatational modes (constraints)
in the mesh. This procedure is successful for plasticity problems in the conventional small strain formulation.
Zienkiewicz pointed out the positive effect of reduced integration for this type of problem and demonstrates the
similarity between modified variational procedures and reduced integration. The lower-order elements, invoking the
constant volumetric strain or the lower-order elements, using reduced integration and hourglass control, behave well
for nearly incompressible materials. Higher order elements in Marc are formulated to be used in large strain analysis
including contact.
Treatment of Boundary Conditions
In many large strain plasticity problems, specifically in the analysis of manufacturing processes, the material slides
with or without friction over curved surfaces. This results in a severely nonlinear boundary condition. The Marc gap-
friction element and CONTACT option can model such sliding boundary conditions.
Severe Mesh Distortion
Because the mesh is attached to the deforming material, severe distortion of the element mesh often occurs, which
leads to a degeneration of the results in many problems. The ERROR ESTIMATE option can be used to monitor this
distortion. To avoid this degeneration, generate a new finite element mesh for the problem and then transfer the current
deformation state to the new finite element mesh. The global adaptive and rezoning procedure in the program is
specifically designed for this purpose.

Instabilities
Elastic-plastic structures are often unstable due to necking phenomena. Consider a rod of a rigid-plastic
incompressible workhardening material. With the current true uniaxial strain rate and the current workhardening,
the rate of true uniaxial stress is equal to
(5-39)
The applied force is equal to , where is the current area of the rod. The rate of the force is therefore
equal to
(5-40)
On the other hand, conservation of volume requires that
(5-41)
Hence, the force rate can be calculated as
(5-42)
Instability clearly occurs if . For applied loads (as opposed to applied boundary conditions), the stiffness matrix
becomes singular (nonpositive definite).
For the large strain plasticity option, the workhardening slope for plasticity is the rate of true stress versus the true
plastic strain rate. The workhardening curve must, therefore, be entered as the true stress versus the logarithmic plastic
strain in a uniaxial tension test.
Computational Procedures for Elastic-Plastic Analysis
Three basic procedures for plasticity exist in Marc. In this section, the variational form of equilibrium equations and
constitutive relations, and incompressibility are summarized. Issues regarding return mapping procedures for stress
calculation in three-dimensional and plane-stress conditions are also discussed.
For notational purpose, three configurations are considered at any point, original ( ), previous ( ) and
current ( ). An iterative procedure, full or modified Newton-Raphson, secant, or arc-length is used to solve
for the equilibrium at .
1. Small Strain Plasticity (reference configuration: ):
In this approach, the basis of variational formulation are 2nd Piola-Kirchhoff stress, and Green-Lagrange
strain, . Equilibrium of the current state can thus be represented by the following virtual work principle:
(5-43)
During the increment, all state variables are defined with respect to the state at and are updated at the
end of increment upon convergence.
=
F A = A
F
A A
+ =
A
+ 0 =
F
H ( )A
=
H >
t 0 = t n =
t n 1 + =
t n 1 + =
t 0 =
S
E
S : E V d
V
n
}
t : dA b : V d
V
n
}
+
A
n
}
=
t n =
123
CHAPTER 5
The linearized form of constitutive equations is given as:
(5-44)
in which is the second Piola-Kirchhoff stress, is the Cauchy stress, and is the elasto-plastic moduli.
This linearization has the advantage that it is fully independent of the rotation increment, but the disadvantage
is that the linearization causes errors equal to the square of the strain increment. Moreover, imposing the
incompressibility condition in terms of the trace of Green-Lagrange strains leads to errors in the form of
fictitious volume changes in fully developed plastic flow.
The above procedure works well in the context of small strain plasticity. However, in many large deformation
problems including metal forming processes, the plastic strain increments can be very large and the above
procedure can lead to large errors in the results. The two finite strain plasticity formulations to model the large
inelastic strains are: rate based (hypoelastic) and total (hyperelastic).
2. Finite Strain Plasticity with additive decomposition of strain rates (reference configuration: ).
This formulation is based on the integration of the constitutive equations in the current configuration. To
maintain objectivity, the notion of rotation neutralized stress and strain measures is introduced. All objective
stress rates, are manifestation of Lie derivative:
(5-45)
where is a deformation measure; for example, deformation gradient or rotation tensor, while is a stress
measure in the current configuration. The general form for an objective stress rate is:
(5-46)
From Table 5-6, it can be seen that there is a possibility of a number of stress rates. It can be observed that while
all the above stress rates are objective, the Truesdell and Durban-Baruch rates would not yield symmetric
matrices.
Table 5-6 Objective Stress Rates
Stress Rates
Truesdell L 1
Cotter-Rivlin L 0
Oldroyd
-L
T
0
Jaumann-Zaremba-Noll W 0
Green-McInnis-Nagdhi 0
Durban Baruch 1
dS L
ep
: dE . dE dE. =
S L
ep
t n 1 + =
L
v
( )
1
T
( )
T
=
R
T
tr d ( ) + =
R
R
1
1
2
--- D W + ( )

Marcs implementation of rate formulation involves the use of Jaumann rate of Cauchy stress which is obtained
as an average of the Oldroyd and Cotter-Rivlin stress rates. Thus, the Jaumann rate can be written as:
(5-47)
(5-48)
where ( ) is the ordinary rate and ( ) is the objective rate with is the spin or the antisymmetric part of the
velocity gradient, . The last term is neglected because of the incompressible nature of plasticity.
The equilibrium in the current state is given by the virtual work principle at :
(5-49)
Linearization of the above form leads to the variational statement:
(5-50)
where,
(5-51)
Within the context of rotation neutralized form of constitutive relations, is defined as:
(5-52)
where, . Also, the rotation neutralized strain can be calculated by:
with being the mid-increment strain, a good approximation for incremental logarithmic strain measure.
Where is the engineering strain and is the stretch tensor obtained by the polar decomposition
of . Admitting an error of the order of in the approximation of the logarithmic strain, one obtains:
(5-53)
where (5-54)
W W +
=
T
t r d ( ) + + + =

W
L
t n 1 + =
: v d
V
n+1
}
t . da b . v d
V
n+1
}
+
A
n+1
}
=
d : 2 d ( ) : tr d ( ) : : u ( ) ( )
T
{ } ] + + [ v d
V
n+1
}
=
d b . v t . a d
A
n+1
}
+ d
V
n+1
}
\ .
|
| |
V
n+1
}
: dv
d R d
RN
R
T
=
d
RN
d
RN
L
RN
: d
RN
=
d
RN
R
T
d R =
RN
U I + ( )
T
F I + ( )
MI D
F I + ( ) U I + ( )
1
=
MI D
e U I = U
F e
2
d L
ep
: d =
L
ep
R R L
RN
R
T
( ) R
T
=
125
CHAPTER 5
During computations, the third term of Equation (5-50), , is neglected due to its nonsymmetric
nature.
In a fully developed plastic flow, the volumetric part of the energy can become extremely large and lead to
volumetric locking. Hence, a special treatment of incompressibility is done to relax the volumetric constraint
in an assumed strain format. The volumetric part of deformation gradient is modified such that, the assumed
deformation gradient is:
(5-55)
Linearization of in the original state relates it to the displacement gradients in the current state as:
(5-56)
where
(5-57)
where the first two terms are evaluated at each integration point and the last term is averaged over the element.
For a lower order element, the procedure leads to mean or constant dilatation approach. Considering Equations
(5-50) to (5-57) the resulting system can be expressed as:
(5-58)
In the event, the elastic strains become large in an elastic-plastic analysis the rate based constitutive equations
do not accurately model the material response. This results from the fact that the elasticity matrix are assumed
to have constant coefficients in the current deformed configuration.
In the next formulation, the rate of deformation tensor is decomposed multiplicatively into the elastic and
plastic parts to resolve these problems.
3. Finite strain plasticity with multiplicative decomposition of deformation gradient.
An alternative formulation, based on the multiplicative decomposition of the deformation gradient has been
implemented in Marc, namely:
(5-59)
where , , and are (elastic, thermal, and plastic) deformation gradients, respectively. The thermo-
mechanical coupling is implemented using the staggered approach.
The above decomposition has a physical basis to it as the stresses are derived from quadratic - logarithmic strain
energy density function:
t r d ( ) :
F J
1
3
---
J
1
3
---
F =
F
DF u ( ) F =
u ( ) u ( )
i j ,
1
3
--- div u ( )
i j
1
3
--- di v u ( )
i j
+ =
s
: L
ep
:
s
u ( ) 2
s
:
s
u ( ) : u ( ) ( )
T
+ [ ] v d
V
n+1
}
=
d b . v t . a d
A
n+1
}
+ d
V
n+1
}
\ .
|
| |
s
[ ]
T
: v d
V
n 1 +
}
F F
e
F
F
p
=
F
e
F
F
p

(5-60)
This function has been chosen due to the availability of the material coefficients from material testing in a small
strain case.
In the metal forming applications, the elastic strains are negligible and the rate-based (or hypoelastic) as well
as the total (or hyperelastic) form of constitutive equations give virtually the same results. However, many
polymers, metals subjected to large hydrostatic pressures, high velocity impact loading of metals and processes,
where shape changes after deformation need to be evaluated precisely, the hyperelastic formulation yields
physically more meaningful results.
The return mapping procedure for the calculation of stresses is based on the radial return procedure. With the
use of exponential mapping algorithm, the incompressibility condition is imposed exactly:
(5-61)
The strain energy is separated into deviatoric and volumetric parts in the framework of mixed formulation. The
general form of three-field variational principle is:
(5-62)
where, (5-63)
and are the deviatoric and elastic volumetric parts of the free energy, p is the pressure, and
are elastic pointwise and average elastic Jacobian of the element, respectively. The volumetric free energy
can be of any form in Equation (5-61). However, if the incompressibility is enforced in a pointwise fashion and
the volumetric free energy is assumed to be of the form:
(5-64)
the perturbed Lagrangian form of the variational principle can be cast in a two-field framework as:
(5-65)
W
1
2
--- J
e
ln ( )
2
tr
1
2
--- b
e
ln
\ .
| |
2
+ =
det F
p
1 =
u p J , , ( ) [W
V
0
}
b ( ) U J
e
( ) p J
e
J
e
( )] + + = dV
b J
2
3
---
b =
W b ( ) U J
e
( ) J
e
J
e
U J
e
( )
9
2
--- K J
e
( )
1
3
---
1
2
=
u P , ( ) W b ( ) 3P J
e
( )
1
3
---
1
)

`

P
2
2K
------- + V d
V
0
}
=
127
CHAPTER 5
Note that , hence, P is not the true pressure in Equation (5-65). Choice of the volumetric free
energy in Equation (5-64) is not arbitrary and is described in more detail in Chapter 7 in the Elastomer section
of this manual.
Linearization of the equilibrium condition arising from the stationary of the variational principle Equation
(5-65), yields:
(5-66)
The linearization of the constraint equation defines the term in Equation (5-66).
The derivation of the spatial tangent follows by a push forward of the material tangent into the current
configuration:
Hence, (5-67)
where is the pull back of the deviatoric Kirchhoff stress tensor. Assuming the entire incremental deformation
to be elastic, a symmetric stress tensor, can be defined with respect to a fixed plastic intermediate
configuration as:
(5-68)
also noting that,
where (5-69)
(5-70)
P p J
e
( )
2
3
---
=
s
:
1
J
--- C
dev
P J
e
( )
1
3
---
1
3
--- 1 1 2I
\ .
| |
+
\ .
|
|
| |
)

`

:
s
u +
V
n+1
}
: u ( ) ( ). ( )
T
{ }
dP
J
------ J
e
( )
1
3
---
1 : + dv =
d b
V
n+1
}
. v t
A
n+1
}
. a d + d
\ .
|
| |
s
( )
T
V
n+1
}
: v d
dP
C
dev
S F
1
dev
F
T
=
S
S
F
n
p
S
F
e
t r
( )
1
dev
F
e
t r
( )
T
=
S
S
A
N
A
N
A
A 1 =
3
= S
Ae
t r
( )
2
----------------- =
dS
dS
A
N
A
N
A
S
B
S
A
( )
BA
N
A
N
B
B A
3
A 1 =
3
+
A 1 =
3
=

where (5-71)
and (5-72)
similarly, with (5-73)
(5-74)
The algorithmic elasto-plastic tangent can be obtained as:
(5-75)
Combining Equations (5-70), (5-75), and (5-76), the spatial form of deviatoric part of the tangent as:
(5-76)
The singularity for two- or three-equal stretch ratios is removed by repeated application of LHospitals rule.
Due to separate interpolation of the pressure, a larger system of equations needs to be solved. A static
condensation of the element variables is carried out before going in the solver. This step renders the singularity
of the system to be inversely proportional to the bulk modulus of the material. Due to the multiplicative split
of the deformation gradient and an additive split of the free energy into deviatoric and volumetric parts, this
framework is eminently suitable for implementation of general nonlinear elastic as well as inelastic material
models. Besides, compressible and nearly incompressible material response can be naturally handled.
dS
A
1
Ae
t r
( )
2
Be
t r
( )
2
-----------------------------------
Be
t r
------------ 2
A

AB
\ .
|
| |

Be
t r
d
Be
t r
=
dN
A

AB
N
B
=
C
F
e
t r
( )
T
F
e
t r
= F
e
t r
Ae
t r
n
A
N
A
A 1 =
3
=
dC
2
Be
t r
d
Be
t r
N
B
N
B
+
B 1 =
3
=

Be
t r
( )
2
Be
t r
( )
2
)
`

BA
N
A
N
B
B A
3
A 1 =
3
C
dev
2F F
S
C
------- F
T
\ .
| |
F
T
2 F
e
t r
( ) F
e
t r
( )
S
------- F
e
t r
( )
T
F
e
t r
( )
T
= =
C
dev
Be
t r
------------ 2
A
AB
\ .
|
| |
n
A
n
A
n
B
n
B

B A
3
A 1 =
3
Ae
t r
( )
2
B

Be
t r
( )
2
Be
t r
( )
2
Ae
t r
( )
2
-------------------------------------------------------- n
A
( n
B
n
A
n
B
+
B A
3
A 1 =
3
+
n
A
n
B
n
B
n
A
)
129
CHAPTER 5
Finally, the stress calculation is done via return mapping to the yield surface. There are several methods to
return the stress. Among them, the popular ones include:
Closest point projection (which reduces to radial return for von Mises), mean normal approach, midpoint
rule, trapezoidal rule, tangent cutting plane algorithm, and multistage method.
Krieg and Krieg (1977) investigated the radial return algorithm which has been proposed by Wilkins
(1964). The third algorithm analyzed was the so-called secant stiffness method proposed by Rice and
Tracy (1973) which represented a special case of the generalized Trapezoidal rule for nonhardening
materials algorithm in contrast to the fully implicit radial return algorithm and the fully explicit tangent
stiffness approach. The radial return method (shown in Figure 5-15) was found to be the most accurate
among the three analyzed, particularly when large strain increments were used. The radial return and mean
normal method are available in Marc and are described next.
Figure 5-15 Radial Return Method
1. Closest Point Projection procedure:
The closest point projection or the backward Euler algorithm reduces to a radial return scheme for cases where
there is no anisotropy or plane stress condition. The trial stress is determined as:
(5-77)
The final stress state is obtained by simply scaling the trial stress by a scale factor,
where, (5-78)
the plastic consistency parameter , is obtained by solving for consistency condition iteratively.
An explicit form of the material tangent for the isotropic, von Mises yield surface can be given as:
(5-79)
fi nal *
el
r r
d
d
2G
d
+ =
1
2G
r r
d
--------------- =
L
ep
2G I
1
3
--- 1 1
\ .
| |
2G N
=

where, (5-80)
Here is the return direction normal to the current yield surface. For the plane stress case, the zero normal
stress condition is explicitly imposed in the yield condition. Since the yield surface is not a circle in the plane,
the return direction is not radial anymore.
2. Mean-Normal method:
Assuming the entire deformation to be elastic in nature, the trial stress is evaluated as:
(5-81)
Due to the use of pressure independent yield function, only the deviatoric stresses are evaluated. As shown in
Figure 5-16, represents the fraction of strain increment for which the plastic flow occurs.
Figure 5-16 Mean-Normal Method
includes the deviatoric stresses at the previous increment scaled such that it satisfies the yield condition.
The yield criterion might not be exactly satisfied due to temperature effects, numerical integration of elastic-
plastic relationships, or accumulated numerical inaccuracy.
indicates that the stress state is exactly on the yield surface. If , scale by a factor
such that:
(5-82)
The equivalent plastic strain, plastic strain tensor and the elasto-plastic moduli are obtained as:
(5-83)
1
1
H
3G
------- +
------------------ 1 + =
N
mn
d
d
2G
d
1 ( )G
d
+ + =
1 ( )
1
d
2
d
3
d
el
*
o

el
+ +
fi nal
*
d
F ( ) 0 = F ( ) 0 >
F ( ) 0 0 1 < < =
p C
e
: N
( ) :
1 ( )
N
: C
e
: N
H +
----------------------------------------------------- =
131
CHAPTER 5
(5-84)
(5-85)
where, is the mean-normal direction to the yield surface, is the hardening modulus and is the elastic
tangent moduli. For nonhardening materials, the mean normal algorithm is a special case of trapezoidal rule.
Marc plasticity algorithms are unconditionally stable and accurate for moderate strain increments. However,
overall global stability of the system can be dictated by other considerations like contact, buckling which then
guide the selection of appropriate load increment size. It should also be recognized that the algorithmically
consistent tangent in the closest point projection algorithm is based on the current state, and a lack of embedded
directionality (unlike secant methods) in the tangent can lead to divergent solutions unless line search or
automatic time stepping algorithms are used. However, when it converges it shows quadratic convergence as
compared to linear convergence for the mean-normal scheme.
3. Multistage Return Mapping:
The return mapping scheme described in this section is used for isotropic and anisotropic plasticity when
LARGE STRAIN is used to describe the Updated Lagrange Procedure. Marc uses this scheme for the Hill and
Barlat yield criteria. This scheme is not available for pressure dependent yield criteria like Mohr Coulomb. In
a large strain analysis, Marc automatically chooses the return mapping scheme that is most suited for the given
material.
The increment of Cauchy stress for elasto-plasticity becomes
. (5-86)
In Equation (5-86), superscript means a materially embedded co-rotational coordinate system. For the
numerical implementation of Equation (5-86), the general midpoint rule can be expressed as follows:
(5-87)
where
(Case 1) if , ,
(Case 2) if , such that
where superscript stands for a trial state. For Case 2, the condition stating that the updated stress stays on
the strain-hardening curve ( ) provides the following condition:
. (5-88)
Equation (5-88) is a nonlinear equation to solve for .
p
N
=
L
ep
C
e
1 ( )
C
e
: N
( ) C
e
: N
( )
N
: C
e
: N
H +
--------------------------------------------------- =
N
H C
e
e
:
e
C
e
:
p
( ) = =
n 1 +

n 1 +
T
C
e
N
n +
=
n 1 +
T
n
C
+ =
F
n 1 +
T
( ) 0 0 =
n 1 +

n 1 +
T
=
F
n 1 +
T
( ) 0 = F
n 1 +
T
( ) 0 =
T

p
( ) =
F ( )
n 1 +
T
C
e
N
n +
( )
n
p
+ ( ) 0 = =

5+
If the strain increment is not small enough, it is difficult to obtain the numerical solution of Equation (5-88)
even though it has a mathematical solution. Therefore, an iterative scheme, which utilizes the control of the
potential residual, is introduced. The method is applicable to a nonquadratic yield function and a general
hardening law. For the implementation of the algorithm, Equation (5-88) has been modified to the following
relationship with ; that is,
(5-89)
where
, ,
and are prescribed values.
As shown in Equation (5-89) and Figure 5-17, the direction of the first substep is guessed from the
direction , which is normal to the yield surface at the trial stress . Then, the exact direction can
be obtained from the first substep nonlinear solution based on Euler backward method. In general, the new
direction for the second substep is guessed from the direction based on the previous substep
stress . This procedure is completed when . In fact, the equation for the N
th
step is
equivalent to Equation (5-89) and . Finally, the proportional logarithmic plastic strain remains
normal to the yield surface at the final stress ; that is,
. (5-90)
Figure 5-17 Schematic View for Multistage Return Mapping Method
1 =
F
k ( )
( )
T
k ( )
C
e
N
k ( )
( )
n
p
k ( )
+ ( ) F
k ( )
= =
F
0 ( )
0 = ( ) F
0 ( )
= F
k ( )
F
0 ( )
F
1 ( )
F
k ( )
F
N ( )
F
N ( )
0 = ( ) k 0 N = , > > > > { }
F F
k 1
F
k
( )
Y

p
0 = ( ) ( ) = ( ) < = F
k 1 N 1 =
N
1 ( )
N
0 ( )

T
N
1 ( )
N
i 1 + ( )
N
i ( )
i ( )
F F
N ( )
0 = ( ) =
n 1 +

N ( )
=
n 1 +
-------
n 1 +
( )
N ( )
N
N ( )
= ( ) =

T
( )

1 ( )
( )

2 ( )
( )

n 1 + ( )
( )

n ( )
( )
n 1 +
2 ( )
p
2 ( )
p
N ( )
p
T
N
1 ( )
N
2 ( )
N
N ( )
133
CHAPTER 5
Therefore, the normality condition of the incremental deformation theory is satisfied at the current state
for .
In order to solve Equation (5-89), the Euler Backward method is employed to solve the iteration procedure for
the k
th
substep. At each iteration, (at substep and iteration) becomes
(5-91)
where
,
,
,
,
and is the hardening modulus in stress-strain curve.
The detailed derivations of Equation (5-91) are shown in the work of Yoon at al. [Ref. 31].
In order to solve the equilibrium equation iteratively, the elasto-plastic tangent modulus consistent with the
current return mapping method is obtained as follows:
(5-92)
where and .
In Equation (5-92), is instantaneous slope and is used.
Creep
Creep is a time-dependent inelastic behavior that can occur at any stress level, either below or above the yield stress
of a material. Creep is an important factor at elevated temperatures. In many cases, creep is also accompanied by
plasticity, which occurs above the yield stress of the material.
n 1 + ( ) 1 =
k ( )
i ( )
k ( ) i ( )
k ( )
i ( )
g
1

k ( )
i ( )
\ .
| |
N
k ( )
i ( )
E
k ( )
i ( )
1
g
2

k ( )
i ( )
\ .
| |
g
3

k ( )
i ( )
\ .
| |
H +
N
k ( )
i ( )
E
k ( )
i ( )
1
N
k ( )
i ( )
H +
-------------------------------------------------------------------------------------------------------------------- =
E
k ( )
i ( )
1
C
e 1
k ( )
i ( )
N
------- +
1
=
g
1

k ( )
i ( )
\ .
| |

( )
n
p
k ( )
i ( )
+
\ .
| |
F
k ( )
=
g
2

k ( )
i ( )
\ .
| |
C
e 1
T
( )
k ( )
i ( )
N
+ =
g
3

k ( )
i ( )
\ .
| |
H
1

n
( )
k ( )
i ( )
=
H
d
j
d
n 1 +
L
ep
d
n 1 +
= =
L
ep
C
CN
n +
CN
n +
n +
CN
n +
h +
-------------------------------------------------
\ .
|
| |
= C C
e 1
n +
n 1 +
------------------- +
1
=
h 1 =

Marc offers two schemes for modeling creep in conjunction with plasticity:
a. treating creep strains and plastic strains separately using an explicit procedure (where the creep is treated
explicitly) or an implicit procedure (where both creep and plasticity are treated implicitly). These procedures
are available with standard options via data input or with user-specified options via user subroutines. More
details are provided below.
b. modeling creep strains and plastic strains in a unified fashion (viscoplasticity). Both explicit and implicit
procedures are again available for modeling unified viscoplasticity. More details are provided in the section
titled Viscoplasticity in this chapter.
The options offered by Marc for modeling creep are as follows:
Creep data can be entered directly through data input or user subroutine. For explicit creep, the CRPLAW user
subroutine is to be used, and for implicit creep, the UCRPLW user subroutine is to be used.
An automatic time stepping scheme can be used to maximize the time step size in the analysis.
Eigenvalues can be extracted for the estimation of creep buckling time.
In addition, for explicit creep, the following additional options can be used:
Creep behavior can be either isotropic or anisotropic.
The Oak Ridge National Laboratory (ORNL) rules on creep can be activated.
The creep analysis option is activated in Marc through the CREEP parameter. The creep time period and control
tolerance information are input through the AUTO CREEP history definition option. This option can be used
repeatedly to define a new creep time period and new tolerances. These tolerances are defined in the section on Creep
Control Tolerances. Alternatively, a fixed time step can be specified through the CREEP INCREMENT history
definition option. In this case, no additional tolerances are checked for controlling the time step.
Creep analysis is often carried out in several runs using the RESTART option. Save restart files for continued analysis.
The REAUTO option allows you to reset the parameters defined in the AUTO CREEP option upon restart.
Adaptive Time Control
The AUTO CREEP option takes advantage of the diffusive characteristics of most creep solutions. Specifically, this
option controls the transient creep analysis. You specify a period of creep time and a suggested time increment. The
program automatically selects the largest possible time increment that is consistent with the tolerance set on stress and
strain increments (see Creep Control Tolerances in this chapter).
The algorithm is: for a given time step , a solution is obtained. Marc then finds the largest values of stress change
per stress, , and creep strain change per elastic strain, . It compares these values to the tolerance
values, (stress change tolerance) and (strain change tolerance), for this period.
The value is calculated as the larger of
(5-93)
t

cr
el
T
s
T
e
p
( ) T
cr
el
( ) T
or
135
CHAPTER 5
If , the program resets the time step as
(5-94)
The time increment is repeated until convergence is obtained or the maximum recycles control is exceeded. In the latter
case, the run is ended.
If the first repeat does not satisfy tolerances, the possible causes are:
excessive residual load correction
strong additional nonlinearities such as creep buckling-creep collapse
incorrect coding in the CRPLAW, VSWELL, or UVSCPL user subroutine.
Appropriate action should be taken before the solution is restarted.
If , the solution is stepped forward to and the next step is begun. The time step used in the next increment
is chosen as
(5-95)
(5-96)
(5-97)
Since the time increment is adjusted to satisfy the tolerances, it is impossible to predetermine the total number of time
increments for a given total creep time.
Creep Control Tolerances
Marc performs a creep analysis under constant load or displacement conditions on the basis of a set of tolerances and
controls you provide.These are as follows:
1. Stress change tolerance This tolerance controls the allowable stress change per time step during the creep
solution, as a fraction of the total stress at a point. Stress change tolerance governs the accuracy of the transient
creep response. If you need accurate tracking of the transient response, specify a tight tolerance of 1 percent or
2 percent stress change per time step. If you need only the steady-state solution, supply a relatively loose
tolerance of 10-20 percent. It is also possible to check the absolute rather than the relative stress.
2. Creep strain increment per elastic strain Marc uses either explicit or implicit integration of the creep rate
equation. When the explicit procedure is used, the creep strain increment per elastic strain is used to control
stability. In almost all cases, the default of 50 percent represents the stability limit, so that you need not provide
any entry for this value. It is also possible to check the absolute rather than the relative strain.
3. Maximum number of recycles for satisfaction of tolerances During AUTO CREEP, Marc chooses its own time
step. In some cases, the program recycles to choose a time step that satisfies tolerances, but recycling rarely
occurs more than once per step. Excessive recycling can be caused by physical problems such as creep
buckling, poor coding of the CRPLAW, VSWELL, or UVSCPL user subroutine or excessive residual
load correction that can occur when the creep solution begins from a state that is not in equilibrium.
p 1 >
t
new
0.8 t
ol d
p =
p 1 < t t +
t
new
t
ol d
= if 0.8 < p < 1
t
new
1.25 t
ol d
= if 0.65 < p< 0.8
t
new
1.5 t
ol d
= if p < 0.65

The maximum number of recycles allows you to avoid wasting machine time under such circumstances. If there
is no satisfaction of tolerances after the attempts at stepping forward, the program stops. The default of five
recycles is conservative in most cases.
4. Low stress cut-off Low stress cut-off avoids excessive iteration and small time steps caused by tolerance
checks that are based on small (round off) stress states. A simple example is a beam in pure bending. The stress
on the neutral axis is a very small roundoff-number, so that automatic time stepping scheme should not base
time step choices on tolerance satisfaction at such points. The default of five percent of the maximum stress in
the structure is satisfactory for most cases.
5. Choice of element for tolerance checking Creep tolerance checking occurs as a default for all integration
points in all elements. You might wish to check tolerances in only 1 element or in up to 14 elements of your
choice. Usually, the most highly stressed element is chosen.
When you enter the tolerances and controls, the following conventions apply:
All stress and strain measures in tolerance checks are second invariants of the deviatoric state (that is,
equivalent von Mises uniaxial values).
You can reset all tolerances and controls upon the completion of one AUTO CREEP sequence.
Background Information
Creep behavior is based on a von Mises creep potential with isotropic behavior described by the equivalent creep law:
(5-98)
The material behavior is therefore described by:
(5-99)
where is the outward normal to the current von Mises stress surface and is the equivalent creep strain rate.
There are two numerical procedures used in implementing creep behavior. The default is an explicit procedure in
which the above relationship is implemented in the program by an initial strain technique. In other words, a pseudo-
load vector due to the creep strain increment is added to the right-hand side of the stiffness equation.
(5-100)
where is the stiffness matrix, and and are incremental displacement and incremental nodal force vectors,
respectively. The integral
(5-101)
cr
f
cr
T t , , , ( ) =
cr
cr
d
--------- =
d
---------
cr
Ku P
T
D
cr
dv
V
}
+ =
K u P
T
D
cr
dv
V
}
137
CHAPTER 5
is the pseudo-load vector due to the creep strain increment in which is the strain displacement relation and is the
stress-strain relation. When plasticity is also specified through a suitably defined yield criterion and yield stress in
Marc, the plasticity is treated implicitly while the creep is treated explicitly.
As an alternative, an implicit creep procedure can be requested with the CREEP parameter. In this case, the inelastic
strain rate has an influence on the stiffness matrix. Using this technique, significantly larger steps in strain space can
be used. In Marc, this option is only to be used for isotropic materials with the creep strain rate defined by
where A is the creep constant that can be defined through input data or through the UCRPLW user subroutine; power
law expression is always to be used for the effective stress with the coefficient provided through the input data or the
UCRPLW user subroutine; and power law coefficients or more general expressions can be provided for the creep
strain, temperature, and time through the input data or the UCRPLW user subroutine, respectively.
When plasticity is also specified through a suitably defined yield stress in conjunction with the von Mises yield
criterion, a sub-iterative scheme within each Newton-Raphson cycle is used to determine the plastic strains needed to
keep the stress state on the yield surface and the creep strains that develop due to the equivalent stress being greater
than a user-defined back stress. The yield stress for the plastic component can be varied as a function of the equivalent
plastic strain, temperature and spatial coordinates. Similarly, the back stress for the creep component can be varied as
a function of the equivalent creep strain, temperature and spatial coordinates.
Creep Buckling
Marc also predicts the creep time to buckling due to stress redistribution under given load or repeated cyclic load.
The buckling option solves the following equation for the eigenvalue
(5-102)
The geometric stiffness matrix, , is a function of the increments of stress and displacement. These increments are
calculated during the last creep time step increment. To determine the creep time to buckle, perform a buckle step after
a converged creep increment. Note that the incremental time must be scaled by the calculated eigenvalue, and added
to the total (current) time to get an estimate as to when buckling occurs.
AUTO THERM CREEP (Automatic Thermally Loaded Elastic-Creep/Elastic-Plastic-Creep Stress
Analysis)
The AUTO THERM CREEP option is intended to allow automatic, thermally loaded elastic-creep/elastic-plastic-creep
stress analysis, based on a set of temperatures defined throughout the mesh as a function of time. The temperatures
and transient times are presented to the program through the CHANGE STATE option, using input option 3 (post file),
and the program creates its own set of temperature steps (increments) based on a temperature change tolerance
provided in this option. The times at all temperature steps are calculated by the program for creep analyses.
At each temperature step (increment), an elastic/elastic-plastic analysis is carried out first to establish stress levels in
the structure. A creep analysis is performed next on the structure for the time period between current and previous
temperature steps (increments). Both the elastic/elastic-plastic stress and the creep analyses are repeated until the total
creep time provided in this option is reached. Convergence controls are provided on the CONTROL option for
D
cr
A
n
f
cr
( )g T ( )h t ( ) =
K K
G
+ ( ) 0 =
K
G

elastic-plastic analysis and in the AUTO THERM CREEP option for creep analysis. The analysis can be restarted at
temperature steps (increments) or at creep steps (subincrements). The results can be saved on a post file (POST option)
for postprocessing.
If no DIST LOADS, POINT LOAD, or PROPORTIONAL INCREMENT option appears with the AUTO THERM CREEP
set, all mechanical loads and kinematic boundary conditions are held constant during the AUTO THERM CREEP.
However, DIST LOADS, POINT LOAD, PROPORTIONAL INCREMENT, or DISP CHANGE can be included in the set
the mechanical loads and kinematic boundary conditions which are then defined are assumed to change in proportion
to the time scale of the temperature history defined by the CHANGE STATE option and are applied accordingly. This
is based on the fact that the increments of load and displacement correspond to the end of the transient time of the
AUTO THERM CREEP input.
Viscoelasticity
In a certain class of problems, structural materials exhibit viscoelastic behavior. Two examples of these problems are
quenching of glass structures and time-dependent deformation of polymeric materials. The viscoelastic material
retains linearity between load and deformation; however, this linear relationship depends on time. Consequently, the
current state of deformation must be determined from the entire history of loading. Different models consisting of
elastic elements (spring) and viscous elements (dashpot) can be used to simulate the viscoelastic material behavior
described in Chapter 7. A special class of temperature dependence known as the Thermo-Rheologically Simple
behavior (TRS) is also applicable to a variety of thermal viscoelastic problems. Both the equation of state and the
hereditary integral approaches can be used for viscoelastic analysis.
To model the thermo-rheologically simple material behavior, the SHIFT FUNCTION model definition option can be
used to choose the Williams-Landel-Ferry equation or the power series expression or Narayanaswamy model.
In Marc, two options are available for small strain viscoelastic analysis. The first option uses the equation of state
approach and represents a Kelvin model. The second option is based on the hereditary integral approach and allows
the selection of a generalized Maxwell model. The thermo-rheologically simple behavior is also available in the
second option for thermal viscoelastic analysis. The Time-independent Inelastic Behavior section in Chapter 7
discusses these models in detail. Automatic time stepping schemes AUTO CREEP and AUTO STEP can be used in a
viscoelastic analysis for first and second options, respectively.
The first option for viscoelastic analysis uses the Kelvin model. To activate the generalized Kelvin model in Marc, use
the VISCO ELAS or CREEP parameter. To input the matrices [A] and [B] for the Kelvin strain rate computations, use
the CRPVIS user subroutine. To input creep time period and the tolerance control for the maximum strain in an
increment, use the AUTO CREEP history definition option.
The Simo model for large strain viscoelasticity can be used in conjunction with the damage and hyperelastic Mooney
or Ogden material model. The large strain viscoelastic material behavior can be simulated by incorporating the
VISCELMOON, VISCELOGDEN, or VISCELFOAM model definition option. Viscoelasticity for Mooney and Ogden
materials is available in both the total and updated Lagrangian framework. Viscoelasticity for foam materials is
available only in the updated Lagrangian framework.
Nonlinear structural relaxation behavior of materials can be modeled by the Narayanaswamy model which accounts
for memory effect. This model allows simulation of evolution of physical properties of glass subjected to complex
time temperature histories. The thermal expansion behavior for the Narayanaswamy model is controlled via
the VISCEL EXP model definition option.
139
CHAPTER 5
Viscoplasticity
There are two procedures in Marc for viscoplastic analysis: explicit and implicit. A brief description of each procedure
follows:
Explicit Method
The elasto-viscoplasticity model in Marc is a modified creep model to which a plastic element is added. The plastic
element is inactive when the stress is less than the yield stress of the material. You can use the elasto-viscoplasticity
model to solve time-dependent plasticity and creep as well as plasticity problems with a nonassociated flow law.
The CREEP option in Marc has been modified to enable solving problems with viscoplasticity. The method is
modified to allow solving elastic-plastic problems with nonassociated flow rules which result in nonsymmetric stress-
strain relations if the tangent modulus method is used.
The requirements for solving the viscoplastic problem are:
CREEP parameter and creep controls
Load incrementation immediately followed by a series of creep increments specified by AUTO CREEP
Use of the CRPLAW user subroutine and/or the NASSOC user subroutine
The following load incrementation procedure enables a viscoplastic problem to be solved:
1. Apply an elastic load increment that exceeds the steady-state yield stress.
2. Relieve the high yield stresses by turning on the AUTO CREEP option.
You may repeat steps 1 and 2 as many times as necessary to achieve the required load history.
The viscoplastic approach converts an iterative elastic-plastic method to one where a fraction of the initial force vector
is applied at each increment with the time step controls. The success of the method depends on the proper use of the
automatic creep time step controls. This means that it is necessary to select an initial time step that will satisfy the
tolerances placed on the allowable stress change.
The allowable stress change is specified in the creep controls. The most highly stressed element usually yields the
maximum strain rate. It is also important to select a total time that gives sufficient number of increments to work off
the effects of the initial force vector. A total time of 30 times the estimated t is usually sufficient.
Marc does not distinguish between viscoplastic and creep strains. A NASSOC user subroutine allows you to specify a
nonassociated flow rule for use with the equivalent creep strains (viscoplastic) that are calculated by the CRPLAW user
subroutine. A flag is set in the CREEP parameter in order to use the viscoplastic option with a nonassociated flow rule.
The viscoplasticity feature can be used to implement very general constitutive relations with the aid of the ZERO and
YIEL user subroutines.
Since the viscoplasticity model in Marc is a modified creep model, you should familiarize yourself with the creep
analysis procedure (see Nonlinear Analysis at the beginning of this chapter).
Note:
The size of the load increments are not altered during the AUTO CREEP process so that further load
increments can be effected by using the PROPORTIONAL INCREMENT option.
The initial time step t =
allowable stress change x 0.7
Maximum viscoplastic strain rate x Youngs modulus

Implicit Method
A general unified viscoplastic material law can be implemented through the UVSCPL. user subroutine When using this
method, you are responsible for defining the inelastic strain increment and the current stress.
Fracture Mechanics
The fracture mechanics capabilities in Marc covers the evaluation of energy release rate and J-integral including
automatic crack propagation. Two methods are offered for the evaluation: the mode separation method through the
LORENZI option and the Virtual Crack Closure Technique (VCCT) with the VCCT option. The VCCT option also
supports automatic crack propagation.
Linear Fracture Mechanics
Linear fracture mechanics presupposes existence of a crack and examines the conditions under which crack growth
occurs. In particular, it determines the length at which a crack propagates rapidly for specified load and boundary
conditions. The concept of linear fracture mechanics stems from Griffiths work on purely brittle materials. Griffith
stated that, for crack propagation, the rate of elastic energy release should at least equal the rate of energy needed for
creation of a new crack surface. This concept was extended by Irwin to include limited amounts of ductility. In Irwins
considerations, the inelastic deformations are confined to a very small zone near the tip of a crack.
The basic concept presented by Griffith and Irwin is an energy balance between the strain energy in the structure and
the work needed to create a new crack surface. This energy balance can be expressed using the energy release rate
as
(5-103)
is defined as
(5-104)
where is the strain energy and is the crack length. depends on the geometry of the structure and the current
loading. is called the fracture toughness of the material. It is a material property which is determined from
experiments. Note that the energy release rate is not a time derivative but a rate of change in potential energy with
crack length. An important feature of Equation (5-103) is that it can be used as a fracture criterion; a crack starts to
grow when reaches the critical value .
The stress and strain fields near the tip of a crack are singular for a linear elastic material model. The stresses and
strains have the principal form
(5-105)
G
G G
c
=
G
G
d
da
-------- =
a G
G
c
G G
c
K
r
------ f ( ) =
K
r
------ g ( ) =
141
CHAPTER 5
in a polar coordinate system centered at the crack tip. Thus, a linear elastic material is said to have a singularity
near a crack tip. It is easy to demonstrate that in both Griffiths and Irwins considerations, the elastic energy release
rate is determined by a single parameter: the strength of the singularity in the elastic stress field at the crack tip. This
is the so-called stress intensity factor, and is usually denoted by capital . The magnitude of depends on the crack
length, the distribution and intensity of applied loads, and the geometry of the structure. Crack propagation occurs
when any combination of these factors causes a stress intensity factor to be equal to or greater than the
experimentally determined material property , which is equivalent to Equation (5-103). Hence, the objective of
linear fracture mechanics calculations is to determine the value of .
There are three possible modes of crack extension in linear elastic fracture mechanics: the opening mode, sliding
mode, and tearing mode (see Figure 5-18).
Figure 5-18 Irwins Three Basic Modes of Crack Extension
The opening mode (see Figure 5-18a), Mode I, is characterized by the symmetric separation of the crack surfaces with
respect to the plane, prior to extension (symmetric with respect to the X-Y and X-Z planes).
The sliding mode, Mode II, is characterized by displacements in which the crack surfaces slide over one another
perpendicular to the leading edge of the crack (symmetric with respect to the X-Y plane and skew-symmetric with
respect to the X-Z plane).
1 r
K K
K
K
c
K
(a) Mode I: Opening
x
z
y
(b) Mode II: Sliding
x
z
y
(c) Mode III: Tearing
x
z
y

The tearing mode, Mode III, finds the crack surfaces sliding with respect to one another parallel to the leading edge
(skew-symmetric with respect to the X-Y and X-Z planes).
It is customary to associate a stress intensity factor with each of these mode: , , and . There is also an
associated fracture toughness associated with each mode: , , and . The most critical mode is usually
mode I and in many cases the other modes are not considered. The connection between the energy release rate and the
stress intensity factors is given by
(5-106)
where
for plane stress (5-107)
and
for plane strain. (5-108)
Marc uses the so-called J-integral for evaluating the energy release rate, see below. The J-integral is similar to but
is more general and is also used for nonlinear applications. J is equivalent to when a linear elastic material model
is used.
Nonlinear Fracture Mechanics
Nonlinear fracture mechanics is concerned with determining under which conditions crack propagation (growth)
occurs. In this sense, nonlinear fracture mechanics is similar to linear fracture mechanics. However, there are
additional questions addressed in nonlinear fracture mechanics. Is the crack propagation stable or unstable? If it is
stable, at which speed does it occur? After some propagation, is the crack arrested?
There also exists a singularity at the crack tip in fracture mechanics problems with nonlinear elastic-plastic material
behavior, though the singularity is of a different nature. If one takes an exponential hardening law of the form
(5-109)
it can be shown that the singularities in the strain and the stresses at the crack tip are of the form
(5-110)
If n approaches infinity, the material behavior becomes perfectly plastic and the singularity in the stresses vanishes.
The singularity in the strains, however, takes the form of
(5-111)
K
I
K
I I
K
I I I
K
I c
K
I I c
K
I I I c
G
K
I
2
E'
-------
K
I I
2
E'
--------
1 +
E
------------- K
I I I
2
+ + =
E' E =
E'
E
1
2
--------------- =
G
G
0
------

0
-----
\ .
| |
1 n /
=
f ( )r
n n 1 + ( )
=
g ( )r
1 n 1 + ( )
=
f ( )r
1
=
143
CHAPTER 5
It has not been possible to establish that the strength of the singularity is the only factor that influences initiation of
crack propagation for nonlinear situations. In fact, it is doubted that initiation of crack propagation is dependent on
only a single factor. The J-integral probably offers the best chance to have a single parameter to relate to the initiation
of crack propagation.
The J-integral was introduced by Rice as a path-independent contour integral for the analysis of cracks. As previously
mentioned, it is equivalent to the energy release rate for a linear elastic material model. It is defined in two dimensions
as
(5-112)
where is the strain energy density, is the kinetic energy density, is the stress tensor and is the displacement
vector. The direction is the same as the x direction in the local crack tip system in Figure 5-19. The integration path
is a curve surrounding the crack tip, see Figure 5-19.The J-integral is independent of the path as long as it starts
and ends at the two sides of the crack face and no other singularities are present within the path. This is an important
feature for the numerical evaluation since the integral can be evaluated using results away from the crack tip.
Figure 5-19 Definition of the J-integral
Numerical Evaluation of the J-integral
The J-integral evaluation in Marc is based upon the domain integration method as described in [Ref. 29]. It is available
as the LORENZI model definition option. A direct evaluation of Equation (5-112) is not very practical in a finite
element analysis due to the difficulties in defining the integration path . In the domain integration method for two
dimensions, the line integral is converted into an area integration over the area inside the path . This conversion is
exact for the linear elastic case and also for the nonlinear case if the loading is proportional, that is, if no unloading
occurs. By choosing this area as a set of elements, the integration is straightforward using the finite element solution.
In two dimensions, the converted expression is
(5-113)
J W T + ( )n
1

i j
n
i
u
j
x
1
--------
\ .
| |
d
}
=
W T
i j
u
i
x
1

n
i
x
y
J
i j
u
j
x
1
-------- W
1i
\ .
| |
q
1
x
i
-------- A d
A
}
=

for the simplified case of no thermal strains, body forces or pressure on the crack faces. For the general expression,
see [Ref. 1]. A is the area inside and is a function introduced in the conversion into an area integral. The function
can be chosen fairly generally, as long it is equal to one at the crack tip and zero on . The form of the function
chosen in Marc is that it has the constant value of one at all nodes inside , and decreases to zero over the outermost
ring of elements in . It can be interpreted as a rigid translation of the nodes inside while the nodes on remain
fixed. Thus, the contribution to Equation (5-113) comes only from the elements in a ring away from the crack tip. This
interpretation is that of virtual crack extension and this method can be seen as a variant of such a technique, although
it is extended with the effects of thermal strains, body forces, and pressure on the crack faces.The set of nodes moved
rigidly is referred to as the rigid region and the function as the shift function or shift vector. For the evaluation of
the J-integral the direction of the shift vector is simply the x axis in the local crack tip system.
In three dimensions, the line integral becomes an area integral where the area is surrounding a part of the crack front.
In this case, the selection of the area is even more cumbersome than in two dimension. The converted integral becomes
a volume integral which is evaluated over a set of elements. The rigid region is a set of nodes which contains a part of
the crack front, and the contribution to the integral comes from the elements which have at least one but not all its
nodes in the rigid region.
Determination of the Rigid Region and the Shift Vector
For the evaluation of the J-integral, Marc requires the nodes along the crack front, the shift vector, and the nodes of
the rigid region. Marc allows three ways of defining the rigid region. The first is direct input; the nodes or elements in
the rigid region are listed explicitly. With this variant, the shift vector is also specified directly. The second and third
variants use an automatic search for the nodes of the rigid region. The second variant is based on the mesh topology
(connectivity) where a number of regions of increasing size are found by Marc. The first region for two dimensions
consists of the nodes of all elements connected to the crack tip node. The second region consists of all nodes in the
first region and the nodes of all elements connected to any node in the first region and so on for a given number of
regions. This way, contours of increasing size are determined. In the third way of determining the rigid region, a radius
is given and all nodes within that radius are part of the rigid region. For the automatic search methods the shift vector
can also be determined automatically. It is then determined using the first element edge on the crack face (for meshes
with notches see below). Symmetry at the crack face is automatically detected by Marc, also for the case that the
symmetry condition is applied by means of a rigid contact body.
In three dimensions, it is a bit more complicated. The crack front is defined by an unsorted list of nodes. In order to
obtain the variation of the J-integral along the crack front, a disk of nodes with the normal of the disk directed along
the tangent to the crack front can be determined. This type of disk can readily be defined if the mesh around the crack
front consists of brick elements in a regular mesh. This mesh is typically created from a two-dimensional mesh which
is extruded along the crack front. The topology based determination of the rigid region assumes such a mesh and
creates disks of increasing size at each crack front node from element faces. The shift vector at each crack front node
is automatically determined to be perpendicular to both the tangent to the crack front and the normal to the crack face
at each crack front node. At the first and last crack front node, where a free surface is assumed to exist, the shift
direction is projected to the tangent to the free surface. This is important since the shift must not change the outer
boundary of the model. The geometry based search method here works with a cylinder with the axis aligned with the
tangent of the crack front and centered at the crack front node. The length of the cylinder is given as a fraction of the
q
1
q
1

A
q
1
145
CHAPTER 5
distance to the neighboring crack front nodes. All nodes within the cylinder are part of the rigid region. This method
is useful if, for instance, the mesh around the crack front is created with an automatic mesh generator. The shift vector
is determined in the same way as for the topology based search method.
In elasto-plastic analyses, it is often advantageous to use a mesh where there are several (multiple) nodes at the crack
tip. If collapsed elements are used at the crack tip, the nodes are kept separate and a notch is formed as the crack tip
deforms. For this kind of mesh, a multiple nodes distance is given, which should be smaller than the smallest element
at the crack tip. All nodes within that distance are then considered part of the crack tip, and the first contour for the
topology based search will consist of all elements connected to any of these nodes (in three-dimensions of element
faces connected to any of these nodes). The normal to the crack face used for determining the shift vector is taken as
the first face outside the crack tip nodes. Thus, it is also possible to model the crack with an initial notch.
Mode Separation
For the linear elastic case, Marc automatically calculates and prints the stress intensity factors , , and for
the three modes I, II, and III. This evaluation is done using analytical functions for the stress field near the crack tip.
The implementation is based upon the procedure outlined in [Ref. 32]. Mode separation is not calculated when
nonlinearities like large deformations, nonlinear material, or general contact are present. However, the use of glued
contact is allowed to connect different parts of the mesh near the crack front. The material is also not allowed to be
temperature dependent. Orthotropic materials are allowed. The calculation of the stress intensity factors is
automatically suppressed if unsupported features are used.
The stress intensity factors are defined in the local crack tip system, see Figure 5-19. will be positive when the
crack is opening and negative when it is overlapping. Similarly, is positive when the shear stress is positive ahead
of the crack in the local system.
Supported Features
The LORENZI J-integral option supports the following features:
Linear and nonlinear materials.
Large deformations. Both total and updated Lagrangian formulation are supported. The crack tip system can
undergo finite rotations and translations.
Thermal loads.
Transient dynamics.
Contact between crack faces including friction.
Loads on the crack faces.
Mode separation for linear elasticity and no loads in the crack region.
Modeling Considerations
The main difficulty in the finite element analysis of linear elastic fracture mechanics is representing the solution near
the crack front. Typically, a focused mesh is used with the elements in a spider web shape. This gives a mesh with rings
of elements in a radial fashion and is particularly useful for studying the variation of the values of with contour
radius. If higher order elements are used for an elastic analysis, it is advantageous to use so-called quarter point
K
I
K
I I
K
I I I
K
I
K
I I
J

elements at the crack tip. These are regular elements where the midside nodes at element sides for which one node is
part of the crack tip are shifted towards the crack tip to the quarter point location along the side. This gives a more
accurate description of the singularity at the crack tip and is important to use if a coarse mesh is used.
In three-dimensions, the mesh along the crack front is typically created from a planar mesh which is extruded along
the crack front region as mentioned above. This allows the rigid regions to be created in a regular manner which is
advantageous for accuracy. Note that Marc allows rigid contact surfaces to be used for applying symmetry conditions.
For a three-dimensional example, see the Marc Users Guide, Chapter 6.2.
Numerical Evaluation of the Energy Release Rate with the VCCT
Method
The VCCT option offers a simpler but more general way for obtaining the energy release rate. The implementation
follows the description in R. Krueger [Ref. 33]. Consider again Equation (5-104).
(5-114)
It states that G is the change in potential energy by a change in crack length. Now, consider the simple finite element
model in Figure 5-20. The models A and B are the same, except that in B, the crack has grown by one element edge
of length a. Suppose we do one analysis for each one of the two models. We can now calculate the energy release rate
G as:
(5-115)
Figure 5-20 Mesh for Illustrating the CCT Method
Here, is the force (obtained from Model A) that keeps the crack together, and the crack opening, , is obtained from
Model B. In order to obtain these quantities, we would need to perform two analyses and this method is often referred
to as CCT (Crack Closure Technique). In the virtual crack closure technique, we only do the analysis with a closed
G
d
da
-------- =
G
Fu
2a
------ =
F
a
u
Model A Model B
F u
147
CHAPTER 5
crack (Model A) and use the opening displacement at the closest nodes to the crack tip. Figure 5-21 shows the case of
pure mode I. The other modes are treated similarly and separately. The displacements and reactions are transformed
into the local crack tip system for this evaluation.
Figure 5-21 The VCCT Method
With x. y, and z denoting the coordinate directions in the local crack tip system (see Figure 5-18), we obtain:
(5-116)
and the total energy release rate as:
(5-117)
For higher-order elements, we need to include the contributions from the midside nodes (see Figure 5-22).
(5-118)
Figure 5-22 Higher-order Elements
F
a
u
G
I
F
y
u
y
2a
------------ = G
I I
F
x
u
x
2a
------------ = G
I I I
F
z
u
z
2a
----------- =
G
t ot
G
I
G
I I
G
I I I
+ + =
G
I
F
1
u
1
F
2
u
2
+
2a
------------------------------- =
a
u
1
F
1
F
2
u
2

For the case that the midside nodes are not at the middle of the element edges (for example, using the 1/4 point position
for increased accuracy, the displacements for these nodes are interpolated to the appropriate locations.
For 3-D solids, we have a situation as shown in Figure 5-23. The situation is similar to the 2-D case, and the evaluation
is done separately for each node along the crack front. The area is given by the shaded part in Figure 5-23.
For the case of higher-order elements, we obtain the following by using the notation in Figure 5-24.
(5-119)
Figure 5-23 3-D Mesh for VCCT
G
F
1
u
1
F
2
u
2
1
2
--- F
3
u
3
F
4
u
4
+ ( ) + +
2a
------------------------------------------------------------------------------- =
Crack Front
F
u is Crack Opening
Displacement At Crack Face
a
149
CHAPTER 5
Figure 5-24 3-D Mesh for VCCT for Higher-order Elements
The above figures show a regular mesh of hexahedral elements. While a mesh designed like this is advantageous for
accuracy, it is not strictly necessary. It is also possible to use a general tetrahedral mesh or a hexahedral mesh with the
crack front as defined in Figure 5-25. The latter is typically obtained in the case of crack propagation as described
below. The program will find the appropriate nodes to use for the forces and crack opening displacement and calculate
an area a. In the current release, the case of higher order tetrahedral elements is not supported.
Figure 5-25 Example of Crack Front Not Following Element Edges
F
1
F
2
F
4
F
3
u
4
u
1 u
3
u
2

The definition of the data involved in the VCCT calculation is done automatically. The user only specifies the crack
tip or crack front. In order to check what the program finds there is debug option available. If the PRINT parameter is
used with a value of 43, the program will print out the nodes it finds, the corresponding forces, crack opening
displacements, and crack areas.
Symmetry is automatically detected and accounted for. The symmetry condition can be enforced by boundary
conditions or by rigid contact.
The mesh in the crack region can contain user tyings or contact. The program automatically detects if the crack tip
node is connected to a node, and the node it is connected to is considered part of the crack tip. The supported
connections are glued contact, user typing, RBE2 and RROD. For the case that the connection is done node-to-node,
it does not matter if the tied or retained node of the tying is selected as the crack node. With glued contact, it is not
necessary to connect the parts node-to-node. If the meshes do not match up, the crack will be treated similarly as a
case of symmetry. This non-matching glued case is automatically detected and for this case, it is necessary to select
the touching node as the crack tip node. A useful option when defining a crack using glued contact is DEACT GLUE.
Suppose two parts are glued together and the crack is defined as part of the interface between the bodies. One can then
use DEACT GLUE to identify the nodes that should be on the crack faces (see Figure 5-26). These nodes will have
regular contact but not be glued.
Figure 5-26 Using DEACT GLUE for Defining a Crack
The calculation of VCCT is done in the current geometry in case of large deformations. The updated crack coordinate
system is calculated at each increment and the calculations are done in this system. Thus, arbitrary rotations and
deformations are allowed. The current crack tip system is available on the post file. The three coordinate axes can be
plotted as vectors for verification.
Figure 5-27 shows some examples of supported crack configurations. There is a large flexibility in how parts can be
tied or glued together to form a crack.
Contact Body 1
Contact Body 2
Crack Tip
Nodes Identified
with DEACT GLUE
Glued Contact
151
CHAPTER 5
Figure 5-27 Examples of Supported Crack Types
Remeshing is supported for 2-D solids and shells. A structure containing a crack can be remeshed and the new crack
tip is automatically found after remeshing. The crack opening displacement is saved before the remesh, and applied
afterwards to make sure that the results are also correct after remeshing takes place. Three-dimensional solids with
cracks cannot currently be remeshed. Remeshing can also be used for automatic crack propagation as described below.
Automatic Crack Propagation
The VCCT option allows automatic crack propagation to be performed. There are two types of crack propagation
analysis types available: fatigue growth and direct growth. There are four methods supported for how the actual crack
growth is performed: growth during remeshing, growth by releasing tyings or glued contact, growth by cutting through
the mesh and growth along element edges. These options are described in the following.
Fatigue Crack Growth
Here, the user specifies a fatigue time period. The analysis is performed without crack growth during this period,
which can be repeated any number of times. During the time period, it records the maximum and minimum energy
release rate and the estimated crack growth direction at the maximum. At the end of each fatigue time period, the crack
growth is applied. For the growth, we need to know the distance and direction the crack should grow. The distance is
either user specified (possibly scaled by a table) or obtained by Paris law. With Paris law we use the following
expression:
(5-120)
Line Crack in Shell Shell Glued to Shell 3-D Solid
Shell Glued to Solid (Line Crack) Shell Glued to Solid (Surface Crack)
Crack Front
Crack Tip
a C G
max
G
mi n
( )
m
G
t h
m
[ ] =

This formula is from T. L. Anderson [Ref. 30], modified to use G instead of the stress intensity K. No growth occurs
if or . The ucrack_paris.f user subroutine can be used for specifying the growth
increment. The crack growth increment is first calculated with Equation (5-120) and then the user subroutine is called.
See UCRACKGROW in Volume D: User Subroutine, Special Routines, and Utility Routines for details.
If growth by remeshing is used, the specified or calculated crack growth increment and estimated crack growth
direction are used for defining the location of the new crack tip. When release of glued contact or tyings are used, then
as many element edges are released as needed for growing the crack according to the specified or calculated growth
increment. If there are multiple choices available for which element edge to release, then the one closest to the crack
growth direction is chosen.
Direct Crack Growth
For this case, a standard transient analysis is performed. The user specifies a crack growth resistance (fracture
toughness) for each crack. When the crack growth criterion is fulfilled, a propagation is done. This takes place within
an increment; the increment iterates until no further crack growth occurs. For unstable crack growth, this may have
the effect that the crack grows through the whole structure within one increment. This ensures that an increment does
not show convergence when a crack should grow which would violate the crack growth criterion.
With remeshing based growth, the user specifies the crack growth increment to use and the current estimated crack
growth direction is used. Repeated remeshing may be used in an increment if more growth is triggered. For the growth
by releasing tyings or glued contact, one element edge at the time is released until no more growth occurs. For both
methods, the growth only occurs when the increment is otherwise converged.
Remeshing Based Growth
Crack growth by means of remeshing is available for 2-D solid elements. The model must be set up to do global
remeshing, but no specific remeshing criterion needs to be set for the crack propagation. Any specified remeshing
criterion will be treated independently from the remeshings due to crack growth. It is, for example, allowed to do
remeshings during the fatigue load cycle.
When a remeshing occurs in a 2-D analysis, the program first finds the outline of the remeshing body. When a crack
is growing, this outline is extended to form a new crack tip location. In order to assure that we obtain a mesh with an
element edge in the x direction of the local crack tip system, we extend the outline a little further than to the new crack
tip. The new mesh is modified such that the new mesh is correct. This way we avoid meshes as shown in Figure 5-28a.
This is also done for shells, so we always get a good mesh at the crack tip after remeshing.
If a crack reaches the boundary of the model, the remeshing body automatically splits. Thus, a crack can grow into
another crack, an internal hole of the body, or the external boundary. Due to the outline extension mentioned above,
the crack will reach the boundary slightly earlier than one would otherwise expect.
The setting of target element length after remeshing will be overridden for the case that it is larger than the growth
increment. This ensures that the mesh resolution is large enough to model the new crack tip. The mesh around the crack
tip is automatically refined when growth occurs. An example of a remeshed crack is shown in Figure 5-28b. This mesh
refinement is not done for shells at this point.
G
max
G
mi n
G
t h
< a a
mi n
<
153
CHAPTER 5
Figure 5-28 Meshes after Remeshing.
Growth by Releasing Tyings or Glued Contact
For this option, the user needs to specify the crack growth path. The crack can only grow in the direction where there
are tyings or glued contact to release. All crack configurations are supported: 2-D, shells and 3-D.
For the case of user tyings, there are three types of tyings supported: user tying type 100, RBE2 and RROD. Here, the
nodes need to be aligned on both sides of the crack faces. The user specifies only the crack tip node; the program
automatically finds the tyings to release. It does not matter if the tied or retained node is selected as the crack tip node
or crack front nodes.
For glued contact, the meshes do not need to match up. As mentioned in the previous section, a special non-matching
glue symmetry case is automatically used if the nodes do not match up. For this case, the tied nodes must be selected
as the crack tip nodes. When nodes in glued contact are released due to crack growth, they switch from glued contact
to regular contact. The DEACT GLUE option is here used internally. Thus, a node which is released can still contact
the new crack face, but it will not glue back again.
If higher-order elements are used, the midside node will be released when the corresponding corner node is released.
If a crack reaches a model boundary, the crack evaluation for this crack will be turned off and the analysis can continue.
The program avoids that a single node is tied at a boundary; if the last element edge is released, then the last node will
also be released.
If the mesh is designed such that there are multiple choices for which element edge to release, then the one closest to
the estimated crack growth direction is chosen.
The crack front for the 3-D case allows for irregular meshes. The front always follows element edges, but in a mesh
with hexahedral elements, a concave corner node is marked as inactive so the actively used front nodes cross the
diagonal of this element face. This inactive front node becomes active if the neighboring front nodes are released due
a) poor mesh which is avoided during
crack growth
b) refined mesh at current crack tip


to growth. Gaps in the crack front are allowed. If part of the front reaches a boundary (for example, a hole in the
structure), then the crack front will contain a gap and this gap may disappear if the front closes again (for example, if
the whole front has passed a hole in the structure).
Growth by Mesh Cutting (beta feature in 2011)
For this option, the crack can grow arbitrarily through a 2-D or shell mesh without global remeshing. Nodes are shifted
and elements are split up in order to create the cut. Only the mesh in the vicinity of the crack is affected when the crack
grows. Lower- and higher-order elements are supported with this option. The data from the previous mesh is mapped
to the new or modified elements similar to the remeshing case, but it is only done in the affected part of the mesh.
This option is similar to the XFEM approach, but new boundaries are explicitly generated. If contact is used, then the
contact boundary is automatically updated so if self contact is used in the part where growth occurs, then the new crack
faces will use contact and friction (if defined).
The mesh cutting procedure is explained below in the Mesh Splitting Along Edges or Faces section. It is done in two
steps. The first step creates a cut up to the new tip node. In the second step a scratch is created ahead of the new tip
in order to ensure an accurate VCCT calculation. This is similar to what is done with remeshing with the extended
outline.
This feature is not officially supported in the 2011 release but is available as a beta feature. The accuracy of the energy
release rate calculation is not always adequate due to the often irregular mesh at the crack tip. This may lead to
inaccurate energy release rates and inaccurate crack growth directions with the maximum hoop stress criterion.
Growth Along Element Edges
For this option, the crack can grow along element edges for 2-D and shell elements. The element edge closest to the
crack growth direction is used. New nodes are automatically inserted and the element connectivity is changed for
elements around the crack tip in order to grow the crack. The new nodes inherit the properties of the respective original
node.
Crack Growth Criteria
There are currently five different crack growth criteria available:
Total G.: The crack will grow if , where is the crack growth resistance (also called fracture toughness).
Total G calculated with LORENZI: For this case, the user either defines a crack using LORENZI with the same
crack tip (or front for 3-D) as used for VCCT or this will be created internally with default settings (topology
search, one rigid region). The value of G used for the growth criterion is taken from the LORENZI results and the
growth direction is determined based upon the VCCT values.
Separately for each mode: Crack growth will take place if either of the following is true: ,
or .
Mixed mode criterion 1:
G G
c
> G
c
G
I
G
Ic
> G
I I
G
I I c
>
G
I I I
G
I I I c
>
G
I
G
I c
--------
\ .
| |
n
1
G
I I
G
I I c
----------
\ .
| |
n
2
G
I II
G
II I c
------------ -
\ .
| |
n
3
+ + 1 >
155
CHAPTER 5
Mixed mode criterion 2:
Calculation of Crack Growth Direction
The program calculates an estimated crack growth direction for each crack front node. Four methods for defining this
are supported: the maximum principal stress criterion ([Ref. 34]), along the pure mode with largest (where
i is mode I, II or III), along mode I and along a specified vector. The maximum principal stress criterion states that a
crack will grow in the direction normal to the direction of greatest tension. The crack growth direction normal to the
direction of greatest tension. The crack growth direction in the local crack tip system can be expressed in
terms of the stress intensity factors in mode I and II as:
(5-121)
For the 3-D case, mode III is assumed to have the same effect as mode I so is used instead of in the
above equation. It is, thus, assumed that crack growth occurs perpendicular to the crack front tangent. The user can
also specify the crack growth direction through the UCRACKGROW user subroutine.
Crack Growth into Intersecting Shells
If a shell structure has shell intersections, then a crack can bifurcate at the intersection. The intersecting part can be
connected to the base part (where the crack is) by shared nodes or glued contact. For the glued contact case, the nodes
of the two parts do not need to match up. The crack bifurcation is done in two steps as shown in Figure 5-29. The first
step is when the growing crack would cross the intersecting shell. At this point, the crack growth increment is
decreased so that the updated crack tip ends up exactly at the intersection. A scratch in the mesh is generated in the
intersecting shell in order to insure an accurate energy release rate calculation. The second step is when this crack is
determined to grow again. Then the crack in the base part grows as usual, and a new crack is introduced in the
intersection. The length of the new crack is the same is the growth increment of the original crack and it grows
perpendicular to the base part. The new crack inherits the properties of the original crack but grows independently.
Any number of new cracks can be generated this way.
G
I
G
I I
G
II I
+ +
G
I c
G
I Ic
G
I c
( )
G
I I
G
I I I
+
G
I
G
I I
G
I I I
+ +
---------------------------------------
\ .
| |
n
1
G
I I I c
G
I I c
( )
G
I I I
G
I I
G
I I I
+
-------------------------
G
I I
G
I II
+
G
I
G
I I
G
II I
+ +
---------------------------------------
\ .
|
| |
n
1
+ +
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 1 >
G
i
G
i c
d
x
d
y
, ( )
d
x
3
K
I I
2
K
I
2
-------- 1 8
K
II
2
K
I
2
-------- + +
1 9
K
I I
2
K
I
2
-------- +
---------------------------------------------- = d
y
K
I I
K
I
-------- 3
K
I I
K
I
-------- 1 8
K
I I
2
K
I
2
-------- +
1 9
K
I I
2
K
I
2
-------- +
------------------------------------------------------- =
K
I
K
I I I
+ K
I

Figure 5-29 Creation of New Crack at Shell Intersection
Modeling Considerations for Crack Propagation
In the previous sections, we have described the different methods to grow a crack. The question is which method to
use in which situations.
The option to grow along a tied or glued interface is designed for situations where the crack path is already known.
This includes cases like delamination growth between layers of a composite material, growth along a material or
structural interface, or simply for cases where the crack is known or assumed to grow straight ahead. The option to
grow along element edges also falls into this category, even though it has the possibility to change growth direction.
For general crack growth directions, one would need a fine mesh to get adequate results.
For cases where the crack path is not known, we recommend the options to grow by remeshing or by mesh cutting.
For these options, the crack growth is not limited by the mesh. The calculated growth direction determines the crack
path. With remeshing based growth, the whole part containing the crack is remeshed. This option is also restricted to
what remeshing supports in general. With the mesh cutting option, only the region around the crack tip is modified. It
also supports higher order elements.
The choice between fatigue and direct propagation is determined by the type of analysis being performed. The fatigue
style propagation is primarily designed for low cycle fatigue, where each loading cycle is analyzed. For applications
with high cycle fatigue, it is usually not realistic to model each loading cycle. Here, one would typically use the
concept of cycle jump, where a few representative cycles are calculated and each calculated cycle represents a larger
number of real cycles.
It is possible to combine fatigue and direct crack propagation for the same crack. Suppose a crack grows in fatigue
mode and at some point the energy release rate becomes so high that a direct propagation would occur. To model this
one would use two VCCT definitions for the same physical crack. One is defined for fatigue and one for direct growth.
Both VCCT definitions will be updated with the new crack tip or front when one is growing.
Dynamic Fracture Methodology
In complete similarity with static fracture mechanics concepts, it is assumed that dynamic crack growth processes for
linear materials are governed by the following condition:
(5-122)
Mesh Before Reaching Intersection Growth Adjusted to Stop at Intersection;
Scratch Added in Intersecting Part
New Crack Added in
Intersecting Part
K
I
t ( ) R
I D
a
T B , , ( ) =
a
0
157
CHAPTER 5
where is the dynamic stress intensity factor for mode I, is the crack velocity, and is the dynamic crack
propagation toughness, which is assumed to be a material parameter that in general depends on crack velocity ,
temperature , and specimen thickness .
The dynamic stress intensity factor depends on crack length ( ), applied loading ( ), time ( ), specimen dimensions
( ), temperature ( ), and initial stress fields ( ) caused by residual stresses or by an initial strain field. The
prediction of the crack propagation history and crack arrest event demands complete knowledge of the vs.
relation. The Dynamic Fracture Methodology procedure consists of the following two phases:
1. Generation phase in this phase, a crack arrest experiment is performed yielding a crack propagation-versus-
time curve. In addition, a numerical simulation of the experiment is carried out by using the measured crack
propagation curve. This is used as input for the numerical model. This allows the calculation of dynamic stress
intensity factors as a function of time. Combination of the latter relation with the measured crack propagation
curve results in a curve, which can be considered the dynamic crack propagation toughness-versus-crack
velocity relation.
2. Application phase in order to predict the crack growth and possible crack arrest point in a structural
component, the inverse problem is solved. Now, the actual stress intensity factors are calculated for the
structural component, that is subjected to a particular loading history, by means of a dynamic finite element
analysis. These calculated values are compared to the fracture toughness curve obtained during the generation
phase, Equation (5-122), and from this the crack growth is predicted.
Dynamic Crack Propagation
The concepts of fracture mechanics discussed in previous sections have been applied to the prediction of crack
initiation, as well as slow stable crack growth of statically loaded structures and for the prediction of fatigue crack
growth in cyclically loaded structures. In problems where inertial effects cannot be ignored, application of quasi-static
fracture mechanics techniques can lead to erroneous conclusions. The use of dynamic fracture mechanics concepts for
these problems is clearly of necessity. The main emphasis on dynamic fracture mechanics is to predict the initiation
of stationary cracks in structures, which are subjected to impact loading. It also focuses on the conditions for the
continuous growth of fast propagating cracks, and on the conditions under which a crack is arrested.
The problem of predicting the growth rate and the possible crack arrest point is quite complicated. It is often treated
by means of a so-called dynamic fracture methodology, which requires the combined use of experimental
measurements and of detailed finite element analyses. An essential step in this approach is formed by the numerical
simulation of propagating cracks by means of the finite element method.
The J-integral as implemented in Marc takes into account the effect of inertial and body forces, thermal and mechanical
loading and initial strains.
The use of the DYNAMIC parameter and the LORENZI model definition option allows for the calculation of dynamic
energy release rates for cracked bodies which are subjected to arbitrary thermal and mechanical loadings including
initial stresses.
K
I
t ( ) a
R
I D
a
T B
a t
D T
i
R
I D
a

Mesh Splitting
Marc has the capability to split up the mesh automatically during the analysis. The splitting can be done in two ways:
split at element edges or faces and by cutting through the mesh.
The first option can be done at nodes and along element edges (2-D and shells) or element faces (3-D). A new node is
created where the split is done. This node inherits the current properties of the original node. The elements at one side
of the split are modified to use the new node to create the opening in the mesh.
The second option cuts through the mesh along a straight line. It is currently supported for 2-D solids and 3-D shells.
Mesh Splitting Along Edges or Faces
The way the mesh splitting works is illustrated in the following. Suppose we have the 2-D mesh shown in
Figure 5-30a. The edges shown in red are targeted for mesh split and this results in the mesh shown in Figure 5-30b.
Figure 5-31 illustrates some different cases and how the mesh is then split. In general, we need to specify edges for
the split so we know how to redefine the element connectivity in order to split the mesh. The examples are shown in
2-D for simplicity of illustration. The mesh split also applies to shells and 3-D solids. For 3-D solids, it is based upon
faces instead of edges.
Figure 5-30 Mesh Split at Red Edges
The mesh splitting technique is used in a number of features in Marc. The VCCT crack propagation along element
edges (see Direct Crack Growth) makes use of this feature.
The mesh can also be split by means of the USPLIT_MESH user subroutine (see Marc Volume D: User Subroutines
and Special Routines for details).
The DELAMIN model definition option allows the specification of mesh splitting between regions and within a region
using a stress criterion. The mesh is split if the following criterion is fulfilled:
if (5-123)
otherwise . (5-124)
a b
n
S
n
------
\ .
| |
m

t
S
t
-----
\ .
| |
n
+ 1 >
n
0
t
S
t
--------
\ .
| |
n
1 >
159
CHAPTER 5
Figure 5-31 Examples of Mesh Split
Here is the normal stress, is the tangential stress and , , m and n are user-defined parameters. Hence, no
splitting takes place due to compressive stresses. If or is zero, the respective term is set to zero. It is also possible
to replace the default stress criterion with a user defined criterion via the UDELAM user subroutine (see Marc Volume
D: User Subroutines and Special Routines for details).
The region where the split is considered can be a material, a composite definition or a PSHELL definition. With the
option to split between regions, the program checks all element edges (faces for 3-D solids) in the interface between
the specified regions. The stresses are extrapolated to the nodes and transformed to a coordinate system aligned with
the region interface. See Figure 5-32. The normal direction is perpendicular to the region interface. With the option to
split within a region, it checks all edges/faces inside a region. The normal direction is perpendicular to each element
edge or face.
No split possible
n

t
S
n
S
t
S
n
S
t

Figure 5-32 Coordinate System for Delamination between Regions
For layered elements like shells and solid composites, the delamination criterion is calculated separately for each layer.
A mesh split is triggered if all layers fulfill the delamination criterion. The same criterion is applied to each individual
layer. An alternative approach to model material failure in each layer is the Progressive Failure Analysis approach
(PFA) available through the FAIL DATA option (also see Progressive Composite Failure in Chapter 7 of this manual).
If contact is used, the program recalculates the contact boundary after mesh splitting takes place. This is important in
order to avoid penetration between the newly generated free surfaces.
Nodal post codes are available for post processing the separate terms in Equation (5-123) as well as the sum. The first
has a nodal post quantity labeled Delamination Index (Normal), the second is called Delamination Index (Tangential)
and the sum Delamination Index.
An option is available in the DELAMINATION option for inserting interface elements using a cohesive zone material
model where the mesh is split. This is automatically done whenever a split takes place. The new elements use the nodes
of the surrounding elements and the appropriate type of the interface element is automatically selected. For example,
if a split occurs between two 10-noded tetrahedral elements, then a 15-noded pentahedral element is inserted. If the
split-up zone grows, more interface elements are added. Collapsed interface elements are used where needed, and
these elements change to regular elements if the split zone grows further. There is no corresponding interface element
for shells connected edge-to-edge, so for this case no interface elements are added.The cohesive material properties
of the interface elements must be defined in the model. The ELEMENTS parameter defining the element type for the
interface elements must be included.
The option is supported for lower- and higher-order 2-D solid, shell and 3-D solid elements. For the case of higher-
order elements, only the corner nodes are considered, and when corner nodes are split, so are the involved midside
nodes. The midside nodes are assigned the average delamination index value of the corner nodes for postprocessing.
region 1
region 2
t
161
CHAPTER 5
Mesh Cutting
With this option, the mesh is modified by making a cut along a line. Nodes are shifted and elements subdivided in
order to create the cut. The cut is done along a straight line and the starting and ending points can be either inside or
outside the mesh. The shifting of nodes is done to avoid thin elements as far as possible. Results are transferred from
the original mesh to the modified mesh using the rezoning capabilities in Marc. This data transfer is only done in the
affected region, as opposed to remeshing where this is done in the whole remeshing body.
The elements in the region of the cut must be of the same type. It is, for example, not possible to combine triangles
and quadrilaterals. This is due to a limitation in the rezoning. If triangles are needed together with quadrilateral
elements, the triangles should be defined as collapsed quadrilaterals.
The updated Lagrange procedure must be used in the element group where mesh cutting takes place.
Mesh cutting is currently supported for 2-D continuum elements that support the updated Lagrange procedure and
shell elements except solid shell elements. Both lower- and higher-order elements are supported.
The mesh splitting technique is used in a number of features in Marc. The VCCT crack propagation by mesh cutting
makes use of this feature (see Growth by Mesh Cutting (beta feature in 2011)). The mesh can also be split by means
of the UCUT_MESH user subroutine (see Marc Volume D: User Subroutines and Special Routines for details). Some
examples of mesh cutting can be found in the description of the UT_SPLIT2D utility routine in Marc Volume D: User
Subroutines and Special Routines.
Dynamics
Marcs dynamic analysis capability allows you to perform the following calculations:
Modal (Eigenvalue) Analysis
Harmonic Response
Spectrum Response
Transient Analysis
The program contains two methods for eigenvalue extraction and three time integration operators. Nonlinear effects,
including material nonlinearity, geometric nonlinearity, and boundary nonlinearity, can be incorporated.
Linear problems can be analyzed using modal superposition or direct integration. All nonlinear problems should be
analyzed using direct integration methods.
In addition to distributed mass, you can also attach concentrated masses associated with each degree of freedom of the
system. You can include damping in either the modal superposition or the direct integration methods. You can also
include (nonuniform) displacement and/or velocity as an initial condition, and apply time-dependent forces and/or
displacements as boundary conditions.
Modal (Eigenvalue) Analysis
Marc uses either the inverse power sweep method or the Lanczos method to extract eigenvalues (natural frequencies)
and eigenvectors (mode shapes). The DYNAMIC parameter is used to determine which procedure to use, and how many
modes are to be extracted. The inverse power sweep method is typically used for extracting a few modes while the
Lanczos method is optimal for several modes. After the modes are extracted, they can be used in a transient analysis
or spectrum response calculation.

In dynamic eigenvalue analysis, we find the solution to an undamped linear dynamics problem:

where is the stiffness matrix, is the mass matrix, are the eigenvalues (frequencies) and are the eigenvectors.
In Marc, if the extraction is performed after increment zero, is the tangent stiffness matrix, which can include
material and geometrically nonlinear contributions. The mass matrix is formed from both distributed mass and point
masses.
Inverse Power Sweep
Marc creates an initial trial vector. To obtain a new vector, the program multiplies the initial vector by the mass matrix
and the inverse (factorized) stiffness matrix. This process is repeated until convergence is reached according to either
of the following criteria: single eigenvalue convergence or double eigenvalue convergence. In single eigenvalue
convergence, the program computes an eigenvalue at each iteration. Convergence is assumed when the values of two
successive iterations are within a prescribed tolerance. In double eigenvalue convergence, the program assumes that
the trial vector is a linear combination of two eigenvectors.
Using the three latest vectors, the program calculates two eigenvalues. It compares these two values with the two
values calculated in the previous step; convergence is assumed if they are within the prescribed tolerance.
When an eigenvalue has been calculated, the program either exits from the extraction loop (if a sufficient number of
vectors has been extracted) or it creates a new trial vector for the next calculation. If a single eigenvalue was obtained,
Marc uses the double eigenvalue routine to obtain the best trial vector for the next eigenvalue. If two eigenvalues were
obtained, the program creates an arbitrary trial vector orthogonal to the previously obtained vectors.
After Marc has calculated the first eigenvalue, it orthogonalizes the trial vector at each iteration to previously extracted
vectors (using the Gram-Schmidt orthogonalization procedure). Note that the power shift procedure is available with
the inverse power sweep method.
To select the power shift, set the following parameters:
Initial shift frequency This is normally set to zero (unless the structure has rigid body modes, preventing a
decomposition around the zeroth frequency).
Number of modes to be extracted between each shift A value smaller than five is probably not economical
because a shift requires a new decomposition of the stiffness matrix.
Auto shift parameter When you decide to do a shift, the new shift point is set to
Highest frequency
2
+ scalar x (highest frequency - next highest frequency)
2
You can define the value of the scalar through the MODAL SHAPE option.
The Lanczos Method
The Lanczos algorithm converts the original eigenvalue problem into the determination of the eigenvalues of a tri-
diagonal matrix. The method can be used either for the determination of all modes or for the calculation of a small
number of modes. For the latter case, the Lanczos method is the most efficient eigenvalue extraction algorithm. A
simple description of the algorithm is as follows.
Consider the eigenvalue problem:
(5-125)
K
2
M ( ) 0 =
K M
K
2
M u K u + 0 =
163
CHAPTER 5
Equation (5-125) can be rewritten as:
(5-126)
Consider the transformation:
(5-127)
Substituting Equation (5-127) into Equation (5-126) and premultiplying by the matrix on both sides of the
equation, we have
(5-128)
The Lanczos algorithm results in a transformation matrix such that:
(5-129)
(5-130)
where the matrix is a symmetrical tri-diagonal matrix of the form:
(5-131)
Consequently, the original eigenvalue problem, Equation (5-126), is reduced to the following new eigenvalue
problem:
(5-132)
The eigenvalues in Equation (5-132) can be calculated by the standard QL-method.
Through the MODAL SHAPE option, you can either select the number of modes to be extracted, or a range of modes
to be extracted. The Sturm sequence check can be used to verify that all of the required eigenvalues have been found.
In addition, you can select the lowest frequency to be extracted to be greater than zero.
The Lanczos procedure also allows you to restart the analysis at a later time and extract additional roots. It is
unnecessary to recalculate previously obtained roots using this option.
Convergence Controls
Eigenvalue extraction is controlled by:
The maximum number of iterations per mode in the power sweep method; or the maximum number of
iterations for all modes in the Lanczos iteration method,
1
2
------ M u M K
1
M u =
u Q =
Q
T
1
2
------ Q
T
M Q Q
T
M K
1
M Q =
Q
Q
T
M Q I =
Q
T
M K
1
MQ T =
T
T
1

2
0
2

2

3
m
0
m

m
=
1
2
------ T =

5+
Modal Stresses and Reactions
After the modal shapes (and frequencies) are extracted, the RECOVER history definition option allows for the
recovery of stresses and reactions at a specified mode. This option can be repeated for any of the extracted modes. The
stresses are computed from the modal displacement vector ; the nodal reactions are calculated from
. The RECOVER option is also used to place eigenvectors on the post file.
Harmonic Response
Harmonic response analysis allows you to analyze structures vibrating around an equilibrium state. This equilibrium
state can be unstressed or statically prestressed. Statically prestressed equilibrium states can include material and/or
geometric nonlinearities. You can compute the damped response for prestressed structures at various states.
In many practical applications, components are dynamically excited. These dynamic excitations are often harmonic
and usually cause only small amplitude vibrations. Marc linearizes the problem around the equilibrium state. If the
equilibrium state is a nonlinear, statically prestressed situation, Marc considers all effects of the nonlinear deformation
on the dynamic solution. These effects include the following:
initial stress
change of geometry
influence on constitutive law
The vibration problem can be solved as a linear problem using complex arithmetic.
The analytical procedure consists of the following steps:
1. Marc calculates the response of the structure to a static preload (which can be nonlinear) based on the
constitutive equation for the material response. In this portion of the analysis, the program ignores inertial
effects.
2. Marc calculates the complex-valued amplitudes of the superimposed response for each given frequency, and
amplitude of the boundary tractions and/or displacements. In this portion of the analysis, the program considers
both material behavior and inertial effects.
3. You can apply different loads with different frequencies or change the static preload at your discretion. All data
relevant to the static response is stored during calculation of the complex response.
To initiate a harmonic response analysis, use the HARMONIC parameter. To define the excitation frequency, use the
HARMONIC history definition option. If you enter the HARMONIC history definition option with a set of incremental
data, Marc assumes those incremental data apply only for the harmonic excitation. This is true for applied boundary
conditions as well as loads.
The small amplitude vibration problem can be written with complex arithmetic as follows
(5-133)
where
= is the complex response vector,
= is the complex load vector,
F K
2
M =
K i D
2
M + [ ]u P =
u u
r e
i u
i m
+
P
P
re
i P
i m
+
165
CHAPTER 5
The notation is further defined below:
(5-134)
where
(5-135)
where
(5-136)
where
If all external loads and forced displacements are in phase and the system is undamped, this equation reduces to
(5-137)
which could be solved without activating the complex arithmetic on the HARMONIC parameter.
The values of the damping coefficients ( , , ) are entered via the DAMPING model definition option. The spring
and damper contributions are entered in either the SPRINGS, PFAST, or PBUSH model definition options and mass
points are specified in the MASSES, CONM1, or CONM2 model definition options.
The element damping matrix ( ) can be obtained for any material with the use of a material damping matrix which
is specified in the UCOMPL user subroutine. You specify the material response with the constitutive equation.
= ,
is the excitation frequency.
are element stiffness matrices,
are the spring stiffness matrices.
are element mass matrices,
are mass point contributions.
are element damping matrices,
are damper contributions,
is the mass damping coefficient,
is the stiffness damping coefficient,
is the numerical damping coefficient.
i
1
K K
el
K
sp
+ =
K
el
K
sp
M M
el
M
mp
+ =
M
el
M
mp
D D
el
D
d
M
2
----- +
\ .
| |
K +
\ .
| |
+ + =
D
el
D
d
K
2
M ( )u
r e
P
r e
=

D
el

(5-138)
where and can be functions of deformation and/or frequency.
The global damping matrix is formed by the integrated triple product. The following equation is used:
(5-139)
where is the strain-displacement relation.
Similarly, the stiffness matrix is based on the elastic material matrix .
A special application of the harmonic excitation capability involves the use of the elastomeric analysis capability in
Marc. Here, the Mooney formulation (used in conjunction with the various Herrmann elements) is used to model the
stress-strain behavior of the elastomeric compound. In Marc, the behavior is derived from the third order invariant
form of the strain energy density function.
(5-140)
with the incompressibility constraint
(5-141)
where , , and are the invariants of the deformation. For the harmonic excitation, the constitutive equation has
the specific form
(5-142)
with as the quasi-static moduli following form the Mooney strain energy density function and
(5-143)
The output of Marc consists of stresses, strains, displacements and reaction forces, all of which may be complex
quantities. The strains are given by
(5-144)
and the stresses by
(5-145)
B C
+ =
B C
D
T
C V
el
d
v
el
}
el
K B
W I
1
I
2
, ( ) C
10
I
1
3 ( ) C
01
I
2
3 ( ) C
11
I
1
3 ( ) I
2
3 ( ) + + =
C
20
+ I
1
3 ( )
2
C
30
I
1
3 ( )
3
+
I
3
1 =
I
1
I
2
I
3
S
i j
D
i j kl
2i
i j kl
+ [ ]E
kl
=
D
i j kl
i j kl

0
C
i k
1
C
j l
1
C
i l
1
C
j k
1
+ [ ]
1

i k
C
j l
1
C
i l
1
j k
+ [ ]
2
C
i k
C
j l
1
C
i l
1
C
j k
+ [ ]
10
i j
C
kl
1
11
i j
kl
12
i j
C
kl

20
+ C
i j
C
kl
1
21
+ C
i j
kl

22
C
i j
C
kl
+ +
+ + +
+
=
u =
B C + =
167
CHAPTER 5
The reaction forces are calculated with
(5-146)
where is only included if requested on the HARMONIC parameter.
The printout of the nodal values consists of the real and imaginary parts of the complex values, but you can request
that the amplitude and phase angle be printed. You do this with the PRINT CHOICE model definition option.
Spectrum Response
The spectrum response capability allows you to obtain maximum response of a structure subjected to known spectral
base excitation response. This is of particular importance in earthquake analysis and random vibration studies. You
can use the spectrum response option at any point in a nonlinear analysis and, therefore, ascertain the influence of
material nonlinearity or initial stress.
The spectrum response capability technique operates on the eigenmodes previously extracted to obtain the maximum
nodal displacements, velocities, accelerations, and reaction forces. You can choose a subset of the total modes
extracted by either specifying the lowest n modes or by selecting a range of frequencies.
Enter the displacement response spectrum for a particular digitized value of damping through the
RESPONSE SPECTRUM model definition or the USSD user subroutine. Marc performs the spectrum analysis based
on the latest set of modes extracted. The program lumps the mass matrix to produce . It then obtains the projection
of the inertia forces onto the mode
(5-147)
The spectral displacement response for the j
th
mode is
(5-148)
Marc then calculates the square roots of the sum of the squares as
(5-149)
(5-150)
(5-151)
(5-152)
R Ku
2
Mu iD
el
u iD
d
u + + =
2
Mu
S
D
( )
M
j
P
j
M
j
=
j
S
D

j
( ) P
j
=
u
j
j
( )
2
j
1 2
= DISPLACEMENT
v
j
j
( )
2
j
1 2
= VELOCITY
a
j
j
2
j
( )
2
j
1 2
= ACCELERATION
f
j
j
2
M
j
( )
2
j
1 2
= FORCE

The internal forces given by Equation 5-152 are identified as reaction forces on the post file. The force transmitted by
the structure to the supporting medium (also referred to as base shear) is only reported in the out file and is given by
(5-153)
Transient Analysis
Transient dynamic analysis deals with an initial-boundary value problem. In order to solve the equations of motion of
a structural system, it is important to specify proper initial and boundary conditions. You obtain the solution to the
equations of motion by using either modal superposition (for linear systems) or direct integration (for linear or
nonlinear systems). In direct integration, selecting a proper time step is very important. For both methods, you can
include damping in the system.
The following sections discuss the seven aspects of transient analysis listed below.
Modal Superposition
Direct Integration
Time Step Definition
Initial Conditions
Time-Dependent Boundary Conditions
Mass Matrix
Damping
Modal Superposition
The modal superposition method predicts the dynamic response of a linear structural system. In using this method, we
assume that the dynamic response of the system can be expressed as a linear combination of the mode shapes of the
system. For the principle of superposition to be valid, the structural system must be linear. Damping can be applied to
each mode used in the superposition procedure.
To select the eigenvalue extraction method and the number of modal shapes, use the DYNAMIC parameter. To select a
fraction of critical damping for each mode, use the DAMPING model definition option. The DYNAMIC CHANGE
history definition option can be used to select the time step.
Marc obtains the transient response on the basis of eigenmodes extracted. The number of modes extracted (rather than
the choice of time step) governs the accuracy of the solution.
Consider the general linear undamped problem
(5-154)
and suppose that n eigenvectors are known.
(5-155)
t f S
D

j
( )
j
2
P
j
2
j
=
TRANSMITTED FORCE
M u
K u + f t ( ) =
1

n
, ,
u
i
u
i
t ( ) =
169
CHAPTER 5
The eigenmodes are orthogonal with respect to the and matrices. After substitution in Equation (5-154), we find
a set of uncoupled dynamic equations
(5-156)
where:
(5-157)
(5-158)
(5-159)
We can introduce damping on each mode as a fraction of critical damping ( ) for that mode. We can rewrite Equation
(5-156) in the form
(5-160)
or if we normalize such that :
(5-161)
The response of a particular mode is then given by the solution of Equation (5-161) which is:
(5-162)
with:
The evaluation of the Duhamel integral, Equation (5-162), can be performed exactly if the load changes linearly in a
specific time increment. Hence, if the load changes rapidly in a specific time period, small load steps have to be taken.
In case initial displacements or initial velocities are given a transformation to the reduced modal system is
needed for those conditions:
and
M K
m
i
u
i
k
i
u
i
+ f
i
t ( ) =
m
i

i
T
M
i
=
k
i

i
T
K
i
=
f
i
t ( )
i
T
f t ( ) =
i
m
i
u
i
2m
i
i
u
i
k
i
u
i
+ + f
i
t ( ) =
i
m
i
1 =
u
i
2
i
i
u
i

i
2
u
i
+ + f
i
t ( ) =
u
i
t ( ) f
i
( )h t ( ) d
0
t
}
=
h t ( )
1
i
d
------- e
i
t
i
d
t ( ) sin = t 0
h t ( ) 0 = t 0 <
i
d
1
i
2
( )
i
=
u
0
u
0
u
i
0
i
T
M u
0
= u
i
0
i
T
M u
0
=

The solution of these initial conditions which must be added to the response given in Equation (5-162) is:
(5-163)
The initial accelerations due to given initial displacements and initial velocities can be obtained by
differentiation of Equation (5-163):
Direct Integration
Direct integration is a numerical method for solving the equations of motion of a dynamic system. It is used for both
linear and nonlinear problems. In nonlinear problems, the nonlinear effects can include geometric, material, and
boundary nonlinearities. For transient analysis, Marc offers four direct integration operators listed below.
Newmark-beta Operator
Houbolt Operator
Generalized-Alpha Operator
Central Difference Operator
To select the direct integration operator, use the DYNAMIC parameter. Specify the time step size through the DYNAMIC
CHANGE or AUTO STEP option. Direct integration techniques are imprecise; this is true regardless of which technique
you use. Each technique exhibits at least one of the following problems: conditional stability, artificial damping, and
phase errors.
This operator is probably the most popular direct integration method used in finite element analysis. For linear
problems, it is unconditionally stable and exhibits no numerical damping. The Newmark-beta operator can effectively
obtain solutions for linear and nonlinear problems for a wide range of loadings. The procedure allows for change of
time step, so it can be used in problems where sudden impact makes a reduction of time step desirable. This operator
can be used with adaptive time step control. Although this method is stable for linear problems, instability can develop
if nonlinearities occur. By reducing the time step and/or adding (stiffness) damping, you can overcome these problems.
Houbolt Operator
This operator has the same unconditional stability as the Newmark-beta operator. In addition, it has strong numerical
damping characteristics, particularly for higher frequencies. This strong damping makes the method very stable for
nonlinear problems as well. In fact, stability increases with the time step size. The drawback of this high damping is
that the solution can become inaccurate for large time steps. Hence, the results obtained with the Houbolt operator
usually have a smooth appearance, but are not necessarily accurate. The Houbolt integration operator, implemented in
Marc as a fixed time step procedure, is particularly useful in obtaining a rough scoping solution to the problem.
u
i
t ( ) e

i
i
t
u
i
0
u
i
0
i
+
i
d
-------------------------------
i
d
t ( ) sin u
i
0
i
d
t ( ) cos + =
u
0
u
0
u
i
0
i
2
u
i
0
( ) 2
i

i
u
i
0
( ) =
171
CHAPTER 5
Single Step Houbolt Operator
Two computational drawbacks of the Houbolt operator are the requirement of a special starting procedure and the
restriction to fixed time steps. In [Ref. 28], a Single Step Houbolt procedure has been presented, being unconditionally
stable, second order accurate and asymptotically annihilating. In this way, the algorithm is computationally more
convenient compared to the standard Houbolt method, but because of its damping properties, the time steps have to
be chosen carefully. This algorithm is recommended for implicit dynamic contact analyses.
Generalized Alpha Operator
One of the drawbacks of the existing implicit operators is the inability to easily control the numerical dissipation.
While the Newmark-Beta method has no dissipation and works well for regular vibration problems, the Single-Step
Houbolt method has numerical dissipation and works well for implicit dynamic contact problems. A single scheme
that easily allows zero/small dissipation for regular structural dynamic problems and high-frequency numerical
dissipation for dynamic contact problems is desirable. In [Ref. 35], a Generalized-alpha method has been presented as
an unconditionally stable, second-order algorithm that allows user-controllable numerical dissipation. The dissipation
is controlled by choosing either the spectral radius S of the operator or alternatively, two parameters
f
and
m
. The
choice of the parameters provides a family of time integration algorithms that encompasses the Newmark-Beta,
Single-Step Houbolt and the Hilber-Hughes-Taylor time integration methods as special cases.
Central Difference Operators
These explicit operators for IDYN = 4 and IDYN = 5 are only conditionally stable. The program automatically
calculates the maximum allowable time step. This method is not very useful for shell or beam structures because the
high frequencies result in a very small stability limit. This method is particularly useful for analysis of shock-type
phenomena. In this procedure, since the operator matrix is a diagonal mass matrix, no inverse of operator matrix is
needed. However, this fact also implies that you cannot use this method in problems having degrees of freedom with
zero mass. This restriction precludes use of the Herrmann elements, gap-friction elements, the pipe bend element, shell
elements 72 and 89 and beam elements 76 and 77. These shell and beam elements are precluded because they have a
rotational degree of freedom, which do not have an associated mass. The mass is updated only if the LARGE STRAIN
or UPDATE parameter or the CONTACT option is used. The elastomer capability can be used with explicit dynamics
in an updated Lagrange framework where the pressure variables are condensed out before going into the solver.
Consider the equations of motion of a structural system:
(5-164)
where , , and are mass, damping, and stiffness matrices, respectively, and , , , and are acceleration,
velocity, displacement, and force vectors. Various direct integration operators can be used to integrate the equations
of motion to obtain the dynamic response of the structural system. The technical background of the three direct
integration operators available in Marc is described below.
The generalized form of the Newmark-beta operator is
(5-165)
Ma Cv Ku + + F =
M C K a v u F
u
n 1 +
u
n
tv
n
1 2 ( )t
2
a
n
t
2
a
n 1 +
+ + + =

(5-166)
where superscript denotes a value at the nth time step and , , and take on their usual meanings.
The particular form of the dynamic equations corresponding to the trapezoidal rule
results in
(5-167)
where the internal force is
(5-168)
Equation (5-167) allows implicit solution of the system
(5-169)
Notice that the operator matrix includes , the tangent stiffness matrix. Hence, any nonlinearity results in a
reformulation of the operator matrix. Additionally, if the time step changes, this matrix must be recalculated because
the operator matrix also depends on the time step. It is possible to change the values of and through the
PARAMETERS option.
Houbolt Operator
The Houbolt operator is based on the use of a cubic fitted through three previous points and the current (unknown) in
time. This results in the equations
(5-170)
and
(5-171)
Substituting this into the equation of motion results in
(5-172)
v
n 1 +
v
n
1 ( )ta
n
ta
n 1 +
+ + =
n
u v a
1 2 = 1 4 = ,
4
t
2
-------- M
2
t
----- C K + +
\ .
| |
u F
n 1 +
R
n
M a
n
4
t
----- v
n
+
\ .
| |
Cv
n
+ + =
R
R
T
dv
V
}
=
u
n 1 +
u
n
u + =
K

v
n 1 +
11
6
------ u
n 1 +
3u
n
3
2
--- u
n 1
1
3
--- u
n 2
+
\ .
| |
t =
a
n 1 +
2u
n 1 +
5u
n
4u
n 1
u
n 2
+ ( ) t
2
=
2
t
2
-------- M
11
6t
--------- C K + +
\ .
| |
u F
n 1 +
R
n
1
t
2
-------- 3u
n
4u
n 1
u
n 2
+ ( )M + + =
1
t
-----
7
6
--- u
n
3
2
--- u
n 1
1
3
--- u
n 2
+
\ .
| |
C
173
CHAPTER 5
This equation provides an implicit solution scheme. By solving Equation (5-169) for , you obtain Equation (5-173),
and so obtain and .
(5-173)
Equation (5-172) is based on uniform time steps errors occur when the time step is changed. Also, a special starting
procedure is necessary since and appear in Equation (5-172).
Single Step Houbolt Operator
The Single Step Houbolt operator according to [Ref. 28] starts with the following equilibrium equation and
expressions for the velocity and acceleration:
(5-174)
(5-175)
(5-176)
Notice that in contrast to the Newmark and the standard Houbolt method, the equilibrium equation also contains terms
corresponding to the beginning of the increment. Without loss of generality, the parameter can be set to 1. Based
on asymptotic annihilation and second order accuracy, the remaining parameters can be shown to fulfill:
, , , , ,
, , ,
In this way, the number of unknown parameters has been reduced to two. Based on a Taylor series expansion of the
displacement about the nth time step, and should be related by , which finally yields
. According to [Ref. 28], should be set to 3/2 (with ) to minimize the velocity
error and to 1/2 (with ) to avoid velocity overshoot. The default values in Marc are and
, but the user can modify and using the PARAMETERS model definition and history definition
option.
u
v
n 1 +
a
n 1 +
u
n 1 +
u
n
u + =
u
n 1
u
n 2
m1
Ma
n 1 +
c1
Cv
n 1 +
k1
Ku
n 1 +
m
Ma
n
c
Cv
n
k
Ku
n
+ + + + + =
f 1
F
n 1 +
a
f
F
n
+
u
n 1 +
u
n
t v
n
t
2
a
n
1
t
2
a
n 1 +
+ + + =
v
n 1 +
v
n
ta
n
1
t a
n 1 +
+ + =
m1
k
0 = =
1

1
+ =
m
1 2 =
k1
1 2
1
=
c
2
1
+ ( ) 4
1
2
=
c1
2 3
1
+ ( ) 4
1
2
=
f
k
=
f 1
k1
=

1

1
+ 1 2 =
1 2 1 2
1
( ) =
1
1 2 =
0 =
1
3 2 =
1 2 =
1

Substitution of the velocity and acceleration into the equilibrium equation results in:
(5-177)
Generalized Alpha Operator
From [Ref. 35], the equilibrium equations for the generalized alpha method can be expressed in the form
(5-178)
where
(5-179)
(5-180)
(5-181)
The displacement and velocity updates are identical to those of the Newmark algorithm
(5-182)
where, as shown in [Ref. 35], optimal values of the parameters and are related to and by
(5-183)
(5-184)
In Equation (5-168), = =0 gives rise to the Newmark-beta method, =0, gives rise to the
HHT - a method [Ref. 36], and =0, gives rise to the WBZ - a method [Ref. 37].
1
1
t
2
k1
------------------------ M

c1
1
t
k1
---------------------- C K + +
)
`

u F
n 1 +
Ku
n
1
t
2
k1
------------------------ M t v
n
t
2
a
n
+ { }

m
k1
--------- Ma
n
c1
k1
--------- C v
n
ta
n
1
1
t
------------ t v
n
t
2
a
n
+
)
`

+
)
`

k1
--------- Cv
n
+ =
M
a
n 1
m
+ +
C
v
n 1
f
+ +
K
u
n 1
f
+ +
+ + F
n 1
f
+ +
=
u
n 1
f
+ +
1
f
+ ( )u
n 1 +
f
u
n
=
v
n 1
f
+ +
1
f
+ ( )v
n 1 +
f
v
n
=
a
n 1
m
+ +
1
m
+ ( )a
n 1 +
m
a
n
=
u
n 1 +
u
n
tv
n
1 2 ( )t
2
a
n
t
2
a
n 1 +
+ + + =
v
n 1 +
v
n
1 ( )ta
n
ta
n 1 +
+ + =

f

m
1
4
--- 1
m

f
+ ( )
2
=
1
2
---
m

f
+ =
f

m

m
0.33
f
0 <
f
0 <
m
1
175
CHAPTER 5
It is seen that the and parameters can be used to control the numerical dissipation of the operator. A simpler
measure is the spectral radius S. This is also a measure of the numerical dissipation; a smaller spectral radius value
corresponds to greater numerical dissipation. The spectral radius of the generalized alpha operator can be related to
the and parameters as follows
(5-185)
(5-186)
S varies between 0 and 1. Accordingly, the ranges for the and parameters are given by and
. = -0.5, = -0.5 corresponds to a spectral radius of 1.0 and = 0, = 1 corresponds to a
spectral radius of 0.0. By substituting the values in equation (5-145) and comparing with Equation (5-164), it can be
noted that the special case of S = 0 ( = 0, = 1) for the generalized alpha operator is identical to the default single
step houbolt operator with and . It can also be noted that the case of S = 1 has no dissipation
and corresponds to a mid-increment Newmark- beta operator.
In the Marc input file, the generalized alpha operator is flagged through the option 8 on the second field of the
DYNAMIC parameter. Options to define the associated generalized alpha parameters are available through the
DYNAMIC parameter and the PARAMETERS model definition and history definition option. On the DYNAMIC
parameter, a default value of 0 on the 8th field indicates contact-optimized parameters ( = 0, = 1) and a 1
on the 8th field indicates non-contact optimized parameters ( = -0.05, = 0). 0 would be recommended for
dynamic contact problems or for any problem where high-frequency numerical dissipation is desired. 1 would be
recommended for noncontact structural dynamic problems and free-vibration problems where a small amount of
numerical dissipation is desired. On the PARAMETERS, the 4th, 5th, and 6th fields of the 5th data block are reserved
for , and S, respectively. S is the main control parameter: indicates that Equations (5-175) and
(5-176) are used to calculate and ; S = -1 indicates that and remain unchanged from previously set
values; S = -2 indicates contact-optimized parameters; S = -3 indicates noncontact optimized parameters; S = -4
indicates that the 4th and 5th fields will be checked and if valid, used to set and directly.
Central Difference Operator
The central difference operator assumes a quadratic variation in the displacement with respect to time.
(5-187)
(5-188)
so that
(5-189)
f

m
f

m
f
S
1 S +
------------- =
m
1 2S
1 S +
---------------- =
f

m
0.5
f
0.0
0.5
m
1
f

m

f

m
f

m
1
3 2 = 1 2 =
f

m
f

m
f

m
0 S 1
f

m

f

m
f

m
a
n
v
n 1 2 +
v
n 1 2
( ) t ( ) =
v
n
u
n 1 2 +
u
n 1 2
( ) t ( ) =
a
n
u
n 1 +
u
n
( ) t
2
( ) =

where
(5-190)
for IDYN=4:
(5-191)
for IDYN=5:
(5-192)
Since the mass matrix is diagonal, no inverse of the operator matrix is needed. Also, since the operator is only
conditionally s, the critical time step is evaluated at the beginning of the analysis. For IDYN = 4, the critical time step
is computed by a power sweep for the highest mode in the system only at the beginning of the analysis. For IDYN =
4, no damping is included. For IDYN = 5, an approximated method based on element geometry is used to compute the
highest eigenvalue. The critical time step is calculated at a user-specified frequency or every 100 steps. The variable
time step can be used only for IDYN = 5. Unless there is significant distortion in an element or material nonlinearity,
the change of critical time step is not significant.
Time Step Definition
In a transient dynamic analysis, time step parameters are required for integration in time. The DYNAMIC CHANGE
option can be used for either the modal superposition or the direct integration procedure. The AUTO STEP option can
be used for the Newmark-beta, the Single Step Houbolt, and the Generalized-alpha operator to invoke the adaptive
time control. Enter parameters to specify the time step size and period of time for this set of boundary conditions.
When using the Newmark-beta operator, decide which frequencies are important to the response. The time step in this
method should not exceed 10 percent of the period of the highest relevant frequency in the structure. Otherwise, large
phase errors will occur. The phenomenon usually associated with too large a time step is strong oscillatory
accelerations. With even larger time steps, the velocities start oscillating. With still larger steps, the displacement
eventually oscillates. In nonlinear problems, instability usually follows oscillation. When using adaptive dynamics,
you should prescribe a maximum time step.
As in the Newmark-beta operator, the time step in Houbolt integration should not exceed 10 percent of the period of
the highest frequency of interest. However, the Houbolt method not only causes phase errors, it also causes strong
artificial damping. Therefore, high frequencies are damped out quickly and no obvious oscillations occur. It is,
therefore, completely up to the engineer to determine whether the time step was adequate.
For the Generalized-alpha operator, depending on the chosen parameters, the integration scheme can vary between the
Newmark-beta operator and the Single-step houbolt operator. For spectral radii < 1, there is artificial damping in the
system. Depending on the type of problem, the Generalized-alpha parameters and the associated time step should be
carefully chosen to reduce phase errors and effects of artificial damping.
In nonlinear problems, the mode shapes and frequencies are strong functions of time because of plasticity and large
displacement effects, so that the above guidelines can be only a coarse approximation. To obtain a more accurate
estimate, repeat the analysis with a significantly different time step (1/5 to 1/10 of the original) and compare responses.
u
n
u
n
u
n 1
=
M
t
2
-------- u
n 1 +
F
n
R
n
M
t
2
-------- u
n
+ =
M
t
2
-------- u
n 1 +
F
n
R
n
M
t
2
-------- u
n
Cv
n
1
2
---
+ =
177
CHAPTER 5
The central difference integration method is only conditionally stable; the program automatically calculates the stable
time step. This step size yields accurate results for all practical problems.
Initial Conditions
In a transient dynamic analysis, you can specify initial conditions such as nodal displacements and/or nodal velocities.
To enter initial conditions, use the following model definition options: INITIAL DISP for specified nodal
displacements, and INITIAL VEL for specified nodal velocities. As an alternative, you can use the USINC
user subroutine.
Dynamics
Time-Dependent Boundary Conditions
Simple time-dependent load or displacement histories can be entered on data blocks. However, in general cases with
complex load histories, it is often more convenient to enter the history through a user subroutine. Marc allows the use
of the FORCDT and FORCEM user subroutines for boundary conditions. the FORCDT user subroutine allows you to
specify the time-dependent incremental point loads and incremental displacements. The FORCEM user subroutine
allows you to specify the time-dependent magnitude of the distributed load.
Mass Matrix
The mass matrix is a discrete representation of the system mass. System mass can be defined through either distributed
masses and concentrated masses.
Distributed masses are defined for elements through the mass density material property. Marc offers both consistent
and lumped element mass matrices. The consistent mass matrix is given by
where is the mass density and is the shape function matrix. The lumped mass matrix is flagged through the
LUMP parameter. Marc uses the Hinton, Rock, Zienkiewicz lumping scheme to produce a diagonal mass matrix. The
salient features of this scheme are as follows: The diagonal coefficients of the consistent mass matrix are computed.
The total mass of the element is also computed. A scale factor s is computed by adding the translational diagonal
coefficients that are mutually parallel and in the same direction. All diagonal coefficients are then scaled by ,
thereby preserving the total mass of the element.
Concentrated masses are defined through the MASSES, CONM1, and/or CONM2 model definition options. The
MASSES option allows the placement of particle masses at degree of freedom of at specific nodes. Both
translational and rotary masses are allowed.
m [ ] N [ ]
T
N [ ] V d
}
=
N [ ]
m
m s
m
i
i

The CONM1 option allows the specification of a diagonal or 6 x 6 symmetric mass matrix at a node in the global
cartesian or in a local coordinate system. The local system is specified through a coordinate system ID. The fully
coupled 3-D form is given below
Each mass term can be varied using tables. Independent variables allowed include time, increment number,
coordinates, temperature. When the diagonal form is specified, only the to

diagonal terms need to be
specified. In 2-D, the mass matrix reduces to 3 x 3 and only the , terms need to be specified. If the
CONM1 mass matrix is specified in a local system, then it is transformed to the global system before assembly.
The CONM2 option allows a concentrated diagonal mass and the mass moments of inertia with respect to the global
cartesian or a local coordinate system. The local system is specified through a coordinate system ID. The mass terms
for each degree of freedom at the node are a function of the concentrated mass, the moments of inertia and the distance
of the center of mass from the node. The offset distance from the node to the center of mass is given by the vector
. The 3-D form of the CONM2 matrix is given below:
where the combined moments of inertia are given as follows:
; ; ;
; ;
Each mass coefficient can be varied using tables. Independent variables allowed include time, increment number,
coordinates, temperature. In 2-D, the mass matrix reduces to 3 x 3 and only , , , and are required. If the
CONM2 mass matrix is specified in a local system, then it is converted to the global system before assembly.
m
conm1
[ ]
M
11

M
21
M
22
sym
M
31
M
32
M
33

M
41
M
42
M
43
M
44

M
51
M
52
M
53
M
54
M
55

M
61
M
62
M
63
M
64
M
65
M
66
=
M
11
M
66
M
11
M
12
M
33
V
1
V
2
V
3
{ }
m
conm2
[ ]
M
0 M sym
0 0 M
0 MV
3
MV
2
I
11
a

MV
3
0 MV
1
I
21
a
I
22
a
MV
2
MV
1
0 I
31
a
I
32
a
I
33
a
=
I
11
a
I
11
M V
2
2
V
3
2
+ ( ) + = I
22
a
I
22
M V
1
2
V
3
2
+ ( ) + = I
33
a
I
33
M V
1
2
V
2
2
+ ( ) + =
I
21
a
I
21
MV
1
V
2
= I
31
a
I
31
MV
1
V
3
= I
32
a
I
32
MV
2
V
3
=
M I
33
V
1
V
2
179
CHAPTER 5
Damping
In a transient dynamic analysis, damping represents the dissipation of energy in the structural system. It also retards
the response of the structural system.
Marc allows you to enter two types of damping in a transient dynamic analysis: modal damping and Rayleigh
damping. Use modal damping for the modal superposition method and Rayleigh damping for the direct integration
method. For modal superposition, you can include damping associated with each mode. To do this, use the DAMPING
option to enter the fraction of critical damping to be used with each mode.
During time integration, Marc associates the corresponding damping fraction with each mode. The program bases
integration on the usual assumption that the damping matrix of the system is a linear combination of the mass and
stiffness matrices, so that damping does not change the modes of the system.
For direct integration damping, you can specify the damping matrix as a linear combination of the mass and stiffness
matrices of the system. You can specify damping coefficients on an element basis.
Stiffness damping should not be applied to either Herrmann elements or gap elements because of the presence of
Lagrangian multipliers.
Numerical damping is used to damp out unwanted high-frequency chatter in the structure. If the time step is decreased
(stiffness damping might cause too much damping), use the numerical damping option to make the damping (stiffness)
coefficient proportional to the time step. Thus, if the time step decreases, high-frequency response can still be
accurately represented. This type of damping is particularly useful in problems where the characteristics of the model
and/or the response change strongly during analysis (for example, problems involving opening or closing gaps).
Element damping uses coefficients on the element matrices and is represented by the equation:
(5-193)
where
If the same damping coefficients are used throughout the structure, Equation (5-193) is equivalent to Rayleigh
damping.
is the global damping matrix
is the mass matrix of i
th
element
is the stiffness matrix of the i
th
element
is the mass damping coefficient on the i
th
element
is the usual stiffness damping coefficient on the i
th
element
is the numerical damping coefficient on the i
th
element
is the time increment
C
i
M
i

i

i
t
----- +
\ .
| |
K
i
+
)
`

i 1 =
n
=
C
M
i
K
i
i
t

The damping associated with springs and mass points can be controlled via the springs and masses input options. The
damping on elastic foundations is the same as the damping on the element on which the foundation is applied. For
springs, a dashpot can be added for nonlinear analysis.
Inertia Relief
Inertia relief refers to an analysis procedure that allows unconstrained systems to be subjected to a quasi-static analysis
by taking rigid body inertia forces into account. Examples of such systems include an aircraft in a steady turn, an
underwater structure in equilibrium under gravity and buoyancy. Conventional static analysis cannot be performed for
such systems since, in the absence of constraints, the stiffness matrix is singular. Inertia relief analysis is applicable to
such free bodies. The response is measured relative to a steady accelerating frame induced by the external loads.
Two important steps are followed by the solver in an analysis with inertia relief:
1. Rigid Body Mode evaluation: This is the central step in an inertia relief analysis since the rigid body modes are
extensively used in the load computations. The Support Method is currently provided in Marc to evaluate the
rigid body modes.
2. Inertia Relief Loads evaluation: The loads associated with the rigid body motion of the system are calculated
using the rigid body modes and the global lumped mass matrix. The loads are calculated such that they balance
the current external load vector in the system.
Next, more details on each of the steps are provided.
Rigid Body Mode Evaluation
Support Method
The Support Method requires the manual specification of nodal degrees of freedom which can be used to evaluate the
rigid body modes of the system. Any number of nodal entries can be provided, each with degrees of freedom ranging
from 1 to the maximum possible degrees of freedom at each node. These degrees of freedom are referred to as the r-
degrees of freedom set. It is the users responsibility to ensure that the r-degrees of freedom set is necessary and
sufficient to evaluate the rigid body modes of the system.
The rigid body modes are evaluated by sequentially setting each r-degrees of freedom to 1 with all other r-degrees of
freedom set to 0 and solving for the system of equations
(5-194)
where is the identity matrix for the r-degrees of freedom and are the modal amplitudes for the remaining
degrees of freedom. Once the rigid body modes are evaluated, the r-degrees of freedom are set to 0 and used as fixed
displacement constraints. Other displacement boundary conditions, specified through normal options like FIXED
DISP, DISP CHANGE, etc. can be used in conjunction with the rigid body degrees of freedom constraints.
The Support Method is computationally inexpensive since the provided Support degrees of freedom are automatically
constrained and no additional ties/constraints are needed while calculating the response. It is available for single-
processor and multi-processor runs (both single-input and multi-input file formats).
K
l l
[ ] K
l r
[ ]
K
r l
[ ] K
rr
[ ]
l
[ ]
I [ ] )
`

0 =
I [ ]
l
[ ]
181
CHAPTER 5
The Support Method requires manual input to evaluate the rigid body modes of the system. For direct solvers, the rigid
body modes are calculated through repeated back-solves of the factorized matrix. The computation of rigid body
modes is more expensive for the conjugate gradient iterative solver. Finally, the Support Method cannot be used in
contact problems with dynamically changing rigid body modes.
Rigid Body Mode Update
In a small displacement analysis, the rigid body modes are evaluated at the first cycle of the first increment of the
inertia relief loadcase and are retained through the loadcase. In a large displacement analysis, the rigid body modes
are updated at the beginning of each increment by using the current stiffness matrix while solving Equation (5-194).
Inertia Relief Loads Evaluation
A general derivation for the inertia relief load vector in a nonlinear analysis is presented here. At the th iteration of
the th increment, the total response is given by
(5-195)
where is the total response, is the rigid body response and is the flexible response. Neglecting
accelerations associated with the flexible response, the governing equations are given as
(5-196)
where and are the mass and stiffness matrices at the th iteration of th increment,
is the external load vector at the th increment, and are the internal load vector
and the inertia relief load vector at the th iteration of the th increment, respectively.
The change in the rigid body acceleration is computed as follows:
(5-197)
where is the matrix of rigid body modes at the th iteration of the th increment and
represents the iterative change in the modal acceleration vector.
Pre-multiplying Equation 5-184 by individual rigid body modes, , and using the fact that the work done
by the rigid body modes is 0, one gets
(5-198)
where the th term of matrix is given by
i
n 1 + ( )
u
t i ,
n 1 +
{ } u
r
n
{ } u
r i 1 ,
n 1 +
{ } u
r i ,
n 1 +
{ } u
n
{ } u
i 1
n 1 +
{ } u
i
n 1 +
{ } + + + + + =
u
t
{ } u
r
{ } u { }
M
i 1
n 1 +
[ ] u
r i ,
n 1 +
{ } K
i 1
n 1 +
[ ] u
i
n 1 +
{ } + P
n 1 +
{ } I
i 1
n 1 +
{ } P
i 1 ,
n 1 +
{ } =
M
i 1
n 1 +
[ ] K
i 1
n 1 +
[ ] i 1 ( ) n 1 + ( )
P
n 1 +
{ } n 1 + ( ) I
i 1
n 1 +
P
i 1 ,
n 1 +
{ }
i 1 ( ) n 1 + ( )
u
r i ,
n 1 +
{ }
i 1
n 1 +
[ ] A
i ,
n 1 +
{ } =
i 1
n 1 +
[ ] i 1 ( ) n 1 + ( )
A
i ,
n 1 +
{ }
k i 1 ,
n 1 +
{ }
T
C
i 1
n 1 +
[ ] A
i ,
n 1 +
{ } R
i
n 1 +
{ } =
kl C
i 1
n 1 +
[ ]
C
i 1 kl ,
n 1 +
k i 1 ,
n 1 +
{ }
T
M
i 1
n 1 +
[ ]
l i 1 ,
n 1 +
{ } =

and the th term of is given by
(5-199)
Note that if the rigid body modes are mass ortho-normalized, matrix is an identity matrix. Solving Equation
(5-198) for , the iterative change in the inertia relief load vector is given by
(5-200)
The governing equations in Equation 5-184 can then be rewritten in the form,
where the inertia relief load vector at the th iteration of the th increment, , is given by
Additional items to be kept in consideration for inertia relief analysis in Marc are as follows:
Convergence Checking
For systems that are unconstrained or just constrained against rigid body motion (statically determinate structures), the
reaction forces at nodes are 0 or very close to 0. In this case, using relative residual force checking may lead to
unnecessary recycling and possible non-convergence. It would be more appropriate to use displacement checking or
absolute residual force checking in such situations.
Time Stepping
Inertia relief can be used in conjunction with all time stepping procedures used for static analyses. Time step cutback
schemes are supported. Inputs specified in either table or non-table formats are also supported.
Parallel Processing
The Support Method is available for serial and parallel runs. For the multi-input file parallel mode, the Support nodes
should be available in all the domains. For the single-input file parallel mode, the inertia relief option should be present
in the model definition section itself in order to allow the Support nodes to be added to domains they are not naturally
present in.
Deactivating Inertia Relief
When inertia relief is active during an analysis, and is subsequently turned off, the user has three choices for treating
the pre-existing inertia relief load vector.
1. The inertia relief load vector can be retained. No new inertia relief loads are computed, but pre-existing inertia
relief loads are included in the right-hand side.
k R
i
n 1 +
{ }
R
i k ,
n 1 +
k i 1 ,
n 1 +
{ }
T
P
n 1 +
{ } I
i 1
n 1 +
{ } P
i 1 ,
n 1 +
{ } [ ] =
C
i 1
n 1 +
[ ]
A
i ,
n 1 +
{ }
P
i ,
n 1 +
{ } M
i 1
n 1 +
[ ]
i 1
n 1 +
[ ] A
i ,
n 1 +
{ } =
K
i 1
n 1 +
[ ] u
i
n 1 +
{ } P
i
n 1 +
{ } I
i 1
n 1 +
{ } P
i ,
n 1 +
{ } =
i n 1 + ( ) P
i ,
n 1 +
{ }
P
i ,
n 1 +
{ } P
i 1 ,
n 1 +
{ } P
i ,
n 1 +
{ } + =
183
CHAPTER 5
2. The inertia relief load vector can be removed instantaneously. Note that this is the option used by Marc if the
INERTIA RELIEF option is present in one loadcase and is absent in the next loadcase. This could lead to
convergence difficulties at the beginning of the second loadcase due to the large change in load vectors.
3. The inertia relief load vector can be removed gradually. This allows the inertia relief load vector to be removed
over the course of the second loadcase.
Inertia Relief Ouput Options
The nodal inertia relief load and moment vectors are written to the post file by default using nodal post codes 51 and
52, respectively. It should be noted that the inertia relief loads are always written out in the global cartesian system.
Also, the command INER can be used on the PRINT NODE option to obtain the inertia relief loads at a particular node.
Also, when the NO PRINT option is revoked, a summary of the total inertia relief load is printed in the out file.
Limitations for Inertia Relief Analyses
1. All inertia relief analyses provide the uninteresting solution of zero displacements when a uniform mesh is
subjected to uniform gravity loads or when a uniform shell mesh is subjected to uniform distributed loads. This
is as expected because the inertia relief equations place an inertia load on each node that is equal and opposite
to the external load.
2. The Support Method is currently not supported for the CASI iterative solver.
3. The Support Method is capable of dealing with material and geometric non-linearities. However, boundary
condition non-linearities like contact between multiple bodies cause the number of rigid body modes to change
in the system. These require relative rigid body positions of the bodies to be continuously updated. These
cannot be handled by the Support Method.
Rigid-Plastic Flow
The rigid-plastic flow analysis is an approach to large deformation analysis which can be used for metal forming
problems. Two formulations are available: an Eulerian (steady state) and Lagrangian (transient) approach. The effects
of elasticity are not included. If these effects are important, this option should not be used.
In the steady state approach, the velocity field (and stress field) is obtained as the solution of a steady-state flow
analysis. The time period is considered as 1.0 and, hence, the velocity is equal to the deformation. In the transient
formulation, the incremental displacement is calculated.
The R-P FLOW parameter invokes the rigid-plastic procedure. This procedure needs to enforce the incompressibility
condition, which is inherent to the strictly plastic type of material response being considered.
Incompressibility can be imposed in three ways:
1. by means of Lagrange multipliers. Such procedure requires Herrmann elements which have a pressure variable
as the Lagrange multiplier.
2. by means of penalty functions. This procedure uses regular solid elements, and adds penalty terms to any
volumetric strain rate that develops. It is highly recommended that the constant dilatation formulation be used
by entering a nonzero value in the second field of the GEOMETRY model definition option.
Penalty factor can be treated as constant or variable through the R-P FLOW parameter. The penalty value is
entered through the PARAMETERS option.

3. In plane stress analysis (shell and membrane elements), the incompressibility constraint is satisfied exactly by
updating the thickness. This capability is not available for steady state analysis.
In R-P flow analysis, several iterations are required at any given increment, the greatest number occurring in the first
increment. Subsequent increments require fewer iterations, since the initial iteration can make use of the solution from
the previous increment.
Due to the simplicity of the rigid-plastic formulation, it is possible to bypass stress recovery for all iterations but the
last in each increment, provided that displacement control is used. In such cases, considerable savings in execution
time are achieved. If nodal based friction is used in a contact analysis, then a stress recovery is always performed after
each iteration.
Steady State Analysis
The steady state R-P flow formulation is based on an Eulerian reference system. For problems in which a steady-state
solution is not appropriate, an alternative method is available to update the coordinates. The UPNOD user subroutine
is used to update the nodal coordinates at the end of a step according to the relationship.
(5-201)
where n refers to the step number, is the nodal velocity components, and t is an arbitrary time step. is selected
in such a way as to allow only a reasonable change in mesh shape while ensuring stability with each step.
Updating the mesh requires judicious selection of a time step. This requires some knowledge of the magnitude of the
nodal velocities that will be encountered. The time step should be selected such that the strain increment is never more
than one percent for any given increment.
The quantities under the title of ENGSTN in the printouts actually refer to the strain rate at an element integration
point. The reaction forces output by the program gives the limit loads on the structure.
Transient Analysis
In the transient procedure, there is an automatic updating of the mesh at the end of each increment. During the analysis,
the updated mesh can exhibit severe distortion and the solution might be unable to converge. Global adaptive meshing
or manual mesh rezoning can be used to overcome this difficulty.
The rigid-plastic flow capability is based on iteration for the velocity field in an incompressible, non-Newtonian fluid.
The normal flow condition for a nonzero strain rate can be expressed as:
(5-202)
where
(5-203)
x
i
n
x
i
n 1
v
n
t + =
v
n
t
i j
2
3
---
---
\ .
|
| |

i j

( )
i j
= =
2
3
---
i j
i j
=
185
CHAPTER 5
is the equivalent strain rate, is the yield stress (which may be rate-dependent) and
(5-204)
gives the deviatoric stress. The effective viscosity is evaluated as:
(5-205)
Note that as . A cutoff value of strain rate is used in the program to avoid this difficulty. An initial
value for is necessary to start the iterations. These values can be specified in the PARAMETERS option. The default
cut-off value is , and the default initial strain rate value is 10
-4
.
The value of the flow stress is dependent upon both the equivalent strain, the equivalent strain rate, and the
temperature. This dependence can be given through the WORK HARD, STRAIN RATE, TEMPERATURE EFFECTS,
or TABLE options, respectively. For steady state analysis, the UNEWTN user subroutine can be used to define a
viscosity. In this manner, a non-Newtonian flow analysis can be performed. For the transient procedure, the URPFLO
user subroutine can be used to define the flow stress.
Superplasticity
Superplasticity is the ability of a material to undergo extensive deformation such as strains of 1000% without necking.
Superplastic behavior has been reported in numerous metal, alloys and ceramics. Every instance of superplasticity is
associated with: (1) A fine grain size ( ), (2) deformation temperatures , and (3) a strain-rate
sensitivity factor .
Using finite element analysis to simulate superplastic fabrication of complex parts used in the aerospace and
automotive industries requires this material behavior and contact with friction. Furthermore, the process pressure
needs to be automatically adjusted to keep the material within a target strain rate. The simulation can be used to predict
thinning, forming time, areas of void formation, and can ultimately be employed in shape optimization; thus, reducing
the number of prototypes of forming trials required to product an acceptable part.
Three mechanisms have been proposed to account for the high strain-rate sensitivity found in superplastic materials:
(1) Vacancy creep, (2) creep by grain boundary diffusion, and (3) grain boundary sliding. According to Ghosh and
Hamilton, the strain-rate sensitivity of metals arises from the viscous nature of the deformation process. The viscosity
is a result of the resistance offered by internal obstacles within the material. In dislocation glide and climb processes,
the obstacles are a fine dispersion of second phase particles within the grain interior, between which the dislocations
are bent around and moved. At high homologous temperatures ( ; where is the melting temperature),
the high diffusivities around grain boundary regions can lead to grain boundary sliding. The overall rate sensitivity of
a material is then a result of the rate sensitivities of the grain boundary and the grain interior. The more the material
behaves as a viscous liquid, the greater its superplasticity.
i j

i j
1
3
---
i j
kk
=
2
3
---
--- =
0 ,
10
6
1 10 m 0.4T
m
>
m 0.3 >
T 0.4T
m
> T
m

The superplastic behavior is characterized by the dependence of the flow stress upon the strain-rate, which is usually
depicted by the logarithmic relationship shown in Figure 5-33.
As indicated in Figure 5-33, the stress-strain rate behavior of a superplastic material can be divided into three regions.
Values of strain-rate sensitivity, (the slope of flow stress versus strain-rate curve) which is a measure of resistance
to localized necking, are relatively low in both the low stress-low strain rate region I and the high stress-high strain
rate region III and superplasticity is not manifested. Rather, superplasticity is found only in region II, a transition
region in which stress increases rapidly with increasing strain-rate. As temperature increases and/or grain size
decreases, region II is displaced to higher strain-rates. Moreover, the maximum observed values of increase with
similar changes in these parameters.
Certainly the forming process innovations evoked will need to be carefully studied and developed. Forming times are
slow, and there will be a critical need for optimizing forming pressures, stress strain-rate and deflection in sheet
forming. Based on the schematic flow stress-strain rate relationships given above, it is apparent that high values of
are requisite for superplastic materials. Since, for a given material and forming temperature, , usually varies with
strain-rate, it is desirable to control strain-rate during forming so that optimum or at least adequate strain-rate
sensitivity is exhibited. Ductility is also dependent upon forming temperature, which must lie within a narrow range.
If forming temperatures and pressure cycle are optimum, then unlike conventional ductile materials, superplastic
materials are much less susceptible to localized necking. Additionally, under such conditions, the flow stress occurring
during forming is much lower than the mechanical yield stress.
Figure 5-33 Function of Strain-rate
m
m
m
m
Decreasing grain size or
increasing temperature
Region III
Region II
Region I
ln
ln
(a) Flow Stress
.
Decreasing grain size or
increasing temperature
Region III
Region II
Region I
(b) Strain-rate
m =
d ln
d
ln
-------------
\ .
| |
ln
.
Sensitivity
187
CHAPTER 5
Thus, the superplastic materials may be viewed as exhibiting time-dependent inelastic behavior with the yield stress
as a function of time, temperature, strain-rate, total stress and total strain. Typical materials used in commercial
superplastic forming applications include Ti-6A1-4V titanium alloy and 5083 aluminium alloy.
The form of constitutive equation used to simulate superplasticity is given as:
(5-206)
The form in Equation (5-206) can be recovered by using appropriate constants in the ISOTROPIC model definition
option to define power law or rate power law. Thus,
Power Law Model: (5-207)
Rate Power Law Model: (5-208)
The superplastic forming process requires the use of the SPFLOW parameter. The use of this parameter automatically
activates the FOLLOW FOR and PROCESSOR parameters. The process parameters are controlled by the use of the
SUPERPLASTIC history definition option.
Typical outputs that are available from the superplastic forming simulation are the thickness distribution for the part,
the equivalent plastic strain rate, and the history of the process pressure.
The process pressure is automatically calculated during the analysis. The pressure magnitude is adjusted such that the
equivalent strain rate in the part is at or close to the user-specified target strain rate. The equivalent strain rate in the
part is an average value calculated by sampling a suitable subset of elements. The recommended scheme is one in
which elements with a strain rate greater than a cut-off factor times the maximum element strain rate are sampled. This
maximum strain rate is based on a smoothing algorithm described below. The cut-off factor can vary between 0 (all
elements below the maximum are sampled) and 1 (only the elements with maximum strain rate are sampled). The
recommended value for the cut-off factor is 0.7 to 0.9 (default value is 0.8). To reduce undesirable oscillations in the
pressure-time history, a pressure smoothing algorithm is incorporated. The basis for this algorithm is a smoothing of
the maximum strain rate in the mesh based on the fact that the maximum strain rate should be typically representative
of a few elements in the mesh, rather than an isolated individual value. The peak strain rates in a few elements are
calculated. The number of elements that are used in this calculation varies with the cut-off factor (for a cut-off factor
of 1, only 1 element is used; for the default of 0.8, 10 elements are used; for a cut-off factor of 0, 50 elements are used).
The strain rate values are successively disregarded in descending order if the difference from the highest strain rate to
the lowest differs by more than 10 percent from the mean.
The value of the cut-off factor has significant influence on the maximum strain rate control and on the smoothness of
the pressure-time curve. Larger the factor (that is, 0.9 or higher) provides more control on the maximum strain rate,
but may potentially cause oscillations in the pressure history. Smaller the factor (that is, 0.7 or lower) provides less
control on the maximum strain rate, but causes smoother pressure-time curves. The default of 0.8 should work in most
cases - in situations where physically realistic localized strain rates occur and one desires good control on these
localized values, a higher value could be used.
y

m
=
y
A
o
+ ( )
m
B
n
+ =
y
A
m
n
B + =

Soil Analysis
This section has the solution procedure for fluid-soil analysis. In the current formulation, it is assumed that the fluid
is of a single phase, and only slightly compressible. This formulation will not be adequate if steam-fluid-solid analysis
is required. The dry soil can be modeled using one of the three models: linear elasticity, nonlinear elasticity and the
modified Cam-Clay model.
There are three types of soil analysis available in Marc. In the first type, you perform an analysis to calculate the fluid
pressure in a porous medium. In such analyses, heat transfer elements 41, 42, or 44 are used. The SOIL model
definition option is used to define the permeability of the solid and the bulk modulus and dynamic viscosity of the
fluid. The porosity is given either through the INITIAL POROSITY or the INITIAL VOID RATIO options and does not
change with time. The prescribed pressures can be defined using the FIXED PRESSURE option, while input mass flow
rates are given using either the POINT FLUX or DIST FLUXES option.
In the third type of soil analysis, a fully-coupled approach is used. Element types 32, 33, or 35 are available. These
elements are Herrmann elements, which are conventionally used for incompressible analysis. In this case, the extra
variable represents the fluid pressure. The SOIL option is now used to define both the soil and fluid properties. The
porosity is given through the INITIAL POROSITY or the INITIAL VOID RATIO option. The prescribed nodal loads and
mass flow rates are given through the POINT LOAD option, while distributed loads and distributed mass flow rates are
given through the DIST LOADS option. The FIXED DISP option is used to prescribe either nodal displacements or
pore pressures.
Technical Formulation
In soil mechanics, it is convenient to decompose the total stress into a pore pressure component and the deviatoric
or effective stress .
(5-209)
Note the sign convention used; a positive pore pressure results in a compressive stress. The momentum balance
(equilibrium) equations are with respect to the total stresses in the system.
(5-210)
where is the density, and , are the body force and the acceleration. The equilibrium equation can then be
expressed as
(5-211)
The fluid flow behavior can be modeled using Darcys law, which states that the fluids velocity, relative to the soils
skeleton, is proportional to the total pressure gradient.
(5-212)
p
d

d
pI =
f + u
=
f u
d
p f + u
=
u
f
K
------- p
f
g + ( ) =
189
CHAPTER 5
5
where
The fluid is assumed to be slightly compressible.
(5-213)
where is the bulk modulus of the fluid.
However, the compressibility is assumed small enough such that the following holds:
(5-214)
It is also assumed that the bulk modulus of the fluid is constant, introducing the fluids compressibility .
(5-215)
The solid grains are assumed to be incompressible. Under these assumptions, the governing equations for fluid flow is
(5-216)
where is the mediums porosity.
It is important to note that the mediums porosity is only dependent upon the original porosity and the total strains.
Letting and stand for the fluid and solids volume
(5-217)
where is the determinant of the deformation gradient and is the original porosity, both with respect to the
reference configuration.
is the fluids bulk velocity
is the soil permeability
is the fluid viscosity
is the fluid density
is the gravity vector.
u
f
K
f
g
p
K
f
f
----- =
K
f

f
K
---- p
f
g + ( )
f
---- p
f
g + ( )
f
1 K
f
=
---- p
f
g + ( )
f
p
+ =
V
f
V
s
dV
f
dV
f
dV
s
+ ( ) 1 J
1
1
0
( ) = =
J
0

Introducing the weighting function and , the weak form, which is the basis for the finite element system, then
becomes
(5-218)
where and are the conventional volumes and surface area and is the applied tractions. Note that the applied
tractions is the combined tractions from both the effective stress and the pore pressure .
and
(5-219)
where the normal volumetric inflow, , is:
(5-220)
The weak form of equilibrium can be written as:
(5-221)
(5-222)
Application of the directional derivative formula yields:
(5-223)
Hence,
(5-224)
Similarly,
(5-225)
u

p
u
f
u
u

u

u
p + [ ]dv
V
n 1 +
}

u
t dA
A
}
+ 0 =
V A t
d
p
p

K
---- p
f
g + ( )
f
p
p
+
p
u
( ) + dv
V
n 1 +
}

p
q
n
dA
A
n 1 +
}
0 =
q
n
q
n
k
--- p
f
g + ( ) n =
r
u

u
p [ ] v d
V
n 1 +
}
=
r
p

p
K
---- p
f
g + ( )
f
p
p

p
u
+ + v d
V
n 1 +
}
=
D
u
r
u
( ) u
d
d
------ r
u
u u + ( )
0 =
=
D
u
r
u
( ) u u
u
p ( )
u
T
up + [ ] v d
V
n 1 +
}
=
D
p
r
u
( ) p
u
p v d
V
n 1 +
}
=
191
CHAPTER 5
(5-226)
(5-227)
with the displacement and pressures interpolated independently as:
(5-228)
we get a linearized system of equilibrium equation,
In the second type of soil analysis, the pore (fluid) pressure is directly defined, and the structural analysis is performed.
Element types 27, 28, or 21 are available. In such analyses, the pore pressure is prescribed using the INITIAL PORE
and CHANGE PORE options. The characteristics of the soil material are defined using the SOIL option. If an elastic
model is used, the Youngs moduli and Poisons ratio are important. If the Cam-Clay model is used, the compression
ratios and the slope of the critical state line is important. For the Cam-Clay model, the preconsolidation pressure is
defined using the INITIAL PC option. For this model, it is also important to define an initial (compressive stress) to
ensure a stable model.
(5-229)
The resulting system of equations is highly nonlinear and nonsymmetric, and is solved by full Newton-Raphson
solution scheme.
Note that it is assumed that the permeability, porosity, viscosity, and the bulk modulus of the fluid are considered
independent of the state variables and . It is evident that, in general, this is not the case; however, in the analysis
that follows, these dependencies are ignored for tangent purposes. Note that they are included in the calculation of the
residuals and ; hence, convergence is always achieved at the true solution.
Three types of analyses can be performed. The simplest is a solution for only the fluid pressure based upon the porosity
of the soil. In this case, a simple Poisson type analysis is performed and element types 41, 42, or 44 are used. In the
second type of analysis, the pore pressures are explicitly defined and the structural analysis is performed. In this case,
the element types 21, 27, and 28 should be used. Finally, a fully-coupled analysis is performed; in which case, you
should use element types 32, 33, or 35. Of course, the soil can be combined with any other element types, material
properties to represent the structure, such as the pilings.
D
u
r
p
( ) u
p
K
---- p
f
g + ( )
f
p
p

p
u
+ + u
p
u
V
n 1 +
}
=
u
p
+ u
u
p
K
---- p
f
g + ( )
p
K
---- u ( )p }dv
D
p
r
p
( ) p
p
K
----p
f
p
p
+ v d
V
n 1 +
}
=
u N
i
u
i
=
and
p N
j
p
j
=
K
uu
K
up
K
pu
K
pp
u
p
)
`

F
R
u
R
p
)
`

=
u p
R
u
R
p

Mechanical Wear
Mechanical wear is an important physical phenomenon in any structure subjected to repeated loadings. Often, this
behavior is modeled by determining the stress on the surface and using it in a subsequent fatigue calculation. For
certain applications, including manufacturing, disk brakes, bearings, gears, tires, and seals, it is important to know the
amount of wear and possibly the change in geometry which would influence the behavior. It should be noted that for
some applications, such as polishing, surface wear is the positive objective. Wear, which is defined as the removal of
material from the surface, may be due to mechanical processes or chemical processes. The latter are not considered
but the UWEARINDEX user subroutine may always be used to incorporate these effects. The wear due to mechanical
behavior can be classified as:
1. Asperity deformation and removal
2. Plowing of the surface
3. Delamination
4. Adhesive
5. Abrasion
6. Fretting
7. Solid particle impingement
These processes lead to either mild wear or severe wear (rough, torn surfaces). The wear model implemented in Marc,
based upon Archard equation, is not applicable for severe wear.
There are several wear models available beginning with Archards equation:
where
This is converted into an incremental form. As the wear is calculated as a local quantity, we use normal stress instead
of the normal force (except for beam elements as discussed below). This leads to a series of models that are written as:

is the wear coefficient
is the normal force
is the sliding distance
is the hardness
Simple Archard
Bayer exponential form
Bayer exponential form with thermal activation
w KF
G
t
H
----- =
K
F
G
t
H
w
K
H
---- V
r el
=
w
K
H
----
m
V
r el
n
=
w
K
H
----
m
V
r el
n
exp
T
0
T
=
193
CHAPTER 5
where
A table may be associated with the coefficient to allow temperature dependence or other effects. The following
quantities can be used as independent variables for the wear coefficients: time, increment, current and original
coordinate position, normal stress and relative sliding velocity.
The incremental wear calculation is performed at the nodal points that are in contact. Hence, the CONTACT option is
required. Although friction is typically used in a wear analysis, it is not needed for the wear calculation itself since it
only uses the normal stress and the relative sliding velocity. The wear model and data is specified for each contact body
that should do wear calculation. The wear can occur on one or both sides of the bodies in contact. Each side uses the
wear model and properties of the respective body. A scaling factor can be used on the contact table allowing the wear
to be different for different contact combinations.
The incremental wear is calculated as:
and the wear is accumulated as:
.
The total accumulated wear can be interpreted as the distance a node would move due to wear and indicates a volume
loss. The total volume loss due to wear is calculated per contact body and printed to the output file. The total
accumulated wear is available for post processing as nodal post code 77, Wear Index. The current wear rate
is available as a vector in the normal direction of the contact point as nodal post code 78, Wear Rate.
For continuum elements, the nodal coordinates can be updated using
where is the normal to the surface and f is a scale factor. If large steps are taken such that the incremental wear
becomes large one might get inaccurate results due to nodes ending up outside the contact tolerance. New contact will
be found again, but small steps are often needed for accurate results. If the total amount of wear becomes large, then
remeshing may be required.
Wear calculation is supported for all continuum, shell, and beam elements. In addition, for thermo-mechanically
coupled analysis, wear calculation is supported on rigid (meshed) bodies with heat transfer. These also support
coordinate update due to wear. The coordinate update option is not supported for shell elements (since the shell
thickness should be changed, not the coordinate positions, but shell thickness update due to wear is not supported).
is the rate of change of wear in the direction normal to the surface
is the normal stress
is the relative sliding velocity
is the activation temperature
is the current temperature
w
V
r el
T
0
T
w
t
w
n 1 +
w
n
w
t + =
w
t ot
w
x
n 1 +
x
n
fw
t n + =
n

Beam elements can as mentioned be used in a wear calculation. This covers both the case that a beam element is
touching a solid or shell element and the case of beam-to-beam contact. Since we do not have a contact area available
in beam contact, we use the contact normal force instead of the contact normal stress in the wear calculation. This
needs to be taken into account when using wear on beam elements. The volume loss due to wear is calculated for each
element as an area obtained by the length of the element and the length offset by the amount of wear, and this area is
multiplied with twice the contact distance radius.
Design Sensitivity Analysis
5
Design sensitivity analysis is used to obtain the sensitivity of various aspects of a design model with respect to changes
in design parameters in order to facilitate structural modifications. The design parameters that are amenable to change
are called design variables. The two major aspects of the design model for which design sensitivity is considered
herein are:
a. Design objective
b. Design model response
As a result, the design sensitivity analysis capability in Marc is currently limited to finite element models of structures
with linear response in the computation of
1. Gradients for
a. An objective function (or the design objective), if one is defined by you (for example, minimizing the
material volume in the model).
b. Various types of design model responses under multiple cases of static mechanical loading, or free vibration,
in linear behavior.
2. Element contributions to the responses of the model.
The gradient of the objective function or of a response function is simply the set of derivatives of such a function with
respect to each of the design variables, at a given point in the design space (that is, for a given design).
For sensitivity analysis to proceed, the design model, the analysis requirements, the design variables, and the functions
for which the gradients are to be found have to be specified by you.
The existing design sensitivity analysis capability in Marc can be applied in one of two ways:
1. As a stand-alone feature, where you are concerned only with obtaining sensitivity analysis results. Such an
application is completed with the output of the sensitivity analysis results.
2. Within a design optimization process, where you are concerned mainly with the optimization of a design
objective related to a finite element model. This type of an application of sensitivity analysis is transparent to
you. The design optimization process is completed with the output of design optimization related data, such as
the optimized objective function, related values of the design variables, and the analysis results for the
optimized design.
195
CHAPTER 5
These two procedures are described below.
1. Sensitivity analysis as a stand-alone feature
This type of solution usually aims at obtaining the derivatives of user prescribed response quantities at a given
design, with respect to each of multiple design variables specified by the user. This set of derivatives is
therefore the gradient of the response function at the given design in the design variable space (or, in short, in
the design space). For example, for a prescribed response function r, given the design variables , , and
, the gradient is defined as
(5-230)
The number of response quantities for which gradients are computed is limited either by the program default
or by a user-specified number. If you are interested in obtaining the sensitivity analysis results in order of
criticality, the option to sort them in this order is also available. The responses are currently prescribed as
constraints with user-defined bounds. If sorting is not required, the bounds can be mostly arbitrary, although
they still have to conform to the type of constraint prescribed. However, if sorting is required and is to be
meaningful, it is important for you to give realistic bounds on the response.
Element contributions to each response quantity are obtained as a by-product of the type of response sensitivity
analysis capability in Marc. Thus, the response r can be represented as a sum of these element contributions:
(5-231)
where the second term, , involves work done elsewhere, such as in support settlement, if any.
This is helpful for a visual understanding of which regions of the model contribute in what manner to each of
the response quantities at the given design, since it can be plotted in a manner similar to, say stress contours.
Finally, as an option, if you also prescribe an objective function, the gradient of the objective function with
respect to the design variables is also computed at the given design. Thus, for the objective function , Marc
obtains
(5-232)
2. Sensitivity analysis within a design optimization process:
The design optimization algorithm in Marc requires the utilization of gradients of the objective function and of
the constraint functions, which are very closely related to the response functions. The current algorithm,
described under design optimization, ignores your initial prescribed design, but instead begins by generating
other designs within the prescribed bounds for the design variables. Once the optimization algorithm is
completed and the optimized design is available, if a sensitivity analysis is required at the optimized state, it
will be necessary for you to modify the model accordingly and to use sensitivity analysis as a stand alone
feature.
x
1
x
2
x
3
r
dr
dx
1
---------
dr
dx
2
---------
dr
dx
3
---------
T
=
r r
e
C +
e 1 =
N
=
C
W
W
dW
dx
1
---------
dW
x
2
---------
dW
dx
n
---------
T
=

During design optimization, sensitivity analysis is performed for a maximum number of constraints either
indicated by the program default or prescribed by you.
Theoretical Considerations
The method currently employed in Marc for response sensitivity analysis is the virtual load method. For sensitivity
analysis of the objective function, finite differencing on the design variables is performed directly.
In the virtual load method, first a design is analyzed for the user-prescribed load cases, and, if also prescribed, for
eigenfrequency response. The response of the structure having been evaluated for each of these analyses, the response
quantities for which sensitivity analysis is to be performed are then decided upon and collected in a database. In
sequence, a virtual load case is generated for each such response quantity. Reanalysis for a virtual load case leads to
virtual displacements.
The principle of virtual work is then invoked. This defines the element contributed part of the response quantity, for
which the virtual load was applied, as a dot product of the structural displacement vector, for the actual load case with
which the response is associated, and the virtual load vector itself.
The j
th
response can be expressed as the dot product of the actual load vector with the virtual displacement vector as
(5-233)
By differentiation of this expression, you can show that the derivative of the response with respect to a given design
variable is given by
where is the stiffness matrix of the structure. The response derivative above is evaluated on the element basis as:
(5-234)
where the vectors and are the vector of element nodal displacements due to the actual load case and the vector
of element nodal forces due to the virtual load, respectively. The case of eigenfrequencies follows the same logic
except that an explicit solution for the virtual load case is not necessary. The derivatives are now evaluated at the
element level via finite differencing. This is known as the semi-analytical approach.
Note that the derivative expression for the virtual load method is quite similar to that for the Adjoint Variable method.
In fact, while they are conceptually different approaches, for certain cases they reduce to exactly the same expressions.
However, for certain other cases, the terms take on different meanings although the end result is the same.
r
j
r
j
p
T
u
v
=
r
j
x
i
dr
j
dx
i
--------
dp
v
T
dx
i
---------- u u
v
T dp
dx
i
--------
dK
dx
i
-------- u
\ .
| |
+ =
K
dr
j
dx
i
--------
dp
v
T
dx
i
---------- u u
v
T dp
dx
i
--------
dK
dx
i
-------- u
\ .
| |
+
e
e
=
u p
v
197
CHAPTER 5
Design Optimization
Design optimization refers to the process of attempting to arrive at certain ideal design parameters, which, when used
within the model, satisfy prescribed conditions regarding the performance of the design and at the same time
minimize (or maximize) a measurable aspect of the design. In Marc, you can ask to minimize
1. total material volume
2. total material mass
3. total material cost.
When there is more than one material in the finite element model, the specification of different mass densities and unit
costs are taken into account in the computation of the objective function.
The performance requirements might not necessarily have to be related to response, but also to different concepts such
as packaging, design envelope, even maintenance. The current capability is based on optimization with constraints on
response. Also, the lower and upper bounds on the design variables themselves define the limits of design
modifications. Hence, the design optimization problem can be posed mathematically in the following format:
where is the objective function, and is the jth constraint function, either specified as an inequality related to a
response quantity or as a limit on a design variable. For example, to limit the x-direction translation component at a
node k, the constraint can first be written as
Assuming that the displacement is constrained for positive values, the normalized constraint expression (dropping
the subscript ) becomes:
(5-235)
with its derivatives as:
(5-236)
Within Marc, the constraints can be imposed on strain, stress, displacement, and eigenfrequency response quantities.
For stresses and strains, the constraints are defined as being on the elements, and for displacements, the constraints are
defined as being at nodes. Stress and strain components, as well as various functions of these components (the von
Mises equivalent stress and principal stresses, stresses on prescribed planes) and generalized stress quantities can be
constrained. Similarly, translation and rotation components of displacement, resultant and directed displacements as
well as relative displacements between nodes can be constrained. For free vibration response, constraints can be placed
on the modal frequencies as well as on differences between modal frequencies. A full listing of such constraints are
given in Marc Volume C: Program Input.
Minimize
Subject to
W
c
j
0.0 j 1 m , , =
W c
j
u
x
( )
k
u
xk
*
c
j
u
u ( ) u
0.0
dc
dx
------
1
u
------
du
dx
------ =

The upper and lower bounds on the design variables are posed as
(5-237)
after which they can be transformed into expressions similar to Equation (5-235).
The response quantities associated with the model are implicit functions of the design variables. Analyses at sample
design points are used to build explicit approximations to the actual functions over the design space. This approach
minimizes the number of full scale analyses for problems which require long analysis times such as for nonlinear
behavior. This method is summarized next.
Approximation of Response Functions Over the Design Space
The design space for the optimization problem is bounded by limits on the design variables of a model to be optimized.
The simplest case is that of a single design variable, where the design space is a straight line, bounded at the two ends.
For higher number of design variables, say n, the design space can be visualized as a bounded hyperprism with
vertices. For such a construct, you can build approximations to the constraint functions by way of analyses conducted
at the vertices. However, this requires analyses.
We now note that the minimum geometrical construct spanning n-dimensional space is a simplex with vertices.
Thus, an approximation based on analyses at vertices requires only analyses. The simplex has already been
used for first order response surface fitting based on only function values [Ref. 3]. However, the use here involves
higher order response functions.
Like the hyperprism, in one-dimension, the simplex degenerates into a straight line. However, in two dimensions it is
a triangle, and in general it is a hyper-tetrahedron. Figure 5-34 compares the simplex to the hyperprism in normalized
two-dimensional design space.
Figure 5-34 Comparison of the Simplex to the Hyperprism in Two-dimensions
While the orientation of the simplex in the design space appears to be a relatively arbitrary matter, once an origin and
the size of the simplex are prescribed, a simple formula will locate all vertices of a simplex in n-dimensions [Ref. 4].
The response gradient information at the simplex vertices is combined with the function values to achieve enhanced
x
i
x
u
and x
i
x
l

n
2
n
2
n 1 +
n 1 + ( )
n = 2
Hyperprism: 22 = 4 vertices
Simplex : 2 + 1 = 3 vertices
199
CHAPTER 5
accuracy. Thus, an analysis at a vertex can be utilized to yield both response function values and, by way of sensitivity
analysis, the response gradients at that vertex. The virtual load method employed in Marc for obtaining the response
gradients is discussed under Design Sensitivity Analysis. The response gradients can then easily be converted to
constraint gradients for use in an optimization algorithm.
As a result, the results of an analysis at a vertex of the simplex can be summarized as the vector of constraint function
values , and the gradient, , of each constraint function ( ) at that vertex.
For the case of a one variable problem, the results of analyses at the two vertices are symbolized in Figure 5-35, for a
hypothetical constraint .
Figure 5-35 Vertex Results for One-dimensional Design Space
From Figure 5-35, it appears almost natural to fit a cubic function to the four end conditions (two function values and
two slopes) depicted in the figure. However, this approach is too rigid, and is not easily generalizable to higher
dimensions. On the other hand, the use of two quadratics, which are then merged in a weighted manner gives higher
flexibility and potential for increased accuracy.
It can be seen that the two equations, and , are designed such that they both satisfy the function values at the
two vertices, but satisfies the slope at vertex 1 only, and satisfies the slope at vertex 2 only. Finally, at any
design point , the response function can be approximated as
(5-238)
where the weight functions and are normally functions of and possibly some other parameters.
This type of approach has the further advantage that it is consistent with the use of the simplex for determining analysis
points and approximating constraint functions in the higher dimensional cases. Therefore, for an n-dimensional
problem, the simplex having vertices, each equation needs to have unknown
parameters. The general quadratic polynomial without the cross-coupling terms satisfies this condition for all .
c
j
c
j
j
c
j
c
j 1
dc
j 1
dx
------------
c
j 2
dc
j 2
dx
------------
x
Possible Actual Function
1
2
E2
E1
E1 E2
E1 E2
x c
j
c
j
W1 E1 W2 E1 + ( ) W1 W2 + ( ) =
W1 W2 x
n 1 + n 1 + ( ) n + 2n 1 + =
n

Improvement of the Approximation
When approximations are used, the results of an optimization process need to be checked by means of a detailed
analysis. As a result, the approximations can be adjusted and the optimization algorithm can be reapplied. Depending
on how accurate the approximations prove to be and how many more detailed analyses are acceptable to you, this
process can be applied for a number of cycles in order to improve upon the results.
The Optimization Algorithm
The optimization algorithm implemented in Marc is the Sequential Quadratic Programming method [Ref. 5]. This
method is employed using the approximation described above. By obtaining response function and gradient values
from the approximate equations, the need for full scale analyses is reduced. The method is summarized below.
The quadratic programming technique is based on the approximation of the objective function by a quadratic function.
When nonlinear constraints exist, as is the case in most practical design optimization problems, the second order
approximation concept is extended to the Lagrangian which is a linear combination of the objective function and the
constraint functions. The solution method for a quadratic programming problem with nonlinear constraints can be
summarized as the following:
At each step, modify the design variables vector by adding a scaled vector
(5-239)
where is a search direction and is the scale factor along the search direction. If the search direction has been
determined, the scale factor can be found by some type of line search along the search direction. The determination of
the search direction constitutes the major undertaking in the quadratic programming method. If is the Hessian of
the Lagrangian and is the gradient of the objective function, then the search direction is that which minimizes the
function
(5-240)
subject to the linearized constraints
(5-241)
where is the Jacobian matrix of the constraints and is the vector of constraint functions at the current iteration.
Due to lack of knowledge about which constraints will be active at the optimum, the Hessian of the Lagrangian is not
always readily available. Thus, some iterations take the form of a gradient projection step.
The coefficients of the constraint functions in the Lagrangian are the Lagrange multipliers which are unknown before
solution has started. At the optimum these multipliers are zero for inactive constraints. Normally, the above problem
is solved using only those constraints which appear to be active at or close to the current design point, with the
assumption that these constraints will be active at the optimum. The selection of these active constraints is done within
the framework of an active set strategy, the set being modified appropriately with the progression of the iterations.
Similarly, the arrays , , and are also modified as the iterations proceed.
x
x x qs +
s q
H
g s
g
T
s s
T
Hs ( ) 2 +
Js c
J c
g H J
201
CHAPTER 5
Marc User Interface for Sensitivity Analysis and Optimization
This is discussed under Input, Output, and Postprocessing in the following paragraphs.
Input
Input features related to design sensitivity and design optimization are similar. However, they differ in the parameter
data blocks and in the optional specification of an objective function for the case of sensitivity analysis. Therefore,
other than these two items, all discussion of input should be construed to refer to both procedures equally. All design
sensitivity and design optimization related parameters and options in a Marc input file start with the word DESIGN.
All load cases and any eigenfrequency analysis request associated with sensitivity analysis or optimization should be
input as load increments in the history definition section after the END OPTION line. The maximum number of nodes
with point loads or the maximum number of distributed load cases should be defined in the DIST LOADS parameter.
The type of solution requested from Marc can simply be indicated by the parameter input DESIGN SENSITIVITY or
DESIGN OPTIMIZATION. Only one of these lines can appear in the input. These parameters also let you change the
sorting of constraints, the maximum number of constraints in the active set, and the maximum number of cycles of
design optimization. The first two items are discussed in previous sections. A cycle of design optimization refers to a
complete solution employing the sequential quadratic programming technique followed by a detailed finite element
analysis to accurately evaluate the new design point reached by way of approximate solution. If required, the
approximation is improved using the results of the last finite element analysis, and a new cycle is started.
The specification of an objective function, being optional for design sensitivity, is made by use of the model definition
option DESIGN OBJECTIVE. This allows you to choose from one of the available design objectives.
The DESIGN VARIABLES option allows you to specify the design variables associated with the finite element model.
You have a choice of three major types of design variables:
1. Element cross-sectional properties as given by way of the GEOMETRY option.
2. Material properties as given by ISOTROPIC or ORTHOTROPIC options.
3. Composite element properties of layer thickness and ply angle, as given by the COMPOSITE option.
The properties which are supported are listed under the DESIGN VARIABLES option. The relevant elements, for which
cross-sectional properties can be specified as design variables, each has a section in Marc Volume B: Element Library,
describing which properties, if any, can be posed as design variables for that element.
Similarly, the list of material properties (currently all related to linear behavior) that can be design variables is given
under the DESIGN VARIABLES option. For composite groups, the layer thicknesses and ply angles can be given as
design variables for composite groups.
Design variables can be linked across finite element entities such that a given design variable controls several entities.
An example is the linking of the thicknesses of several plane stress elements by means of a single design variable.
Thus, when this variable changes, the thicknesses of all linked elements reflect this change. On the other hand, for the
unlinked case, all thicknesses are associated with separate design variables. This feature is controlled by the LINKED
and UNLINKED commands.
Response quantities, or constraints on response quantities, are specified by means of the DESIGN DISPLACEMENT
CONSTRAINTS, DESIGN STRAIN CONSTRAINTS, DESIGN STRESS CONSTRAINTS, and DESIGN FREQUENCY
CONSTRAINTS options. There is no limit to the number of constraints. Displacement constraints are posed at nodes

or groups of nodes, while the strain and stress constraints are posed over elements or groups of elements, and
frequency constraints are posed for free vibration modes. A complete list can be found under the above mentioned
options in Marc Volume C: Program Input.
When strain or stress constraints are prescribed, it is useful to know that the program evaluates such constraints at all
integration points of all layers of an element and proceeds to consider the most critical integration point at the related
layer for the element. Thus, a strain or stress constraint on an element normally refers to the most critical value the
constraint can attain within the element. During optimization, the most critical location within the element may change
and any necessary adjustment takes place internally.
For certain responses, the limiting values can be the same in absolute value for both the positive and negative values
of the response. For constraints on such response functions, you have the choice of prescribing either separate
constraints on the positive and negative values, or a combined constraint on the absolute value. The first approach is
more accurate albeit at a higher computational cost.
Output
For sensitivity analysis, the output file contains the following information:
Echo of input, any warnings or error messages.
A of numbers and definitions for your prescribed design variables. It is important to note that the variable
values are always output in the internal numbering sequence which is defined in this. Search for the words
Design variable definitions to reach this in the output file.
Analysis results for your prescribed design.
The value of the objective function and its gradient with respect to the design variables prescribed by you.
Sensitivity analysis results for the response functions in the default or user-defined set, sorted in order of
criticalness (if specified). It should be noted that although the responses are sorted across multiple load cases,
the sensitivity results are output for each load case. The related output file consists of a check on the actual
response value (obtained by sensitivity analysis) and the gradient of the response with respect to the design
variables. For eigenfrequency results, the check values on the response can be somewhat more accurate than
the results from eigenfrequency analysis since the latter is iterative, but the check uses the Rayleigh quotient on
top of the iteration results. A constraint reference number allows you to track the sensitivity analysis results
plots when postprocessing. Search for the word Sensitivity to reach this output.
For design optimization, the following is written into the output file:
Echo of input, any warnings or error messages.
Certain indications that some analyses are being done, but no analysis results except for the best design
reached during optimization.
The objective function values and the vector of design variable values:
At each of the simplex vertices.
For the starting simplex vertex (from this point on, the information also includes whether the design is
feasible or not).
At the end of each cycle of optimization.
For the best design found.
Analysis results for the best design.
203
CHAPTER 5
Postprocessing
This requires that you ask the program to create a post file. The following plots can be obtained by way of
postprocessing.
For sensitivity analysis:
Bar charts for gradients of response quantities with respect to the design variables.
Contour plots of element contributions to response quantities. Thus, a finite element model contour plot gives
the element contributions to a specific response quantity which was posed in the form of a constraint in
the input file.
The increment number of the sensitivity analysis results is the highest increment number available in the post file. The
information for each response quantity is written out as for a subincrement. The zeroth subincrement corresponds to
the objective function information. Numbers for the other subincrements correspond to the constraint reference
number(s) in the output file.
For design optimization:
Path plots showing the variation of the objective function and of the design variables over the history of the
optimization cycles. The best design (feasible or not) is not necessarily the last design point in the plot. The
values at the starting vertex are considered as belonging to the zeroth subincrement. Each optimization cycle is
then another subincrement. The increment number corresponding to these subincrements is taken as zero. The
analysis results for the best design start from increment one.
Bar charts where each chart gives the values of the design variables at the optimization cycle corresponding to
that bar chart.
Defined Initial State with Result Data from Previous Analysis
(including AXITO3D)
In many cases, it is necessary to analyze a nonlinear process in several stages. Each stage may involve different contact
bodies and boundary conditions, but history data such as temperature, displacement, stress and strain have to be carried
over for contact bodies to be passed from one stage to another. In Marc, it is possible to start a numerical simulation
as a two-dimensional axisymmetric or plane strain problem and switch to a full three-dimensional analysis in
subsequent stages. This makes multiple stage analysis more efficient. The PRE STATE option is designed to read
result data from Marc and use the data as initial conditions in the new analysis. For contact analysis, it allows the
selection of contact bodies by names. The PRE STATE option includes the capability of the AXITO3D option available
in the 2003 release.
The methodology of the PRE STATE option can be outlined as the following:
Scan through the result file for the desired time step or time instant.
Collect the list of nodes and elements corresponding to the contact body transfer list.
Collect node and element data as required (see the PRE STATE option in Marc Volume C: Program Input for
the data type).
Copy data to the new model as the initial conditions with 2-D to 3-D expansion if applicable.

The PRE STATE option takes several steps:
Run first stage analysis.
Open the result file.
Extract and expand (if 2-D to 3-D is required) the mesh from the previous result file at the increment. The
expansions of boundary conditions can also be done through Mentat.
Build second stage model based on the mesh extracted and add other contact bodies and conditions.
Specify the PRE STATE option under initial condition menu.
Run 3-D analysis.
Most quadrilateral or hexahedral elements (including Herrmann elements) as well as commonly most available
materials (such as metal and rubber) can be used in the PRE STATE option. For large deformation problems, either
total or updated Lagrangian formulation can be used. Thermal and dynamic effects can also be included.
For more detailed information on this feature, see the PRE STATE model definition option, in Marc Volume C:
Program Input.
Load and Displacement Boundary Conditions Transfer for
PRE STATE Option
In general, the PRE STATE option can only transfer history data from the result file as the initial conditions in the new
model. Users are responsible for reapplying the boundary and load conditions as well as the material properties. If
users use Mentat, the boundary conditions defined in the previous 2-D analysis automatically expand to 3-D when the
mesh is expanded from 2-D to 3-D. However, if a load function is used and is required in the new model, curve shifting
is required. That is, if, in the previous model, a load is applied from 0 to 10 in the new model based on the last step of
the previous analysis, the load is applied starting from 10. For prescribed displacement, as nodal positions are updated
to the final configuration in the previous analysis, the load function should start from 0 in the new model.
Compatible element types for PRE STATE
It is important to select the correct element types compatible with the previous model for a 2-D to 3-D transfer. For
the same element type transfer, not all the element types are supported.
For axisymmetric to 3-D transfer, supported element types are:
10 -> 7 20 -> 7
82 -> 84 83 -> 84
28 -> 21 67 -> 21
33 -> 35 66 -> 35
55 -> 57 59 -> 61
48 -> 23 142 -> 23
144 -> 146 145 -> 146
166 -> 147 147 -> 147
169 -> 148 170 -> 148
40 -> 43 172 -> 174
205
CHAPTER 5
For 2-D plane strain to 3-D solid elements, supported elements types are:
For 2-D to 2-D and 3-D to 3-D data transfer, the PRE STATE option supports most standard Marc element types for
plane stress, plane strain, axisymmetric, 3-D and (solid) shell applications. The supported structural element types are:
The supported heat transfer element types are:
Steady State Rolling Analysis
Rolling contact analysis of a cylindrical deformable body in a Lagrangian framework can be computationally
expensive because it may require not only time-dependent transient process but also a fine mesh in the entire body to
accurately capture contact characteristics. However, some problems involve only fully axisymmetric structures with
constant moving/spinning velocities. These problems can be considered steady state if a reference configuration,
which moves with the body but does not spin around the rolling axis, is used.
Marc provides the capability of steady state rolling analysis. The feature is characterized by a mixed
Eulerian/Lagrangian formulation with inertia effects in spinning/cornering deformable bodies. Using a non-spinning
reference frame attached to the wheel axel, the analysis becomes purely space dependent. It presents a better
alternative to the unnecessary computational burden of arriving at a steady state condition through a transient analysis.
Furthermore, a finer mesh only needs to be used in the contact region as opposed to the entire rolling surface.
11 -> 7 80 -> 84
19 -> 7 81 -> 84
27 -> 21 32 -> 35
29 -> 21 34 -> 35
39 -> 49
plane stress: 3, 26, 53, 124, 201
plane strain: 6, 11, 19, 27, 29, 32, 34, 54, 58, 60, 80, 81, 125, 128, 155
axisymmetric: 2, 10, 20, 28, 33, 55, 59, 66, 67, 82, 83, 119, 126, 129, 156
3-D solid: 7, 21, 35, 57, 61, 84, 127, 130, 134, 136, 157, 202
membrane: 18, 30, 158, 200
(solid) shell: 22, 49, 72, 75, 138, 139, 140, 185
rebar: 23, 46, 48, 142-148, 165-170
planar: 37, 39, 41, 69, 121, 131
axisymmetric: 38, 40, 42, 70, 122, 132
3-D solid: 43, 44, 71, 123, 133, 135, 137, 203

Kinematics
We consider the axisymmetric body shown in Figure 5-36. The body spins at an angular velocity around the
axisymmetric axis at point and, simultaneously rotates with a cornering angular velocity around an axis
at point . Assume that a particle in the body has a location at time . At time , its motion contains
three parts:
1. from to a location because of the spinning
2. from to because of a deformation , where is time independent function resulting from
the steady state condition
3. from to because of the cornering.
Figure 5-36 Kinematics
The three motions can be described as
(5-242)
(5-243)
(5-244)
with
and (5-245)
In Equation (5-245), and are the skew-symmetric tensors associated with the rotation vectors and ,
respectively, with
and (5-246)
for any vector .
Time derivative of Equation (5-245) gives
and (5-247)
s
T
s
P
s

c
T
c
P
c
P
0
t 0 = t
P
0
X
X Y Y D X ( ) = D
Y Z
P
s
T
s

s
T
c
P
c
c
X R
s
P
o
P
s
( ) P
s
+ =
Y D X ( ) =
Z R
c
Y P
c
( ) P
c
+ =
R
s

s
t ( ) exp = R
c

c
t ( ) exp =
s

c

s
T
s

c
T
c
s
r
s
T
c
r =
c
r
c
T
c
r =
r
R
s

s
R
s
= R
c

c
R
c
=
207
CHAPTER 5
Making use of Equations (5-242), (5-243), (5-246), and (5-247), the velocity of the particle can be obtained by the first
time derivative of Equation (5-244) as
(5-248)
where , is the spinning angle.
Similarly, the acceleration is obtained by the second derivative of Equation (5-244) with respect to time:
(5-249)
where denotes the tensor product and is the unit tensor.
Transformation of Equations (5-248) and (5-249) into the reference configuration defined by by premultiplying
, gives
(5-250)
(5-251)
where and are velocity and acceleration of the particle with respect to the reference frame.
Inertia Effect
The contribution of inertia effect into the right-hand-side of the system equation can be calculated using the weak form
(5-252)
where is the density, is the volume, and is the displacement.
Linearization of Equation (5-252) gives
(5-253)
which can be used to calculate the contribution of the inertia effect to stiffness matrix.
Z
R
c

c
T
c
Y P
c
( )
s
D
------- + =

s
t =
Z
R
c

c
2
T
c
T
c
1 ( ) Y P
c
( ) 2
s
c
T
c
D
-------
s
2
2
D
2
---------- + + =
1
X
R
c
T
v
c
T
c
Y P
c
( )
s
D
------- + =
a
c
2
T
c
T
c
1 ( ) Y P
c
( ) 2
s
c
T
c
D
-------
s
2
+

2
D
2
---------- + =
v a
a u v d
v
}
=
c
2
T
c
T
c
1 ( ) Y P
c
( ) u v 2
s
c
T
c
D
------- u v d
V
}
d
V
}
=
s
2
D
-------
u
--------- v d
V
}
+
v u

c
2
u T
c
T
c
1 ( ) u v 2
s
c
T
c
u
---------- u v d
V
}
d
v
}
=
s
2
u
----------
u
--------- v d
v
}
+

Rolling Contact
To take into account the rolling effect in contact, the velocity vector in Equation (5-250) is decomposed into a normal
and a tangential component, with respect to the contact surface, for all nodes in contact with ground. The normal
component is then forced to become zero because of the contact conditions. The relative slipping velocity used in
friction calculation is the difference between the tangential component and the ground moving velocity.
Steady State Rolling with Marc
The capability of steady state rolling analysis in Marc has taken into account the effects of rolling frictions and the
inertia effects resulting from both spinning and cornering. The deformable rolling body can contact with multiple, flat
or nonflat rigid surfaces. A typical example is a tire model which is in contact with a rigid rim and a rigid road surface.
The spinning effects are involved only for selected contact body pair (for example, tire/road) which is defined with the
SS-ROLLING history model definition option. The steady state rolling analysis can follow a static stress analysis with
various loadcases and can also combine with any steady state loads. The feature is available for 3-D analysis only.
The element types supported for steady state rolling analysis include 7, 9, 18, 21, 35, 57, 61, 84, 117, 120, 146, 147,
and 148.
The friction type supported in contact for steady state rolling analysis, based on nodal force, is Coulomb friction for
rolling.
Because the system matrix becomes nonsymmetric in steady state rolling analysis, it is recommended that the
multifrontal direct sparse solver be used.
The feature requires the so-called streamline along the circumferential direction of the spinning body. Therefore, the
3-D mesh must be generated by revolving its corresponding axisymmetric mesh. For a 3-D brick element, the
circumferential direction must be from the element face defined by nodes 1-2-3-4 to the element face defined by nodes
5-6-7-8.
Below is a list of options required in a steady state analysis, in addition to the others in a standard static stress analysis.
Structural Zooming Analysis
Local variations, such as the changes in model geometries or in the degrees of finite element refinement to achieve a
better evaluation of the local gradients in the solution, often need a complete re-analysis of the entire model. However,
in cases that these local changes have negligible influence on the solution a certain distance away from the changes,
it is computationally more efficient to model only the part with the local changes. It can be realized by applying the
existing loads or/and boundary conditions in the local model along with properly defined kinematic conditions to the
local boundaries connecting to the global model.
SS-ROLLING: Parameter activating the steady state rolling analysis
ROTATION A: Model definition option defining the spinning axis
CORNERING AXIS: Model definition option defining the cornering axis
SS-ROLLING: History definition option defining the spinning body, ground body, spinning body motion
(spinning/cornering/moving relative to the ground) and others. All velocities/forces defined
in the option are total values.
209
CHAPTER 5
A typical structural zooming analysis contains two steps:
1. Global run to obtain a post file containing global results.
2. Local run to define kinematic boundary conditions in the local model and to obtain refined results in the local
model.
This procedure can be repeated as many times as desired. Any local analysis can be the global analysis of next level
refinement.
The GLOBALLOCAL option (see Marc Volume C: Program Input for details) is used in the input of the local run to
define the list of nodes connecting to the global model. Marc calculates the deformation (temperature) history of these
nodes, based on their locations in the global model and on the solution of the global analysis. The obtained deformation
(temperature) history is then applied to the nodes as prescribed kinematic boundary conditions. The detailed steps
include:
a. Reading in the GLOBALLOCAL option to get the list of connecting nodes.
b. Reading in the global model (element types, node coordinates, element connectivity, thickness if shell
elements) and solution of the model from the generated post file in global run.
c. Finding out the locations of local connecting nodes in the global model (the element each node is associated
with and its isoparametric location within the element).
d. Calculation of the deformation (temperature) of each connecting node for every increment available in the
global post file, based on its location in the global elements associated using the interpolation techniques,
and storage of the deformation (temperature) history in the format of time-dependent tables.
e. Applying the deformation (temperature) history of the connecting nodes to the local model as prescribed
kinematic boundary conditions.
All the steps above are performed automatically inside Marc once the GLOBALLOCAL option is used in the input file
of the local run.
If the global adaptive meshing is used in the first analysis, then the GLK model definition should also be used.
Element Types Supported
The global to local modeling can be used in the following four cases:
One can use composite continuum elements in the local model if the global model had composite continuum elements
with the same number of layers or was composed of shell elements
See Figure 5-37 for examples from 2-D solid elements to 2-D solid elements and from 3-D shell elements to 3-D solid
elements.
Global Model Local Model
2-D Solid 2-D Solid
3-D Solid 3-D Solid
3-D Shell/Membrane 3-D Shell/Membrane
3-D Shell/Membrane 3-D Solid

Figure 5-37 Structural Zooming
Uncoupled Thermal Stress Analysis
A typical chaining analysis in Marc for thermal stress problems includes two steps:
Heat transfer analysis to generate a post file containing temperature distribution on the model and history over
the time period considered.
Thermal stress analysis based on the temperature distribution/history obtained from the previous heat transfer
analysis
To perform a chaining analysis for a thermal stress problem two options can be used in Marc. They are the CHANGE
STATE option and the MAP TEMP option.
Using the CHANGE STATE option, the mesh in stress analysis has to be the same as that used in the heat transfer
analysis. The structural element type has to correspond with the heat transfer element type. Please see Marc Volume
B: Element Library for this correspondence. The basic idea of the option is to read in the integration point temperatures
for each element from the post file of the heat transfer analysis and use them directly in the stress analysis. No spatial
data mapping is required. Since this option is integration point based, INITIAL STATE should be used if one needs to
specify initial temperatures in the stress analysis. This option can be used with AUTO THERM, AUTO THERM CREEP,
and AUTO STEP.
The data mapping used with the MAP TEMP option contains the following steps:
1. Read in the MAP TEMP option to get the list of structural nodes. The temperature history of the nodes is
expected to be obtained from the heat transfer post file.
2. Read in heat transfer model (element types, node coordinates, element connectivity, etc.), and solution of the
model in the heat transfer analysis.
3. Find the location of the structural nodes in the heat transfer model (the element each node associated and its
isoparametric location within the element).
global model
local model
cracks
Cylinder
(a) 2-D Solid to 2-D Solid (b) 3-D Shell to 3-D Solid
211
CHAPTER 5
5
4. Calculate the temperature of each structural node for every increment available in the post file of the heat
transfer analysis, based on its location in the heat transfer element associated. Store the temperature history in
the format of a time-dependent table.
5. Apply the temperature history to the structural node in stress analysis.
Since this option is node-based, INITIAL TEMP should be used if one needs to specify initial temperatures in the stress
analysis. This option cannot be used with AUTO THERM and AUTO THERM CREEP. It can be used with AUTO STEP.
Compared to the CHANGE STATE option, the MAP TEMP option provides more flexibility. It allows a temperature
distribution/history obtained from a previously heat transfer simulation to be used in the stress analysis even though
the meshes are dissimilar. The element types or even element classes used in heat transfer and stress analyses can be
different. The model or the time period covered in heat transfer and stress analyses can also be different. For time
period outside the range in heat transfer analysis, a time shift or end values or extrapolation option can be used. For
the part of the structural model or time period not covered by the post file of the heat transfer analysis, POINT TEMP
can always be used to directly input nodal temperatures.
Cure-Thermal-Mechanically Coupled Analysis
When resin is used in the manufacturing of composite parts, the curing process during and after the forming processes
has a significant effect on the thermal and mechanical behavior of the formed part. The shrinkage caused by curing
may severely distort the final geometry of the part. Marc has implemented a capability that allows you to perform cure-
thermal-mechanical coupled analysis to predict this behavior.
The curing analysis is incorporated into the existing staggered coupled thermal-mechanical analysis procedure (see
Thermal Mechanically Coupled Analysis in Chapter 6: Nonstructural and Coupled Procedure Library). Before the heat
transfer analysis pass takes place, the curing analysis is performed based on the estimated temperatures at the
beginning and end of the increment. The cure rate is then calculated according to the cure kinetics of the resin
materials. Using the cure rate, a heat flux due to cure reaction heat generation is calculated and added into the heat
transfer system of equations. Volumetric curing reaction heat flux is calculated according to the cure rate
(5-254)
where:
The governing matrix equation for cure-thermal coupled analysis can be expressed as:
(5-255)
resin cure reaction heat
resin density
resin degree of cure
fiber volume fraction
Q
c
d
dt
------- 1 V
f
( )
r
H
r
=
H
r
V
f
C T ( )T
K T ( )T + Q Q
I
Q
F
Q
C
+ + + =

In Equation (5-255), and are the temperature-dependent heat capacity and thermal conductivity
matrices, respectively. is the nodal temperature vector. is the time derivative of the temperature vector, is the
external heat flux vector, is the heat flux vector due to plastic work, represents the heat generated due to
friction, and is the heat generated due to curing.
With the CURING parameter included in the input data, Marc first activates the pass to calculate the degree of cure and
curing reaction heat flux at the beginning of each cycle of heat transfer analysis. Then, Marc considers the cure
shrinkage strain in the mechanical pass of the thermal-mechanical coupled analysis.
To include curing into a heat transfer or thermal-mechanical coupled analysis, it is necessary to add the CURING
parameter into the input data. In addition, the CURE RATE model definition option is also needed to define the curing
properties, usually called cure kinetics, of resin materials as one kind of property for heat transfer analysis.
To include cure shrinkage strain into the mechanical pass of a thermal-mechanical coupled analysis, in addition to
having both the CURING parameter and the CURE RATE model definition option in the input data, the CURE
SHRINKAGE model definition must be included to define the cure shrinkage properties of the resin material. Note that
the cure shrinkage effect is calculated based on the degree of cure, so the cure shrinkage effect is ignored if the degree
of cure is not calculated during the heat transfer pass.
Cure Kinetics
In Marc, the cure kinetics defines the cure rate as a function of the degree of cure and the temperature of resin
materials. There are different approaches to define the cure kinetics of resin materials: embedded cure kinetics models,
table definition and user subroutine.
In a curing analysis, the cure rate is calculated for each time step. Assuming that the cure rate is defined as the function
of the degree of cure and temperature of:
(5-256)
The time integration of the degree of cure is done using backward Euler method:
(5-257)
where
iteration number for cure
current increment number
previous increment number
function defined by cure kinetics model
time step size of the increment
C T ( ) K T ( )
T T
Q
Q
I
Q
F
Q
C
d
dt
------- f T , ( ) =
n
i
t f
n
i 1
T , ( )
n 1
+ =
i
n
n 1
f
t
213
CHAPTER 5
The four cure kinetics models implemented in Marc are:
Cure kinetics model 1: by Lee, Loos and Springer (1982) [Ref. 42];
Cure kinetics model 2: by Scott(1991) [Ref. 49];
Cure kinetics model 3: by Lee, Chiu, and Lin (1992) [Ref. 43];
Cure kinetics model 4: by Johnston and Hubert (1996) [Ref. 51].
Tables 5-7 and 5-8 summarize the details of these cure kinetics models and the definition of the required parameters.
Table 5-7 Cure Kinetics Models Embedded in Marc
Model Equations Parameters
Model 1
Lee, Loos and Springer
(1982) [Ref. 42]
(Included in Model 2)
Lee, Chiu, and Lin
(1992)[Ref. 43]; White and
Hahn (1992) [Ref. 47]
(Included in Model 2)
Kenny (1992) [Ref. 48];
Scott (1991)[Ref. 49]
Model 2 (combined Model)
Scott (1991)[Ref. 49]
Model 3
Lee, Chiu, and Lin
(1992)[Ref. 43]
Model 4
Johnston and Hubert
(1995)[Ref. 51]
d
dt
------- K
1
K
2
+ ( ) 1 ( ) B ( ) =

c
d
dt
------- K
3
1 ( ) =
c
>
K
i
A
i
e
E
i
RT ( )
=
A
1
A
2
A
3
E
1
E
2
E
3
B
C
H
R
, , , , , , , ,
d
dt
------- K
m
1 ( )
n
=
K Ae
E RT ( )
=
A E m n H
R
, , , ,
d
dt
------- K 1 ( )
n
=
K Ae
E RT ( )
=
A E n H
R
, , ,
d
dt
------- K
1
K
2
m
+ ( ) 1 ( )
n
=
K
i
A
i
e
E
i
RT ( )
=
A
1
A
2
E
1
E
2
m n H
R
, , , , , ,
d
dt
------- K
1
1 ( )
l
K
2
m
1 ( )
n
+ =
K
i
A
i
e
E
i
RT ( )
=
A
1
A
2
E
1
E
2
l m n H
R
, , , , , , ,
d
dt
-------
K
m
1 ( )
n
1 e
C
C0

CT
T + ( ) { }
+
--------------------------------------------------------------- =
K
i
A
i
e
E
i
RT ( )
=
A E m n C
C0

CT
H
R
, , , , , , ,

In Marc, the table definition allows the user to define the cure rate as a function of the degree of cure and temperature
if the resin kinetics is not defined by the four embedded models.
In addition, the degree of cure and cure rate can also be calculated through the UCURE user subroutine.
Cure Shrinkage Strain
The cure shrinkage strain is calculated according to the volumetric shrinkage due to curing process. The resin degree
of cure shrinkage is defined as the ratio of volumetric cure shrinkage and maximum volumetric cure shrinkage
of the resin material, as . Equation (5-258) then calculates the cure shrinkage strain:
(5-258)
Considering the anisotropic shrinkage behavior of composite with resin, the strain components of the composite are
calculated by using the directional Cure Shrinkage Coefficient (CSC) matrix by Equation 5-259.
(5-259)
Table 5-8 Parameters used in the Embedded Resin Cure Kinetics Models
Variable Description Units
Resin degree of cure. -
Resin temperature K or R
Total resin heat of reaction ( = 0 to 1)
J/kg or BTU/lb
m
Pre-exponential factor. /s
Activation energy. J/mol or BTU/mol
Equation superscript. -
Gas constant J/(mol K) or BTU/(mol R)
Diffusion Constant -
Critical Resin Degree of Cure -
The Increase in Critical Resin Degree of Cure with Temperature -
T
H
R
A
i
E
i
l
m
n
R
C
C0
CT
V
r
S
( )
V
r
S
( )
S
V
r
S
V
r
S
----------- =
r
S
1 V
r
S
+ ( )
1 3
1 =
i j
S
CS C
i j

r
S
( ) =
i j , 1 2 3 , , =
215
CHAPTER 5
The calculation of volumetric cure shrinkage, , is conducted based on the cure shrinkage models defined by users.
Marc has two cure shrinkage models available. Note that a positive cure shrinkage strain indicates a reduction of
volume.
Tables 5-9 and 5-10 summarize the details of the cure kinetics models and the definition of the requires parameters.
Table 5-9 Resin Cure Shrinkage Models
Model Equations Parameters
Model 1
Bogetti and Gillespie (1992)
[Ref. 41]
Model 2
White and Hahn 1992
[Ref. 47]
Table 5-10 Parameters used in the Resin Cure Shrinkage Models
Variable Description Units
Resin volumetric cure shrinkage
-
Total volumetric resin shrinkage from = 0 to 1.
-
The degree of cure shrinkage.
-
Degree of cure after which the resin shrinkage begins (model 1).
-
Degree of cure after which the resin shrinkage stops (model 1).
-
Linear cure shrinkage coefficient
-
Degree of cure after which the resin shrinkage begins (model 2).
-
Cure shrinkage model superscript.
-
V
r
S
V
r
S
0.0 =

C1
<
V
r
S
A
*
S
V
r
S
A ( )
*
S
2
+ =
C1

C2
<
V
r
S
V
r
S
=

C2
S

C1
C2

C1
---------------------------
V
r
S
C1

C2
A , , ,

C
V
r
S
V
r
S
=
V
r
S
V
r
S
*10
B
c
( )
=

C
>
V
r
S
C
B , ,
V
r
S
V
r
S
C1
C2
A
C
B

References
1. Key, S. W. and R. D. Krieg, 1982, On the Numerical Implementation of Inelastic Time-Dependent, Finite
Strain Constitutive Equations in Structural Mechanics, Computer Methods in Applied Mechanics in
Engineering, V.33, pp. 439452, 1982.
2. Bathe, K. J., E. Ramm, and E. L. Wilson. Finite Element formulation for Large Deformation Dynamic
Analyses, International Journal for Numerical Methods in Engineering, V. 9, pp. 353-386, 1975.
3. Montgomery, D. C., Design and Analysis of Experiments, (2nd ed.) John Wiley and Sons, 1984.
4. Spendley, W., Hext, G. R., Himsworth, F. R., Sequential Application of Simplex Designs in Optimisation and
Evolutionary Operation, Technometrics, Vol. 4, No. 4,
pp. 441-461 (1962).
5. Gill, P. E., Murray, W., Saunders, M. A., Wright, M. H., Sequential Quadratic Programming Methods for
Nonlinear Programming, NATO-NSF-ARO, Advanced Study Inst. on Computer Aided Analysis and
Optimization of Mechanical System Dynamics, Iowa City, IA, Aug. 1-2, 1983, pp. 679-700.
6. Bathe, K. J. Finite Element Procedures Prentice-Hall, 1996.
7. Wilkinson, J. H. The Algebraic Eigenvalue Problem. Oxford: Clarendon Press, 1965.
8. Zienkiewicz, O. C., and Taylor, L. C. The Finite Element Method. Fourth Ed., Vol. 1 & 2. London: McGraw-
Hill, 1989.
9. Barsoum, R. S. On the Use of Isoparametric Finite Elements in Linear Fracture Mechanics. Int. J. Num.
Methods in Engr. 10, 1976.
10. Dunham, R. S. and Nickell, R. E. Finite Element Analysis of Axisymmetric Solids with Arbitrary Loadings.
No. 67-6. Structural Engineering Laboratory, University of California at Berkeley, June, 1967.
11. Hibbitt, H. D., Marcal, P. V. and Rice, J. R. A Finite Element Formulation for Problems of Large Strain and
Large Displacement. Int. J. Solids Structures 6 1069-1086, 1970.
12. Houbolt, J. C. A Recurrence Matrix Solution for the Dynamic Response of Elastic Aircraft. J. Aero. Sci. 17,
540-550, 1950.
13. McMeeking, R. M., and Rice, J. R. Finite-Element Formulations for Problems of Large Elastic-Plastic
Deformation. Int. J. Solids Structures 11, 601-616, 1975.
14. MARC Update. The Inverse Power Sweep Method in MARC. U.S. Ed., vol. 3, no. 1, February, 1984.
15. Simo, J. C., Taylor, R. L., and Pister, K. S. Variational and Projection Methods for the Volume Constraint in
Finite Deformation Elasto-Plasticity, Comp. Meth. in App. Mech. Engg., 51, 1985.
16. Simo, J. C. and Taylor, R. L. A Return Mapping Algorithm for Plane Stress Elasto-Plasticity, Int. J. of Num.
Meth. Engg., V. 22, 1986.
17. Wriggers, P., Eberlein, R., and Reese, S. A Comparison of Three-Dimensional Continuum and Shell Elements
for Finite Plasticity, Int. J. Solids & Structures, V. 33, N. 20-22, 1996.
18. Simo, J. C. and Taylor, R. L. Quasi-Incompressible Finite Elasticity in Principal Stretches, Continuum Basis
and Numerical Algorithms, Comp. Meth. App. Mech. Engg., 85, 1991.
217
CHAPTER 5
19. Marcal, P. V. Finite Element Analysis of Combined Problems of Nonlinear Material and Geometric
Behavior. in Proceedings of the ASME Computer Conference, Computational Approaches in Applied
Mechanics, Chicago, 1969.
20. Melosh, R. J., and Marcal, P. V. An Energy Basis for Mesh Refinement of Structural Continua. Int. J. Num.
Meth. Eng. 11, 1083-1091, 1971.
21. Morman, K. N., Jr., Kao, B. G., and Nagtegaal, J. C. Finite Element Analysis of Viscoelastic Elastomeric
Structures Vibrating about Nonlinear Statically Stressed Configurations. SAE Technical Papers Series
811309, presented at 4th Int. Conference on Vehicle Structural Mechanics, Detroit, November 18-20, 1981.
22. Morman, K. N., Jr., and Nagtegaal, J. C. Finite Element Analysis of Small Amplitude Vibrations in Pre-
Stressed Nonlinear Viscoelastic Solids. Int. J. Num. Meth. Engng, 1983.
23. Nagtegaal, J. C. Introduction in Geometrically Nonlinear Analysis. Int. Seminar on New Developments in
the Finite Element Method, Santa Marherita Ligure, Italy, 1980.
24. Nagtegaal, J. C., and de Jong, J. E. Some Computational Aspects of Elastic-Plastic Large Strain Analysis. in
Computational Methods in Nonlinear Mechanics, edited by J. T. Oden. North-Holland Publishing Company,
1980.
25. Newmark, N. M. A Method of Computation for Structural Dynamics. ASCE of Eng. Mech. 85, 67-94, 1959.
26. Parks, D. M. A Stiffness Derivative Finite Element Technique for Determination of Elastic Crack Tip Stress
Intensity Factors. International Journal of Fractures 10, (4), 487-502, December 1974.
27. Timoshenko, S., Young, D. H., and Weaver, Jr., W. Vibration Problems in Engineering. John Wiley, New
York: 1979.
28. Chung, J. and Hulbert, G.M., A family of single-step Houbolt time integration algorithms for structural
dynamics, Comp. Meth. in App. Mech. Engg., 118, 1994.
29. Shih, C.F., Moran B. and Nakamura K., Energy release rate along a three-dimensional crack front in a
thermally stressed body, International Journal of Fracture, vol. 30, pp. 79102, 1986.
30. Anderson, T.L., Fracture Mechanics: Fundamentals and Applications (2nd ed.) CRC Press, 1995.
31. Yoon, J.W., Yang, D.Y. and Chung, K., Elasto-plastic finite element method based on incremental
deformation theory and continuum based shell elements for planar anisotropic sheet materials, Comp.
Methods Appl. Mech. Eng., 174, 23 (1999).
32. Shih, C.F. and Asaro, R.J., Elastic-Plastic Analysis of Cracks on Bimaterial Interfaces: Part I Small Scale
Yielding, Journal of Applied Mechanics, Vol. 110, pp. 299316, June 1988.
33. Krueger, R., Virtual Crack Closure Technique: History, Approach and Applications, Appl. Mech. Rev., Vol.
57:2, pp. 109143, March 2004.
34. Knops, B., Numerical Simulation of Crack Growth in Pressurized Fuselages, Doctoral thesis, Faculty of
Aerospace Engineering, Delft University of Technology, The Netherlands, 1994.
35. Chung, J. and Hulbert, G.M., A time integration algorithm for structural dynamics with improved numerical
dissipation: The Generalized- Method, Journal of Applied Mechanics, Vol. 60, pp. 371 - 375, June 1993.
36. Hilber, H.M., Hughes, T.J.R., and Taylor, R.L., Improved Numerical Dissipation for Time Integration
Algorithms in Structural Dynamics, Earthquake Engineering and Structural Dynamics, Vol. 5, pp. 283-292,
1977

37. Wood, W.L., Bossak, M., and Zienkiewicz, O.C., An Alpha Modification of Newmarks Method,
International Journal for Numerical Methods in Engineering, Vol., 15, pp. 1562-1566, 1981.
38. J. Reeder, 3D Mixed-Mode Delamination Fracture Criteria - An Experimentalist's Perspective. Proceedings of
American Society for Composites, 21st Annual Technical Conference, Dearborn, Michigan, 2006.
39. M.L. Benzeggagh and M. Kenane, Measurement of Mixed-Mode Delamination Fracture Toughness of
Unidirectional Glass/Epoxy Composites with Mixed-Mode Bending Apparatus, Composites Science and
Technology, vol. 56, pp. 439-449, 1996.
40. Yih-Farn Chen, Modeling Analysis of Defect Formation during Thick Laminate Compression Molding
Process, Presented at the American Helicopter Society 58
th
Annual Forum, Montreal, Canada, June 11-13,
2002.
41. Bogetti, T. A., Gillespie, J. W., 1992 Process-Induced Stress and Deformaton in Thick Section thermoset
Composite Laminates, Journal of Composite Materials, 26, 626-660.
42. Lee, W.I., A. Loos and G.S. Springer. 1982 Cure Kinetics and Viscosity of Fiberite 976 Resin, Journal of
Composite Materials, 21; 243-261.
43. S.N. Lee, M.T. Chiu, and H.S. Lin, 1992 Kinetic Model for the Curing Reaction of a Tetraglycidyl Diamino
Diphenyl Methane/Diamino Diphenyl Sulfone (TGDDM/DDS) Epoxy Resin System, Polymer Engineering
and Science 32 (15); pp. 1037-1046.
44. T. G. Gutowski, et. al. The consolidation of laminate composites, Journal of Composite Materials, Vol. 21,
February 1987, pp. 172 187.
45. W. B. Young, Consolidation and cure simulations for laminated composites, Polymer Composites, February
1996, Vol. 17, No. 1, pp. 142 148
46. Min Li and Charles Tucker, Optimal Curing for Thermoset Matrix Composites: Thermochemical and
Consolidation Considerations, Polymer Composites, Aug. 30, 2001
47. S. R. White and H.T. Hahn, 1992 Process Modeling of Composite Materials: Residual Stress Development
during Cure. Part I. Model Formulation, Journal of Composite Materials, 26, 2402-2422.
48. J.M. Kenny, 1992 Integration of Process Models with Control and Optimization of Polymer Composites
Fabrication, Proceedings of the Third Conference on Computer Aided Design in Composite Materials
Technology, pp. 530-544.
49. E.P. Scott, 1991 Determination of Kinetic Parameters Associated with the Curing of Thermoset Resins Using
Dielectric and DSC Data, Composites: Design, Manufacture, and Application, ICCM/VIII, Honolulu, 1991, pp.
10-0- 1-10.
50. George S. Springer, Resin Flow During the Cure of Fiber Reinforced Composites, Journal of Composite
Materials 1982 16: 400-410.
51. P. Hubert, 1996 Aspects of Flow and Compaction of Laminated Composite Shapes During Cure, Ph.D. Thesis,
The University of British Columbia, Vancouver, B.C.
Chapter 6 Nonstructural Procedure Library
6
Nonstructural Procedure
Library
Heat Transfer 220
Diffusion 280
Hydrodynamic Bearing 285
Electrostatic Analysis 289
Magnetostatic Analysis 293
Magnetodynamic Analysis 305
Piezoelectric Analysis 310
Acoustic Analysis 314
Fluid Mechanics 315
Coupled Analyses 324
References 344

This chapter describes the nonstructural analysis procedures available in Marc. These are comprised of several areas
including heat transfer, hydrodynamic bearing, electrostatic, magnetostatic, electromagnetic, piezoelectric, fluid
mechanics, and coupled analysis. This chapter provides the technical background information as well as usage
information about these capabilities.
Heat Transfer
Marc contains a solid body heat transfer capability for one-, two-, and three-dimensional, steady-state and transient
analyses. This capability allows you to obtain temperature distributions in a structure for linear and nonlinear heat
transfer problems. The nonlinearities in the problem may include temperature-dependent properties, latent heat (phase
change) effect, heat convection in the flow direction, and nonlinear boundary conditions (convection and radiation).
The temperature distributions can, in turn, be used to generate thermal loads in a stress analysis.
Marc can be applied to solve the full range of two- and three-dimensional transient and steady-state heat conduction
and heat convection problems. Marc provides heat transfer elements that are compatible with stress elements.
Consequently, the same mesh can be used for both the heat transfer and stress analyses. Transient heat transfer is an
initial- boundary value problem, so proper initial and boundary conditions must be prescribed to the problem in order
to obtain a realistic solution. Marc accepts nonuniform nodal temperature distribution as the initial condition, and can
handle temperature/time-dependent boundary conditions. The thermal conductivity can be isotropic, orthotropic, or
anisotropic. Both the thermal conductivity and the specific heat in the problem can be dependent on temperature;
however, for conventional heat transfer, the mass density remains constant at all times. Latent heat effects (solid-to-
solid, solid-to-liquid phase changes) can be included in the analysis. A time-stepping procedure is available for
transient heat transfer analysis. Temperature histories can be stored on a post file and used directly as thermal loads in
subsequent stress analysis. User subroutines are available for complex boundary conditions such as nonlinear heat
flux, directional heat flux, convection, and radiation.
A summary of Marc capabilities for transient and steady-state analysis is given below.
Selection of the following elements that are compatible with stress analysis:
1-D: three-dimensional link (2-node, 3-node)
2-D: planar and axisymmetric element (3-, 4-, and 8-node)
2-D: axisymmetric shells
3-D: solid elements (4-, 6-, 8-, 10-, 15-, and 20-node)
3-D: membrane elements (3-,4-, 6-, and 8-node)
3-D: shell elements (4- and 8-node)
Specification of temperature-dependent materials (including latent heat effects) is performed with the
ISOTROPIC, ORTHOTROPIC, ANISOTROPIC, TEMPERATURE EFFECTS, ORTHO TEMP, TABLE, and
LATENT HEAT model definition options.
Selection of initial conditions is done using the INITIAL TEMP option.
Selection of the temperature and time-dependent boundary conditions (prescribed temperature history,
volumetric flux, surface flux, film coefficients, radiation, change of prescribed temperature boundary
conditions during analyses) is done using the FIXED TEMPERATURE, TEMP CHANGE, DIST FLUXES,
POINT FLUX, QVECT, and FILMS options. Moving heat sources due to welding can be specified using the
WELD FLUX, WELD PATH, and WELD FILL options.
Import of viewfactors calculated by Mentat for radiation analyses.
221
CHAPTER 6
Nonstructural Procedure Library
Selection of time steps using the TRANSIENT or AUTO STEP history definition option.
Application of a tying constraints on nodal temperatures using the TYING model definition option.
Generation of a thermal load (temperature) file using the POST option, which can be directly interfaced with
the stress analysis using the CHANGE STATE model definition option.
Use of the ANKOND user subroutines for anisotropic thermal conductivity, FILM or UFILM for convective and
radiative boundary conditions or FLUX or QVECT for heat flux boundary conditions.
Selection of nodal velocity vectors for heat convection is done using VELOCITY and VELOCITY CHANGE
options.
Use of the UVELOC user subroutines for heat convection.
In addition, a number of thermal contact gap and fluid channel elements are available in Marc. These elements can be
used for heat transfer problems involving thermal contact gap and fluid channel conditions.
The CONRAD GAP option, used in conjunction with 4- and 8-node 2-D continuum elements or 8- and 20-node 3-D
continuum elements, provides a mechanism for perfect conduction or radiation/convection between surfaces,
depending on the surface temperatures. The perfect conduction capability in the elements allows for the enforcement
of equal temperatures at nodal pairs and the radiation/convection capability allows for nonlinear heat conduction
between surfaces, depending on film coefficient and emissivity. The perfect conduction is simulated by applying a
tying constraint on temperatures of the corresponding nodal points. An automatic tying procedure has been developed
for such elements. The radiation/convection capabilities in the elements are modeled by one-dimensional heat transfer
in the thickness direction of the elements with variable thermal conductivity.
For the purpose of cooling, the CHANNEL option allows coolant to flow through passage ways that often appear in the
solid. The fluid channel elements are designed for the simulation of one-dimensional fluid/solid convection conditions
based on the following assumptions:
1. Heat conduction in the flow direction can be neglected compared to heat convection;
2. The heat flux associated with transient effects in the fluid (changes in fluid temperature at a fixed point in
space) can be neglected.
For high velocity air flow, these assumptions are reasonable and the following approach is used:
Each cooling channel is modeled using channel elements. On the sides of the channel elements, convection is
applied automatically. The film coefficient is equal to the film coefficient between fluid and solid, whereas the
sink temperature represents the temperature of the fluid. A two-step staggered solution procedure is used to
solve for the weakly coupled fluid and solid temperatures.
Thermal Contact
The CONTACT and THERMAL CONTACT options may also be used to define the fluxes entering the surfaces. For
more information, see Chapter 8 in this volume.
Use the CONTROL model definition option to input convergence controls for heat transfer analysis. These options are:
1. Maximum allowable nodal temperature change. This is used only for transient heat transfer analysis in
conjunction with an automatic time stepping scheme. It controls the time step size.

2. Maximum allowable nodal temperature change before properties are re-evaluated and matrices reassembled.
This is used only for transient heat transfer analysis in conjunction with an automatic time stepping scheme.
For mildly nonlinear problems, if the time step remains constant, the operator matrix is not reassembled until
this value is reached.
3. Maximum error in temperature estimate used for property evaluation. This control provides a recycling
capability to improve accuracy in highly nonlinear heat transfer problems (for instance, for latent heat or
radiation boundary conditions). When nonzero, this control is used for both steady state and transient heat
transfer analysis when either the fixed or adaptive stepping procedures are used.
For steady state problems, use the STEADY STATE history definition option. If the problem is nonlinear, use the
tolerance for temperature estimate error on the CONTROL model definition set to obtain an accurate solution.
You must distinguish between two cases of nonlinearity in steady-state solutions: mild nonlinearities and severe
nonlinearities. In the case of mild nonlinearities, variations are small in properties, film coefficients, etc., with respect
to temperatures. The steady-state solution can be obtained by iteration. After a small number of iterations, the solution
should converge.
The technique described above is not suitable for severe nonlinearities. Examples of severe nonlinearities are radiation
boundary conditions and internal phase-change boundaries. In these cases, you must track the transient with a
sufficiently small time step ( ) to retain stability until the steady-state solution is reached. Clearly, the choice of
is dependent on the severity of the nonlinearity. The number of steps necessary to obtain the steady-state solution can
often be reduced by judicious choice of initial conditions. The closer the initial temperatures are to the steady-state,
the fewer the number of increments necessary to reach steady-state.
Transient Analysis
Both fixed stepping and adaptive stepping schemes are available for transient heat transfer analysis. The fixed stepping
scheme is TRANSIENT NON AUTO. The adaptive stepping schemes are TRANSIENT and AUTO STEP.
In the fixed stepping scheme, the program is forced to step through the transient with a fixed time step that is user
specified. Only the error in temperature estimate (convergence control 3) is used with the fixed stepping scheme and
the first two convergence controls are not checked.
The scheme for the TRANSIENT option is as follows:
1. After the program obtains a solution for a step, it calculates the maximum temperature change in the step and
checks this value against the allowable temperature change per step given on the CONTROL option
(convergence control 1).
2. If the actual maximum change exceeds the specified value, the program repeats the step with a smaller time
step and continues repeating this step until the maximum temperature change is smaller than the specified value
or until the maximum number of recycles given on the control option is reached (in which case the program
stops).
3. If the actual maximum temperature change is:
a. between 80 percent and 100 percent of the specified value, the program goes on to the next step, using the
same time step.
t t
223
CHAPTER 6
b. between 65 percent and 80 percent of the specified value, the program tries the next step with a time step of
1.25 times the current step.If the actual maximum is below 65 percent of the specified value, the program
tries the next step with a time step that is 1.5 times the current step.
The objective of the scheme is to increase the time step as the analysis proceeds.
The TRANSIENT option can be used for both heat transfer as well as thermo-mechanical coupled analysis. The time
step for coupled analyses is only based on the convergence characteristics of the heat transfer pass.
The AUTO STEP option is a unified time-stepping scheme that is available for thermal, mechanical and thermo-
mechanically coupled analysis. The time step for coupled analyses is based on the convergence characteristics of each
of the heat transfer and structural passes. Details of the AUTO STEP scheme are provided in Chapter 11.
Let the temperature within an element be interpolated from the nodal values of the element through the
interpolation functions ,
(6-1)
The governing equation of the heat transfer problem is
(6-2)
In Equation (6-2), and are the temperature-dependent heat capacity and thermal conductivity matrices,
respectively, is the nodal temperature vector, is the time derivative of the temperature vector, and is the heat
flux vector.
The selection of the backward difference scheme for the discretization of the time variable in Equation (6-2) yields the
following expression:
(6-3)
Equation (6-3) computes nodal temperatures for each time increment .
For the evaluation of temperature-dependent matrices, the temperatures at two previous steps provide a linear
(extrapolated) temperature description over the desired interval
(6-4)
This temperature is then used to obtain an average property of the material over the interval to be used in Equation
(6-3), such that
(6-5)
T x ( ) T
N x ( )
T x ( ) N x ( )T =
C T ( )
dt
dt
----- K T ( )T + Q =
C T ( ) K T ( )
T T
Q
1
t
----- C T ( ) K T ( ) + T
n
Q
n
1
t
----- C T ( )T
n 1
+ =
t ( )
T ( ) T t t ( ) =

t
----- T t t ( ) T t 2t ( ) ( ) +
f
f
1
t
----- f T ( ) [ ]d
t t
t
}
=

During iteration, the average property is obtained based on the results of the previous iteration:
(6-6)
are the results of the previous iteration.
Temperature Effects
The thermal conductivity, specific heat, and emissivity in a heat transfer analysis can depend on temperatures;
however, the mass density remains constant. Specify reference temperature values of thermal conductivity, specific
heat, mass density, and emissivity with the ISOTROPIC option. Enter temperature variations of both the thermal
conductivity and specific heat using the TEMPERATURE EFFECTS or the TABLE option. This option also allows
input of latent heat information. The temperature-dependent data can be entered using either the slope-break point
representation or the property versus temperature representation. During analysis, an extrapolated/interpolated
averaging procedure is used for the evaluation of temperature-dependent properties.
Latent heat can be induced because of a phase change that can be characterized as solid-to-solid, solid-to-fluid, fluid-
to-solid, or a combination of the above, depending on the nature of the process.
Phase change is a complex material behavior. Thus, a detailed modeling of this change of material characteristic is
generally very difficult. The use of numerical models to simulate these important phenomena is possible; several major
factors associated with phase change of certain materials have been studied numerically.
The basic assumption of the latent heat option in Marc is that the latent heat is uniformly released in a temperature
range between solidus and liquidus temperatures of the materials (see Figure 6-1). Marc uses a modified specific heat
to model the latent heat effect. If the experimental data is sufficient and available, a direct input of the temperature-
dependent specific heat data (see Figure 6-2) can be used. Results of both approaches are comparable if the
temperature increments are relatively small.
Figure 6-1 Modeling Phase Changes with the Latent Heat Option
T ( ) T t t ( )

t
----- T* t ( ) T t t ( ) ( ) + =
T* t ( )
S
p
e
c
i
f
i
c

H
e
a
t
Temperature
Latent Heat Uniformly Release
Between ST and LT
ST Solidus Temperature
LT Liquidus Temperature
ST LT
225
CHAPTER 6
Figure 6-2 Modeling Phase Changes with the Specific Heat Option
Initial Conditions
In a transient heat transfer analysis, Marc accepts nonuniform nodal temperature distribution as the initial condition.
Enter the initial condition through the INITIAL TEMP model definition option or through the USINC user subroutine.
Initial conditions are not required in steady-state heat transfer analysis, even though they can improve convergence
when temperature-dependent properties are included.
Boundary Conditions
There are two types of boundary conditions in transient/steady-state heat transfer analysis: prescribed nodal
temperatures and nodal/element heat fluxes. These boundary conditions are entered directly through input or through
user subroutines.
Prescribe nodal temperatures using the FIXED TEMPERATURE model definition option. These prescribed
temperatures are constant with time unless a time dependent table is referenced, the FORCDT user subroutine is
invoked, or a TEMP CHANGE history definition option is encountered.
The POINT FLUX model definition option allows you to enter concentrated (nodal) heat fluxes such as heat source and
heat sink. These applied fluxes are constant with time unless a time dependent table is referenced, the FORCDT user
subroutine or a POINT FLUX history definition option is encountered.
The DIST FLUXES or QVECT model definition option allow the definition of distributed (surface or volumetric) heat
flux. These applied fluxes may be spatially varying by an invoked spatial table or by the use of the FLUX or UQVECT
user subroutine. These applied fluxes are constant with time unless a time dependent table is referenced, the FLUX or
UQVECT user subroutine is invoked, or a DIST FLUXES history definition option is encountered. A special case is the
volumetric heat generated by inelastic loads. This will be calculated by the program when the distributed flux type is
101. It may also be necessary to include the CONVERT option. For spatially varying fluxes associated with a welding
heat source, it is more convenient to use the WELD FLUX model definition option. Specialized welding related options
are discussed in the next section.
S
p
e
c
i
f
i
c

H
e
a
t
Temperature

Use the FILMS model definition option to input the constant film coefficient and the ambient temperature associated
with the convective boundary conditions. Use the FILM user subroutine for time/temperature-dependent convective
boundary conditions. When not using the table input option, the expression of the convective boundary condition is
(6-7)
where , , , and are heat flux, film coefficient, unknown surface temperature, and ambient temperature,
respectively.
The radiative boundary condition can be expressed as
(6-8)
where is the heat flux, is the Stefan-Boltzmann coefficient, is emissivity, and and are unknown
surface and ambient temperatures, respectively. The radiative boundary condition can be rewritten as
(6-9)
This shows that the radiative boundary condition is equivalent to a nonlinear convective boundary condition, in which
the equivalent film coefficient depends on the unknown surface temperature . When using the
nontable driven input, this case requires the FILM user subroutine.
When using the table driven input, the distributed thermal flux into an element may be defined as:
(6-10)
where:
Notes:
In heat transfer analysis, you must always specify total values; for example, total temperature boundary
conditions or total fluxes. This specification is to be used consistently for the heat transfer portion of
analysis on coupled thermal-mechanical (thermal-solid), fluid-thermal, and fluid-thermal-solid.
The time variation used for thermal boundary conditions is a function of the stepping procedure that is used.
If adaptive stepping (either TRANSIENT or AUTO STEP) is used, thermal boundary conditions are applied
instantaneously. If fixed stepping (TRANSIENT NON AUTO) is used, the magnitudes of the thermal
boundary conditions are applied in accordance with the tables used to define them.
is the conventional film coefficient
is the unknown surface temperature
is the sink temperature
q H T
s
T
( ) =
q H T
s
T
q T
sa
4
T
a
4
( ) =
q T
sa
T
a
q T
sa
3
T
sa
2
T
a
T
sa
T
a
2
T
a
3
+ + + ( ) T
s
T
( ) =
H T
s
T
, , , ( ) T
s
T
( ) =
H T
s
T
, , , ( ) T
s
q H T
s
T
( ) H
nv
T
s
T
( )
e
vf
T
sa
4
T
a
4
( ) f + + + =
H
T
s
T
227
CHAPTER 6
The absolute temperature is the temperature in user units plus the offset, which is defined in the PARAMETERS model
definition option.
A surface flux may be entered on the FILMS model definition option using the table driven input or the DIST FLUXES
model definition option. If both convective behavior and other terms contributing to the flux are present it is
computationally efficient to do it here.
When using table driven input, all coefficients given above may be temperature dependent by referencing a table. If
more complex physical boundary conditions exist, then the UFLIM user subroutine should be used.
When using the table driven input, an alternative method is available for defining the sink temperature through the use
of the SINK POINTS model definition option. This allows the user to interface to exterior fluid/thermal programs to
input spatially varying environment temperatures. In Figure 6-3, temperatures may be known at the points marked
with an X. Using the SINK POINTS model definition option when evaluating the flux into an element, the program
will first determine the closest sink point to the surface integration point, and then use this sink point temperature in
the calculation.
Figure 6-3 Temperature Points
If the heat flux defined in the FLUX or FORCDT user subroutine is temperature dependent, then the convergence
behavior of the solution process of the set of nonlinear equations can be improved by defining not only the current flux
value, but also the derivative of the flux with respect to temperature. Based on this derivative, Marc adapts the
conductivity matrix as well as the heat flux vector . This is done by rewriting the heat flux using a Taylor
series expansion. For a temperature dependent point flux on node , the flux during iteration is thus given by:
(6-11)
is the unknown surface temperature in absolute units
is the sink temperature in absolute units
is the natural convection coefficient
is the natural convection exponent
in the effective view factor to the environment; usually equal to 1.0
is the emissivity
is the Stefan Boltzman constant
is the distributed flux from other sources
T
sa
T
a
H
nv
e
vf
f
x
x
x
x
x
x
x
x
K T ( ) Q
i n
q
n
i
q
n 1
i q
i
d
T d
-------
\ .
| |
n 1
T
n
i
T
n 1
i
( ) + =

Since all the terms in Equation (6-11) except for are known from iteration , the corresponding entries in the
conductivity matrix and heat vector are updated as:
(6-12)
(6-13)
The temperature dependent distributed fluxes are treated in a similar way and require the FILM user subroutine.
If a film coefficient is temperature dependent, then, similar to temperature dependent fluxes, defining the derivative
of the film coefficient with respect to temperature may speed up the convergence of the iterative process. This
derivative should be defined in the FILM user subroutine.
The QVECT option may be used to define a directed flux into the surface. The user specifies a magnitude and an
orientation . The magnitude of the flux applied is then where is the area of the surface,
is the face absorptivity, and in the outward unit normal into the surface. The absorptivity is defined through the
EMISSIVITY option. Note that when using heat transfer membranes or shell element, the normal is based upon the
right-hand rule.
For both the FILMS and QVECT option using table driven input, the coefficients used in the expressions may be
temperature dependent or dependent upon other quantities.
For FILMS, this lead to:
and for QVECT, this leads to:
The user can specify that the evaluation temperature be either at the surface (TEMTYP=0), the average of the surface
and environment temperature (TEMPTYP=10), or the environment temperature (TEMPTYP=20).
Furthermore, for the FILMS option, one can specify that the environment temperature at a control node; in which case,
Control Systems
Often in heat transfer, one desires that a boundary condition is dependent upon the temperature at another point. In
HVAC systems, this is driven by the temperature at the thermostat (multiple thermostats). If multiple thermostats are
involved, an effective temperature should be obtained by using a Servo Link.
T
n
i
n 1
K
n
i i
K
n
i i q
i
d
T d
-------
\ .
| |
n 1

Q
n
i
Q
n
i q
i
d
T d
-------
\ .
| |
n 1
T
n 1
i

q
0
n
Q q
0
A n
n
s
= A
n
s
q H T
eval
( ) T
s
T
( ) =
q q
0
T
eval
( ) T
eval
( ) A n
n
s
( ) =
q H T
eval
( ) T
s
T
T
cnt r l nd
( ) ( ) =
229
CHAPTER 6
If a control node is specified, then the following boundary conditions are applied:
The FLUXTYPE is specified in the respective options. FLUXTYPE=4 is compatible with Nastran definition of control
node.
Velocity and Pressure Dependent Convection
In many applications, the convective coefficients are dependent upon the external fluid boundary conditions. In such
cases, one can express:
where and are the pressure and the velocity of the flow field. This simulation can be done by using the VELOCITY
option to define the velocity and the FIXED PRESSURE option to define the pressure at the nodal points on the
boundary. This data is then interpolated to the integration points where it is used to evaluate the function/table
describing H. Furthermore, in a transient simulation where the velocity and pressure are a function of time, the
VELOCITY and FIXED PRESSURE option may be used to define the temporal variation. In the most general case, one
then has:
If spatial variation is also required, then the UFILM user subroutine is required.
Point Flux
FLUXTYPE=6
FLUXTYPE=4
Distributed Flux
FLUXTYPE=6
FLUXTYPE=4
Film
FLUXTYPE=6
FLUXTYPE=4
QVECT
FLUXTYPE=6
FLUXTYPE=4
q Q T
cnt r l nd
( ) =
q Q T
cnt r l nd
=
q q
a
T
cnt r l nd
( ) A =
q q
a
T
cnt rl nd
A =
q H T
eval
( ) T
s
T
T
cnt r l nd
( ) ( ) =
q H T
eval
( ) T
cnt r l nd
T
s
T
( ) =
q q
o
T
cnt r l nd
( ) T
eval
( ) A n
s
( ) =
q q
o
T
cnt r l nd
T
eval
( ) A n
s
( ) =
q H T
eval
V P , , ( ) T
s
T
( ) =
P V
q H T
eval
V t ( ) P t ( ) t , , , ( ) T
s
T
t ( ) ( ) =

Surface Energy
The SURFACE ENERGY option calculates contributions to the thermal boundary conditions, including the recession
due to thermochemical ablation at the surface of a material which is subjected to very high thermal fluxes. This
includes the effects of convective heat flux with a possible blowing effect by mass injection, an enthalpy flux due to
molecular diffusion, a thermochemical ablation due to heterogeneous chemical reactions with gases, a possible
thermal internal decomposition, a thermochemical ablation and a mechanical erosion by liquid or solid particles
impacts. Surface energy allows one to obtain the recession due to thermochemical ablation by gases or particles. The
mechanical erosion must be calculated by use of a UFLUXMEC user subroutine, and added to the thermochemical
ablation in order to obtain the total surface recession. Surface energy is coupled with the heat transfer into the material
by the conductive heat flux, and possibly with the mass flow rate towards the surface due to water evaporation and/or
thermal decomposition gases generated inside the material.
The SURFACE ENERGY option may be used in any heat transfer analysis, but depending on the model, some terms
could be null. When used in conjunction with the THERMO-PORE option for materials undergoing pyrolysis, most of
the terms are nonzero.
The PRINT,15 parameter allows the printing of each contribution to the surface energy.
Thermochemical Ablation and Surface Energy Balance
Physical Presentation
Figure 6-4 Schema of the Surface Energy Balance
Surface Energy Balance
The surface energy balance is summarized by this equation:
Convection + Enthalpy flux due to external chemical species diffusion effects + Enthalpy flux due to
thermochemical ablation by gases + Enthalpy flux due to internal thermal decomposition products flow
Enthalpy flux due to blowing of gaseous products + Enthalpy flux due to thermochemical ablation by impacting
particles + Thermal effects of mechanical erosion by impacting particles Enthalpy flux due to mechanical
removal of liquid phases at the surface + Radiative heat transfer balance Thermal conduction = 0

surface
wall
flow
convection
conduction decomposition ablation by
particles
ablation by
gases
radiative
balance
particles
impact
diffusion
blowing
231
CHAPTER 6
Mathematical Presentation
The equation below is noted
It is homogeneous with a power per surface unit .
Details on Terms:
Each time that a upper-case is used for an enthalpy, it means that is the total enthalpy, including both the formation
enthalpy and the so-called sensible enthalpy. If we refer only to the sensible part, a lower-case is used. In all cases,
we refer to specific enthalpies .
Convection: enthalpy flux at the surface due to processes of heat conduction in the
boundary layer.
Diffusion: enthalpy flux at the surface due to processes of chemical species
diffusion in the boundary layer.
Thermochemical ablation by gases: enthalpy flow associated to mass flow of thermochemically ablated
material by gases.
Decomposition products flow: enthalpy flow from inside of the material and ejected to the boundary
layer without change of phase (generally pyrolysis gases and water).
Blowing: enthalpy flux due to evacuation of gases freed both by thermochemical
ablation (by both gaseous products and chemical effects of particle
impacts) at the surface, and internal thermal decomposition (including
possibly water evaporation).
Liquid phases removal: enthalpy flow associated with the blowing of liquid phases possibly
appearing at the surface, by external actions (flow, )
Thermochemical ablation by external
particles:
enthalpic flow associated with mass flow of thermochemically ablated
material by surface impacting particles.
Thermal effects of mechanical erosion by
impacts of particles:
enthalpy flow associated with energy given to the surface by impacting
particles (fraction of conversed kinetic energy).
Radiative balance: balance of radiative heat flux, resulting from absorption and emission.
Conduction: apparent conductive flow within the material.
conv
x t T
s
, , ( )
di ff
x t T
s
, , ( )
s t h g , ,
x t T
s
m
s t h ,
, , , ( )
g
x t T
s
m
g p ,
m
g w ,
, , , , ( )
bl ow
x t T
s
m
g p ,
m
g w ,
m
s t h ,
, , , , , ( )
l i q
x t T
s
m
l j ,
, , , ( )
s t h p , ,
x t T
s
m
p j ,
, , , , ( )
par t ki n ,
x t T
s
m
p j ,
, , , , ( )
r ad
x t T
s
, , ( )
cond
x t T
s
, , ( )
+ + +
+
+ + 0 =
W. m
2
[ ]
H
h
J. kg
1

Equation of Each Term of the Energy Balance:
Convection:
with:
Often, a 'blowing correction' modifies the heat transfer coefficient. In fact, there are two possibilities:
1. the CFD software allows one to take into account the blowing in some manner, for instance by using a
simplified subroutine representing the material behavior as a boundary condition. In that case, the coefficient
is obtained directly.
2. the CFD software allows one to calculate only a heat transfer coefficient for 'inert' wall (no blowing). So,
this coefficient must be corrected in the thermal software. The usual correction is:
with:
is the transpiration factor. Usual values are for laminar flow and for turbulent flow.
heat transfer coefficient (homogeneous with surface mass flow rate
/specific enthalpy ). In some papers, is sometimes written
as , with

being the Stanton number for heat transfer, and the mass
flow rate of the external flow parallel to the surface at the edge of the boundary layer.
specific recovery enthalpy of the external flow . This quantity depends also
implicitly on the external pressure. But this dependence is hidden behind the time
dependence (because the pressure is itself a function of the time and the space variable).
specific enthalpy of the external flow, calculated for the frozen chemical composition
existing at the edge of the boundary layer, but evaluated at the surface temperature
. This term is calculated by neglecting the variation of the chemical composition of
the external flow between the edge of the boundary layer and the wall, due to
both chemical reactions in the gas phase, and molecular diffusion of chemical species
in the boundary layer with possibly unequal mass diffusion coefficients for the species.
Other diffusion effects of the second-order are neglected (Soret effect, Dufour effect).
The non-neglected diffusion effects are taken into account in the diffusion term. This
quantity depends also implicitly on the external pressure. But this dependence is hidden
behind the time dependence (because the pressure is itself a function of the time and
the space variable).
H
H
r ec
H
e
T
s
;
\ .
| |
H

H
x t T
s
, , ( ) =
kg. m
2
. s
1
[ ] J. kg
1
[ ]
H
e
u
e
C
H
C
H

e
u
e
e
H
r ec
H
r ec
x t , ( ) =
J. kg
1
[ ] ( )
H
e
T
s
; H
e
T
s
x t T
s
, , ( ) ; =
e
T
s
H
x t T
s
, , ( )
H0
C
H
C
H
0
----------

H
H
0
----------
2B'
v
0
;
2B'
v
0
; 1
\ .
| |
exp
---------------------------------------------
Ln 1 2B'
v
; +
\ .
| |
2B'
v
;
----------------------------------------- = = =
B'
v
0
;
v ( )
w
H
0
--------------- B'
v
v ( )
w
H
--------------- ;
0.5 = 0.4 =
233
CHAPTER 6
Marc has two options to enter the heat transfer coefficient: either directly or

with the above correction. For
the latter, there is both the standard form above with as an entry in the data deck, and a UFAH user subroutine
allowing one to enter another kind of correction.
Diffusion:
with:
The quantities and are tabulated as a whole.
mass transfer coefficient (homogeneous with surface mass flow rate
/specific enthalpy ). In some papers, is sometimes written as
;

being the Stanton number for mass transfer and the mass flow rate of
the external flow parallel to the surface at the edge of the boundary layer. Very often,
(or ) is supposed to be correlated to (or ) by , being the
Lewis number for the frozen external flow. This Lewis number is often supposed to be equal
to 1, so that and is meaningless Otherwise, depends on the configuration of
the external flow, a value is generally accepted. For the sake of generality, one
can enter separate entry tables for and .
specific enthalpy of the chemical component in the external flow, evaluated at the surface
temperature

.

special fraction characterizing the component in the chemical composition of the external
flow at the edge of the boundary layer. has no dimension (no unit). It is a rather
complicated algebraic combination of the mole fractions and the mass fractions of all the
chemical species existing in the flow. These quantities depend also implicitly on the
external pressure and on the temperature. But these dependencies are hidden behind the
time dependence (because the pressure and the temperature of the flow are themselves
functions of the time and the space variable).
special fraction characterizing the component in the chemical composition of the external
flow near the surface of the material. The difference between and
characterizes the diffusion effects across the boundary layer. These quantities depend also
implicitly on the external pressure and on the temperature. But these dependencies are
hidden behind the time dependence (because the pressure and the temperature of the flow
are themselves functions of the time and the space variable).
H

H
0
M
Z
i
*
e
Z
i
*
s
; ;
\ .
| |
H
T
s
i
i
M

M
x t T
s
, , ( ) =
kg. m
2
. s
1
[ ] J. kg
1
[ ]
M
e
u
e
C
M
C
M

e
u
e
e C
M
M
C
H

H
C
M
C
H
Le ( )
n
Le
C
M
C
H
n n
n 2 3 =
H

M
H
i
T
s
H
i
T
s
x t T
s
, , ( ) =
i
T
s
J. kg
1
[ ] ( )
Z
i
*
e
; Z
i
*
e
x t , ( ) ; =
i
e Z
i
*
e
;
Z
i
*
s
; Z
i
*
s
x t , ( ) ; =
i
s Z
i
*
e
; Z
i
*
s
;
Z
i
*
e
H
i
T
s
x t T
s
, , ( ) ;
i
Z
i
*
s
H
i
T
s
x t T
s
, , ( ) ;
i

Thermochemical Ablation by External Gases:
A table gives as a function of , for each material possibly present at the surface.
Decomposition Products Flow:

Blowing:
with:
For each point and time (discrete values), a table gives as a function of the mass flow rate of gaseous products
coming from the inside (pyrolysis, water evaporation), the total mass flow rate coming from thermochemical ablation,
and the surface temperature .
Thermochemical Ablation by Impacting Particles:
with:
surface mass flow rate of thermochemically ablated surface material by the action
of external gases .
specific enthalpy of the solid material of the surface, evaluated at the surface
temperature .

surface mass flow rate of gaseous products produced internally by thermal decomposition
(both pyrolysis and water evaporation, ). Note that if there is no pyrolysis in the
model defined by the THERMO-PORE option, this term is zero.
specific enthalpy of the gases taken into account in, evaluated at the surface temperature
. This quantity is tabulated.
surface mass flow rate of all gaseous products leaving the surface of the
material (including gas coming from internal thermal decomposition
[pyrolysis, water, ], thermochemical ablation by external gases,
thermochemical ablation by impacting particles) .
specific enthalpy of the mix of all gaseous products existing at the
surface, including blowed gases and chemical species from the external
flow, evaluated at the surface temperature .
surface mass flow rate of particles for the family
.
empiric law for thermochemical ablation by impacting particles
(without unit)
m
s t h g , ,
H
s
m
s t h g , ,
m
s t h g , ,
x t m
g
T
s
, , , ( ) =
kg. m
2
. s
1
[ ] ( )
H
s
H
s
T
s
( ) =
T
s
J. kg
1
[ ] ( )
H
s
T
s
m
g
H
g
m
g
m
g
x t , ( ) =
kg. m
2
. s
1
[ ]
H
g
H
g
T
s
( ) =
T
s
J. kg
1
[ ] ( )
v ( )
w
H
w
v ( )
w
v ( )
w
m
g
m
s t h g , ,
m
s t h p , ,
, ( ) =
kg. m
2
. s
1
[ ]
H
w
H
w
x t T
s
m
g
m
s t h g , ,
m
s t h p , ,
, ( ) , , =
T
s
J. kg
1
[ ] ( )
x t H
w
T
s
m
s t h p , ,
f
t h p ,
T
s
( ) G
t h p j , ,
m
p j ,
H
r p j , ,
=
m
p j ,
m
p g ,
x t , ( ) j
kg. m
2
. s
1
[ ] ( )
G
t h p j , ,
G
t h p j , ,
V
p j ,
x t , ( ) D
p j ,

p j ,
x t , ( ) , , , ( ) =
235
CHAPTER 6
The summation is based on families of particles. Each family is characterized by an impacting mass flow rate, a
velocity of impact , an angle of impact , and the mean diameter of the family . The three first quantities
are generally dependent on time, and on the location on the surface. The mass flow rate of the surface material ablated
thermochemically by the family is generally given by a correlation of these parameters
(often called G-Law):
The correlation is established by experiments. The G-Law can be entered by a UGLAW user subroutine or a table.The
total ablated mass flow rate is obtained by summing the above formula for the families.For the sake of
generality, all the families are not obliged to be composed by particles of the same chemical species. So it is more
general to consider an enthalpy of reaction different for each family : .
The enthalpy of reaction depends on the temperature of the surface

This enthalpy of reaction is a property of both
the surface material (which can be different depending on the location on the part to be calculated), and the family of
particles under consideration. It is attached to the material, and the family of particle is considered as a parameter.
This model does not take into account the effect of the temperature of the surface, as far as the ablated mass flow rate
is concerned. An empiric correction has been proposed by several authors. The value given above (after summation
on ) must be corrected by multiplying it by a function of the surface temperature. Several functions have been
proposed:
with and .
These two corrections are implemented in standard input, and an alternate UFTHP user subroutine allows one to enter
other kinds of corrections.
A last point is that generally comes from an aerothermal analysis (performed with a CFD software) at
discrete times and are calculated on a mesh which does not match with the thermal mesh. The analysis provides as
well other parameters such as , , . For 3-D analysis, the proper angle will be
specific enthalpy of reaction for the interaction between the
surface material and the family of particles
.
empiric correction for the effect of surface temperature (without
unit)
H
r p j , ,
H
r p j , ,
T
s
J , ( ) =
j J. kg
1
[ ] ( )
f
t h p ,
f
t h p ,
T
s
( ) =
j
V
p j ,

p j ,
D
p j ,
j G
t h p j , ,
V
p j ,
D
p j ,

p j ,
, , ( )
m
s t h p j , , ,
x t , ( ) G
t h p j , ,
V
p j ,
x t , ( ) D
p j ,

p j ,
x t , ( ) , , ( )m
p j ,
x t , ( ) =
m
s t h p , ,
j
j H
r p j , ,
T
s
p j
j
f
1
T
s
( ) 0 if T
s
T
1

T
s
T
1
T
2
T
1
------------------- if T
1
T
s
T
2
1 if T
2
T
s
< ; < < ; <
)
`

=
f
2
T
s
( )
1
2
--- h
T
s
T
0
T
-------------------
\ .
| |
1 + tan =
T
0
1
2
--- T
1
T
2
+ ( ) = T n T
2
T
1
( ) =
m
p j ,
x t , ( )
V
p j ,
x t , ( ) D
p j ,

p j ,
x t , ( )
p j ,
x t , ( )

calculated within the CFD software (angle between the normal direction at the surface and the velocity vector, in the
common plan defined by these two vectors).
Thermal Effects of Particle Impacts:
with:
Liquid Phases Removal:

This term takes into account possible liquid phases appearing at the surface of the ablating material, by several
mechanisms (phase change of a component of the material [silica, ], oxidation at the surface of gaseous products
coming from the inside forming liquid phase , for instance), and the blowing of these phases (gravity,
shear stress, ). Moreover, this kind of phenomena would require a more complex modeling, for example with the
explicit modeling of the liquid film running at the surface. Finally, there could exist a liquid film at the surface
produced by other mechanisms such as impacts of many liquid particles.
These quantities are tabulated, and a user-subroutine can be used to provide this flux as a whole.
Radiative Balance:
This is defined through the FILMS or RAD-CAVITY option.
Mass Conservation Equation:
Mass flow blown (injected in the outflow) + mass flow of removed liquid phases = mass flow of internally produced
gaseous products + mass flow of the gaseous products of thermochemical ablation reactions with external gases + mass
flow of gaseous products of thermochemical ablation reactions with impacting particles.

This term takes into account a possible energy deposit by the
particles impacting the surface, besides the phenomena of
thermochemical ablation and mechanical erosion. For
instance, it could be due to a partial conversion into heat of the
kinetic energy of the particles, during inelastic collisions.
There does not seem to exist any general accepted law or
formulation for this term. This effect is even neglected in
many ablation models. So this term can only be introduced by
a UTIMP user-subroutine.
surface mass flow rate of liquid products formed at the surface of the material, and blowed out
by the action of the external flow or other mechanisms .
specific enthalpy of the liquid phase 'l' taken into account in, evaluated at the surface temperature
.
p k ,
p k ,

p k ,
m
p j ,
x t , ( )V
p j ,
x t , ( )D
p j ,

p j ,
x t , ( ) , ( ) =
f
t h p ,
m
1
H
1
m
1
1
x t , ( ) =
kg. m
2
. s
1
[ ] ( )
H
1
H
1
T
s
( ) =
T
s
J. kg
1
[ ] ( )
Si O
g
Si O
2
[ ]
r
ab
r
em
v ( )
w
m
1
1
+ m
s t h g , ,
m
g
m
s t h p , ,
+ + m
s t h g , ,
m
g
f
t h g ,
G
t h p j , ,
m
p j ,
+ + = =
237
CHAPTER 6
Velocity of Surface Recession Due to Thermochemical Ablation:
The surface recession due to thermochemical ablation (and by extension due to other possible 'physical-chemical'
mechanisms such as phase change or blowing of liquid phase) is only one part of the total surface recession . This
part is calculated from the SEB. To obtain the total surface recession, it must be added the contribution of mechanical
erosion :
The contribution must be calculated by other means.
Velocity of surface thermochemical recession velocity of chemical thermochemical ablation by gases + velocity
of thermochemical ablation by impacting particles (unit: m s-1)
with
is the density of the solid, when pyrolysis is included through the THERMO-PORE option, this is the current
density.
Mechanical Erosion
The mechanical erosion is currently considered to be due both to impacts of particles, and to other external actions
such as the shear stress of the flow, the vibrations of the part,
Mechanical Erosion by Particles
The expression is similar to the one for thermochemical ablation by particles described above:
with here a mechanical 'G-Law' , and a similar temperature correction . These two quantities may be
defined through the UGMEC and UFMEC user subroutines, respectively.
Alternatively, a UTIMP user-subroutine is provided for calculating .
unit:
without unit
without unit
S
t h
S
mec
S
t h
S
mec
+ =
S
mec
S
t h
S
t h
m
s mec p , ,
m
s h p , ,
+ [ ]
s
=
m
s t h p , ,
f
t h p ,
G
t h p j , ,
m
p j ,
=
s
S
mec
m
s mec p , ,
m
s mec ot h , ,
+ [ ]
s
=
m
s mec p , ,
f
mec p ,
G
mec p j , ,
m
p j ,
=
m
s mec p , ,
kg m 2 s 1
f
mec p ,
G
mec p j , ,
G
mec p j , ,
f
mec p ,
m
s mec p , ,

Mechanical Erosion by Other Actions
There is no generally accepted expression for. And there are many possibilities depending on the kind of material and
the environment. The UFLUXMEC user-subroutine is provided for calculating .
Pyrolysis
Pyrolysis involves the decomposition of materials due to thermal processes. There are two aspects of these
simulations: the thermal-chemical decomposition of the material and the creation of gas and transport of this gas. The
creation of the gas is based upon the conservation of mass. The material models used in these simulations have
increasing levels of complexity. There are two models used to simulate the transport of the gas known as the streamline
model and the Darcy fluid model. The choice of these two models is made on the PYROLYSIS parameter.
In the streamline model, the gas moves along streamlines that are aligned with the mesh, as shown in Figure 6-5. This
procedure is available for planar, axisymmetric and solid elements using either linear or quadratic interpolation
functions. But, because the need for regular meshes, triangular and tetrahedral elements are not available. The user
must identify which element contain streamlines.
The streamline method introduces two new terms: streamlines and streamline integration points (SIP). For lower order
elements, the SIP are at the intersection of the streamlines and the element edges. For higher order elements, an
additional SIP exists midway between the two. The pyrolysis calculation occurs at the SIP. An extrapolation and
interpolation procedure is used to move data from the SIP to conventional integration points (CIP). The pyrolysis gas
moves from SIP
i
to SIP
i+1
. If the temperature becomes hotter in the interior, such that flow of gases would reverse,
an error would occur.
When using the Darcy law model, any solid heat transfer element may be used. When this method is used, the pyrolysis
calculation occurs at the CIP. The pyrolysis gas may move in any direction.
Figure 6-5 Streamlines
m
s mec ot h , ,
m
g
node
SFIP
SIP
i+1
SIP
i
CIP
Regular Mesh
SIP Streamline Integration Point
CIP Conventional Integration Point
SFIP Surface Integration Point
239
CHAPTER 6
Presentation of the Mass Equation
The conservation of mass equation for the streamline model yields:
where the first term is the conventional generation (source) term due to a change in density of the solid, the second
term represent the change in density of initial water vapor, and the final term is due to change in density due to coking.
The drying of the initial water vapor occurs before the pyrolysis occurs so:
The global mass equation is defined as the sum of these. The global mass flow rate is equal to .
Principle of Pyrolysis
When pyrolysis occurs, the solid material is degraded due to the effect of the temperature. This degradation is done in
steps (Figure 6-6), according to the temperature. We represent these steps on the drawing below. With each step a loss
of density corresponds: .
mass flow rate of the gases of decomposition .
mass density of the pyrolysis gas.
mass density of the solid undergoing pyrolysis.
mass density of the liquid vapor.
mass density of the coked solid.
velocity of the pyrolysis gas.
source term of decomposition.
source term of carbon deposit.
source term of water drying.
. m
s p* ,
t
---------------

l
t
--------

s c c* , ,
t
------------------- =
m
g
=
g
s p* ,
s c c* , ,
s p* ,
t
---------------
s c c* , ,
t
-------------------
l
t
--------
. m
g w ,
l
t
-------- =
m
g
m
g p ,
m
g w ,
+
s p j , ,

Figure 6-6 Degradation Steps
The density of material thus varies from , density of virgin material to , the density of charred (entirely
pyrolyzed material). If coking occurs the density increases because carbon is redeposited in the solid
We define a dimensionless variable which varies from 1 to 0 during pyrolysis (it represents the advancement of
pyrolysis).
It is calculated by the law of Arrhenius:
.
The coefficients , , and are determined by thermogravimetry (TGA). The material properties associated
with each phase (virgin, charred, coked, liquid, gas) are entered through the ISOTROPIC or ORTHOTROPIC options.
The THERMO-PORE option is used to reference these phases and provide additional material data including the terms
of the Arrhenius series.
This variable facilitates the calculation of the variation in solid density due to pyrolysis, , which is the sum
of the preceding

law of Arrhenius:
or
s p 3 , ,
s p v , ,
s p 1 , ,
s p c , ,
Temperature
s p v , ,

s p c , ,
j
t
-------- B
j
T
a j ,
T
s
------------
\ .
| |
j
exp =
B
j
T
a j ,

j
s p* ,
t
---------------
s p* ,
t
---------------
( )
s p j , ,
j
t
--------
j 1 =
Nd
( )
s p j , ,
B
j
T
a j ,
T
s
------------
\ .
| |
j
exp
j 1
Nd
= =
s p* ,
t
---------------
( )
s p j , ,
j
t
--------
j 1 =
Nd
( )
s p j , ,
B
j
E
a j ,
R
*
T
------------
\ .
| |
j
exp
j 1 =
Nd
= =
241
CHAPTER 6
where
It is assumed that at a point thermal equilibrium occurs and the temperature of the solid and the gas are the same.
Alternative Formulation
We see in the literature an alternative writing of the Arrhenius law know as the rho law:
where:
The user can enter the input for Arrhenius law for the both laws. If he chooses the rho law, then he must give:
. If he chooses the law, then he must give: .
One can define an alternate model for pyrolysis using the UPYROLSL user subroutine.
Coking
The processes considered when the material is heated are very complex. For instance, we can consider the primary
pyrolysis chemical reactions producing gases, and the subsequent secondary chemical reactions, between gases,
between gases and solid, and between constituents in the solid phase. We focus here on one class of secondary
chemical reactions, namely the heterogeneous reactions between gases and porous solid leading to a carbon deposit,
known as coking phenomenon.
is the number of terms in the Arrhenius series.
is the temperature of activation of the reactions of decomposition
is the order of the reaction
is the dimensionless variable of decomposition varying from 1 to 0 during pyrolysis
is a factor to the exponential
is the energy of activation of the reaction j of decomposition,
is the representation of the reduction of density (counted in positive quantity)
is the local temperature of the solid phase.
is the perfect gas constant.
is the virgin material density.
is the charred material density.
is the current material density.
Nd
T
a j ,
j
B
j
E
a j ,
( )
s p j , ,
T
s
R
s p ,
t
------------
j
B'
j
T
a j ,
T
---------
\ .
| |
s v j , ,
s p j , ,

s c j , ,
s v j , ,
--------------------------------
\ .
|
| |

j
exp
j 1 =
Nd
s v j , ,
j
s c j , ,
j
s p j , ,
j
Nd
j
B'
j
E
a j ,

j

s v j , ,

s c j , ,
, , , , ,
j
Nd
s p j , ,
B
j
E
a j ,

j
, , , ,

The processes mentioned above are governed by complex sets of kinetically controlled chemical reactions, which are
neither well known, nor the constants associated with them.
One Term Arrhenius Model
In this model, we consider the total mass fraction of atomic element carbon in the gaseous phase, independently
of the molecular configuration (chemical specie) in which it is present. The initial value , corresponding to the
gases generated by the initial primary pyrolysis chemical reactions, is data that can be determined experimentally by
calculating the difference in carbon content between the virgin material and the one of the char before coking. We
should note that is the mass flow rate of atomic element carbon independently of the molecular configuration.
This model is inspired by a paper written by R.A. Rindal [Ref. 20].
During the subsequent process of secondary gas/gas chemical reactions, this quantity does not change, because only
the molecular configurations are changing in the gaseous phase, but not the content in any atomic element:
.
It is known that the gaseous phase at lower temperature contains more carbon than it would contain in the case of
chemical equilibrium. When the gases flow towards the heated surface, the temperature is rising; the rate of reactions
is rising too, and the composition of gases approaches the one corresponding to chemical equilibrium. It results in a
deposit of solid carbon (coking), and then the quantity is changing in the zone where coking occurs.
From the considerations above, it seems reasonable to assume that the potential of coking is driven by the difference
between the local value and the value of at chemical equilibrium, and that the rate with which chemical
equilibrium is approached is of Arrhenius kind. Furthermore, the homogenous chemical reactions in the gaseous phase
are generally not equimolar, so that a pressure dependence and an order of reaction different from the unity are
assumed. So, the following kinetic equation for the evolution of the quantity is proposed:
with:
The quantity can be calculated using standard thermochemical calculation software.
:
kinetic coefficients for coking reactions.
:
total mass fraction of the atomic element carbon in the pyrolysis gas.
:
total mass fraction of the atomic element carbon in the pyrolysis gas when chemical equilibrium
is achieved.
K
cg
K
cg0
K
cg
m
g
K
cg
K
cg0
=
K
cg
K
cg
K
cgE
K
cg
K
cg
t
------------- k
c
E
ac
RT
------------
\ .
| |
P ( )
n
c
K
cg
K
cgE
( )
n
c
exp =
k
c
E
ac
n
c
, ,
K
cg
K
cgE
K
cgE
243
CHAPTER 6
The streamline model does not give access to the value of the local pressure . There are two assumptions that can be
made:
either neglect the pressure dependence, by setting equals to 1
assume that is equal to the external pressure.
In the current implementation the pressure
When using the Darcy law model, the pressure to be considered will be the calculated pore pressure.
Remarks
The quantity will always be between and , and its variation rate will approach zero when
becomes close to .
Based upon the assumption made in this analysis that the coking only begins when the material is completely
pyrolysed, we control the coking area by the advancement of the pyrolysis. Coking occurs if the advancement variable
of pyrolysis at this location is bigger than a critical value, named pyrmax. It is usually just smaller than 1., which
means that pyrolysis is over. This may be defined through the THERMO-PORE option; the default is 0.96.
Beside this Arrhenius model to calculate the mass fraction of carbon in the pyrolysis gas, a linear model is proposed.
Linear Model
This model determines the mass fraction of carbon in the pyrolysis gas as a function of the temperature. The evolution
is linear from a low temperature where the pyrolysis gas has an excess of carbon to a higher temperature where it is in
equilibrium:
1. For temperatures below the low temperature , there is no coking: .
2. For temperatures between the low temperature and the equilibrium one:
If the temperature is increasing, is a linear function of temperature. ((2))
If the temperature is decreasing, the coking is stopped and retains its last value. This forbids the
inverse coking reaction that would lead to an increase of carbon in the pyrolysis gas.
For temperatures larger than the equilibrium temperature, there is no more chemical reaction:
. Figure 6-7 illustrates this:
The user can also specify a coking model through the UCOKSL user subroutine.
P
P
P
P ( ) 1.0 =
K
cg
K
cg0
K
cgE
K
cg
K
cgE
T
l
K
cg
K
cg
T
l
( ) =
K
cg
K
cg
K
cg
K
cgE
=

Figure 6-7 Calculation of Mass Fraction of Carbon in Pyrolysis Gas by Linear Function of Temperature
Water Drying Model
The water drying model chosen for the streamline model is based upon the same principle than the pyrolysis one,
namely an Arrhenius law.
Written in function of the liquid mass density variable , the Arrhenius law for water drying is:
If we set , we have:
,
So the temporal gradient of is:
K
cg
K
cgE
T
low
T
high
=T
E
K
cg
(T
l
)=
0
~
cg
K
Temperature
l
t
-------- B
w
*
E
w
R*T
-----------
\ .
| |
*
l 0 ,
l 0 ,
-------- -
\ .
|
| |

w
exp =
w
1

l 0 ,
--------- =
l 0 ,
1
w
( )
t
------------------------- B*
E
w
R*T
-----------
\ .
| |
*
l 0 ,
1
w
( )
w
exp =
w
t
---------- B
w
*
E
w
R*T
-----------
\ .
| |
* 1
w
( )
w
exp =
245
CHAPTER 6
where:
Moreover, we have:
where is the mass density of liquid at the beginning of the analysis.
The user can define its own law to calculate the term via a UWATERSL user subroutine.
Presentation of the Energy Equation
To take into account pyrolysis and coking and water evaporation in the calculation of the temperature, a convective
and three enthalpic terms are added to the standard equation. The phenomena is included in:
the convective term by the mass flow rate . The latter decreases during coking because of the mass loss of
gaseous carbon to solid carbon.
an enthalpic term that represents the energy consumed by the pyrolysis reaction.
an enthalpic term that represents the energy consumed by the coking reaction.
an enthalpic term that represents the energy consumed by the evaporation of the water vapor.
is defined by , and represents the drying state. It varies from 0 to 1 during evaporation,
is a pre-exponential factor,
is the energy of activation of the reaction of water drying,
is the order of the reaction,
is the local temperature of the solid phase.
is the perfect gas constant.
w

w
1

1 0 ,
---------- =
B
w
E
w
w
T
R
l
t
--------
l 0 ,
*
w
t
---------- =
l 0 ,
l
t
--------
m
s p* ,
t
--------------- H
g p ,
H
s p vc , ,
( )
s c c* , ,
t
------------------- H
c*
H
g p ,
( )
l
t
-------- H
v
H
l
( )

The equation becomes:
where:
The effective specific heat is based upon a mixture law between heat capacity of pyrolysis gas and
the one of vapor :
We have
And so
To take into account the effect of water drying in the material properties, we use the drying state defined above
by .
The effective heat capacity is defined by:
.
The effective conductivity
The heat capacity of the virgin material, charred material, deposit by coking effect and the liquid part
is effective heat capacity of the material during the analysis.
is the effective specific heat of the gas (pyrolysis and vapor).
is the effective conductivity of the material during the analysis.
is the enthalpy of the gas, including only gas of decomposition.
is the enthalpy of the solid in course of pyrolysis (no liquid influence).
is the enthalpy of the water vapor.
is the enthalpy of the liquid water.
s i * ,
c
pi

s p* ,
c
s p* ,

s c c* , ,
c
c*
+ + ( )
T
t
------- c
pg p ,
m
g
. T +
. *T ( )

s p* ,
t
--------------- H
g p ,
H
s p vc , ,
( )

s c c* , ,
t
------------------- H
c*
H
g p ,
( )

l
t
-------- H
v
H
l
( ) + +
=
c
p
( )
ef f
c
p g ,
*
H
g p ,
H
s p vc , ,
H
v
H
l
c
p g ,

g p ,
c
p g p , ,
( )
g w ,
c
p g w , ,
g
c
p g ,

g p ,
c
p g p , ,

g w ,
c
p g w , ,
+ =
c
p g ,
m
g p ,
c
p g p , ,
m
g w ,
c
p g w , ,
+
m
g
------------------------------------------------------------ =
w
1

l 0 ,
--------- =
c
p
( )
ef f
1
p
( )
c
p
( )
v

p
1
c
( )
c
p
( )
c

p
c
p
( )
cd
1
w
( )
l 0 ,
c
p l ,
+ + + =
* 1
p
( )
v

p
1
c
( )
c

p
cd
+ + =
c
p
( )
v

c
p
( )
c

c
p
( )
cd

l 0 ,
c
p l ,
( ) , , ,
247
CHAPTER 6
The conductivity of the virgin material, charred material, deposit by coking effect
We can note that:
the vapor is not taking into account, which is consistent with the fact that we neglect also the effect of the
pyrolysis gas,
the conductivity does not depend on the liquid water. Indeed, we can consider that the liquid belongs to small
porosity, and that water evaporates before the beginning of the pyrolysis.
We have, respectively:
the rate of pyrolysis, coking
the drying state
The enthalpy of material in the course of pyrolysis function of the densities and enthalpies of virgin and charred
material: .
The enthalpies, written in upper case, are the absolute enthalpies. They are equal to:
.
Ablation
In many mechanical and thermal processes, there is the loss or removal of surface material by an erosive process such
as melting or vaporization. Mechanical processes, such as particle impact or friction, may be involved as well. This
process is particularly significant in high temperature applications, such as spacecraft reentry, where the intentional
removal of material contributes to the thermodynamic cooling. For high temperature applications, special ablative
materials have been created which may be defined using the THERMO-PORE option. In the current release, ablation
has been designed to be used in thermal analyses only. The capability exists for 2-D, axisymmetric and 3-D solid
elements, but the 3-D capability is in pre-release stage.
is the enthalpy of formation, is an input of the program.
is the temperature of reference for enthalpy of formation, is an input of the program.
is the sensible enthalpy.
v

c

cd
, ,
p

c
,
c
t
--------
1
c d* ,

s p c* , ,
-----------------------------------
s c c* , ,
t
------------------- =
H
s p vc , ,
s p v* , ,
H
s p v* , ,

s p c* , ,
H
s p c* , ,
s p v* , ,

s p c* , ,
--------------------------------------------------------------------------- =
H H
f
0
h + H
f
0
c
p
T d
T
0
T
}
+ = =
H
f
0
T
0
h

There are seven aspects that control the receding surface calculation in Marc.
1. Specifying which surfaces are to undergo recession.
The RECEDING SURFACE option is used to specify which surface is to be subjected to recession. For nodes
that reside on elements edges (2-D) of faces (3-D) specified by the RECEDING SURFACE, the incremental
displacement will be calculated during the recovery phase.
2. Specifying the magnitude of the recession rate.
The recession rate is based upon either:
a. Simple model, that is either a constant, or evaluated by a table.
b. Based upon the data given in the SURFACE ENERGY model definition block.
c. Specified in UABLATE user subroutine.
3. Specifying the direction of the movement of the nodes on the surface.
The direction of the recession is based upon the normal evaluated at the node. If the ABLATION,1 parameter is
used, then the normal is based upon applying a unit distributed load to all external surfaces. The magnitude of
the equivalent force is the effective area, and the direction is the normal. This procedure is based upon lower
order elements. The result is that at corners of the mesh the calculated normal is not perpendicular to either
surface, but a weighted average (based upon the area of the edges).
This procedure may lead to more physically reasonable results, as corners tend to be rounded off, but leads to
more meshing problems.
If the ABLATION,2 parameter is used, then the normal is based upon applying a unit distributed load to those
element edges/faces specified by the ABLATION option. This results in the movement of the node lying at the
corner to be perpendicular to the surface.
If the ABLATION,3 parameter is used, then the direction of recession is based upon the direction of the
streamline, assuming that the streamline method is specified on the PYROLYSIS parameter.
If the ABLATION,4 parameter is used, then the direction of recession is based upon the direction of the
streamline, assuming that the streamline method is specified on the PYROLYSIS parameter. The magnitude of
the recession is adjusted based upon the difference between the true normal and this direction.
4. Specifying when remeshing is to occur.
Remeshing is based upon three criteria specified via the ADAPT GLOBAL option. The tolerance values are
relevant, but not critical unless the recession rate is very high, the time step is large, or the elements are small.
The two key issues are:
a. The element should not be so distorted at the beginning of the increment that the recession during the
increment would lead to an inside-out element. This may lead to an Exit 1005.
b. The ablation criteria should only find at most one element along each streamline that should be shaved off.
This may lead to an Exit 1111.
249
CHAPTER 6
With this in mind, the following remeshing criteria have been developed:
a. Element size reduction two ratios are calculated. The first is the ratio of the current element edge length
divided by the original element edge length. The second ratio is the ratio of the projected current element
edge length with the surface normal divided by the projected original element edge length with the surface
normal. For a 2-D quadrilateral mesh for each element on the receding surface, two edges are checked; while
for 3-D hexahedral mesh, four edges are checked. If any edge satisfies the criteria, remeshing occurs. The
recommended value of the tolerance is 0.4.
b. Motion of a node. The magnitude of the displacement since the last remeshing is compared with the user-
specified tolerance. This method is not recommended.
c. The program also uses the previous displacements to predict if in the next increment the element will go
inside out. As long as the next increment is similar to the previous increment this is a reasonable procedure.
If may fail if the either the recession rate suddenly increases, or if the time step is increased. The user does
not need to provide any tolerance value. While in theory, this method could be used exclusively, without the
ones mentioned above, this is not recommended, because it leads to very thin elements. Two procedures are
used to determine if an element has gone inside out. The first is based upon the Jacobian of the isoparametric
shape functions. This method was determined not to be conservative. Hence additionally the area of four
triangles (2-D) or volume of eight tetrahedrals is used.
5. Specifying the mesher used to create a new mesh
Currently the user needs to specify whether the shaver mesher, relax mesher, or stretch mesher is to be used.
These meshers may be used with quadrilateral and hexahedral elements.
a. The shaver (type 9) is advantageous because rezoning is required in only a smaller region of the mesh.
Elements are only modified if they have failed any of the criteria specified above, or they are adjacent to a
modified element. The shaved mesher is also advantageous because the original biased mesh is retained
during the process. The shaver mesher results in degenerated quadrilateral elements (triangles), or
degenerated hexahedral elements (wedges), which may not be esthetically pleasing.
b. The relax mesher modifies ever element in a body, as soon as any one of them fails the mesh criteria
specified above. The relax mesher itself is simpler than the shaver mesher. The relax mesher also requires
the use of the SPLINE option to identify edges that are to be fixed.
c. The stretch mesher modifies ever element in a body, as soon as any one of them fails the mesh criteria
specified above. The stretch mesher is similar to the relax mesher, but it only works in one direction, and it
reasonably assures that all elements have a uniform thickness in this direction.
6. Which element type is being used and control of midside nodes
Either mesher may be used with either lower order or higher order elements for both 2-D and 3-D problems.
When using higher order elements, two considerations are made.
a. During the recession itself the midside node on the surface is positioned such that the surface geometry is
linear within and element, bi-linear for 3-D geometries. While this is not strictly consistent with the idea of
higher order elements, it resulted in less distortion problems. Temperature fields are still quadratic. The
midside nodes on the edges perpendicular to the surface that are in the interior of the element are adjusted
such that they are mid-edge. This was done to make sure the quadratic shape functions did not break down.
b. The remeshing process itself is based upon using lower order elements, and the midside nodes are then
repositioned to the midpoints.

7. Region geometry.
For the streamline model the element edges normal to the surface in theory should follow the streamline
behavior; i.e., be normal to the surface. This is also advantageous from a surface recession perspective. When
the recession, which by default is normal to the surface, is not aligned with the element edges, addition mesh
distortion occurs. To overcome this problem a projection procedure can be requested. This is only available in
conjunction with the shaver mesher, and is still not as good as creating a good mesh to begin with. It is highly
advantageous to have initially a good mesh.
Examples
1. Axisymmetric analysis with edges not perpendicular to surface). This is an example of a nonoptimal meshing
technique because the element edges are not perpendicular to the surface.
Figure 6-8 Original Mesh Note vertical edges are not normal to the surface
Figure 6-9 Time=44.0 seconds
251
CHAPTER 6
Figure 6-10 Close Up Showing Effect of Edges Not Aligned Perpendicular to Surface
2. Axisymmetric analysis of a nozzle using four node element and shaver mesher. Here the mesh is well aligned
with the geometry, which leads to the ability to ablate further, and improves the remeshing.
Figure 6-11 Initial Mesh
Figure 6-12 Time = 220. seconds

Figure 6-13 Close Up at Leading Edge
Figure 6-14 Final Temperatures
3. Example of nozzle using 2-D quadratic elements tolerance = 0.2
Figure 6-15 Initial Mesh, Showing Nodes
253
CHAPTER 6
Figure 6-16 Time = 220 Seconds
Figure 6-17 is a close-up showing nodes, including those shaved off. Note that outer edges remain straight, and
that the midside nodes for the elements on the receding surface also remain at the midside of the edges. The
element distortion is due to the fact that the element edges are not aligned perpendicular to the surface.
Figure 6-17 Time = 220 Seconds

4. Example of 3-D nozzle.
Figure 6-18 Initial Mesh has 6480 Elements, 7657 8-node Bricks
Figure 6-19 Time = 44.0 seconds
255
CHAPTER 6
Figure 6-20 Close Up of Leading Edge
5. Example using linear elements and relax mesher.
Figure 6-21 Final Configuration
Note that the number of elements using the relaxed mesher does not change so elements become thinner during
the ablation process.

6. Example 3-D Linear hexahedral elements Relax mesher extruded nozzle.
Figure 6-22 Final Configuration - Linear 3-D Elements
7. Example 3-D Quadratic hexahedral elements Relax mesher extruded nozzle.
Figure 6-23 Time = 44.0 Seconds
257
CHAPTER 6
Note that in Figure 6-24, the outer edges remain straight and that the midside nodes for the elements on the
receding surface also remain at the midside of the edges.
Figure 6-24 Close Up of Leading Edge Showing nodes
Welding
Welding is a thermal process with specialized boundary conditions. These boundary conditions are specified through
the WELD FLUX model definition option in conjunction with the WELD PATH and WELD FILL model definition
options. Using these options, the welding heat input can be specified through two different techniques:
Modeling the heat input as a spatially varying distributed flux through the WELD FLUX model definition
option applied to the base metal and filler elements. No temperatures are specified on the WELD FILL model
definition option in this case.
Modeling the heat input as a spatially varying temperature boundary condition through the WELD FILL model
definition option applied to the nodes of filler elements. No flux value (zero power) is specified on the WELD
FLUX model definition option in this case.
It is also possible to combine a nonzero flux applied to base metal elements through the WELD FLUX model definition
option with a nonzero temperature applied to filler element nodes through the WELD FILL model definition option. In
all these cases, the heat input path for the welding heat source or for the filler element deposition is defined by the
WELD PATH model definition option.
Weld Flux
A disc shaped surface/edge heating source or a double ellipsoidal shaped volume heating source can be specified using
the WELD FLUX model definition option. For more complex shapes, the heating source can be specified through the
UWELDFLUX user subroutine. The associated weld path (defined through the WELD PATH model definition option)
and the associated filler element set (defined through the WELD FILL model definition option) are also referred to
through the WELD FLUX model definition option.

Pavelics disc shaped weld heat source is expressed as
[Ref. 16] (6-14)
where is the heat flux rate per unit area; is the applied power with = efficiency, = Applied Voltage
and = Applied Current; is the radius of the disc; is the local coordinate along the weld path and is the local
coordinate along the tangent to the weld path. The disc shaped model can be used for specifying edge fluxes in 2-D
and surface fluxes in 3-D. It is particularly useful for situations where the depth of weld penetration is not significant.
Goldaks double ellipsoidal shaped weld heat source can be used to specify volume fluxes in 2-D and 3-D [Ref. 16]:
(6-15)
where and are the weld flux rates per unit volume in the front and rear weld pools respectively; is
the applied power; is the weld width along the tangent direction x; is the weld penetration depth along the arc
direction y; and are the forward and rear weld pool lengths in the weld path direction z; and are
dimensionless factors given by
(6-16)
The double ellipsoidal source is also shown below in Figure 6-25.
Figure 6-25 Double Ellipsoidal Weld Flux showing the Local Coordinate Dimensions
The weld pool dimensions in Equations (6-14) and (6-15) are to be estimated and provided by the user. The width,
depth, forward length and rear length are used for the volume weld flux. The surface radius is used for the edge or face
q x y z , , ( )
3Q
r
2
--------
3x
2
r
2
------------
\ .
| |
3z
2
( )
r
2
----------------- exp exp =
q Q VI = V
I r z x
q
f
x y z , , ( )
6 3f
f
Q
abc
f

-----------------------
3x
2
a
2
------------
\ .
| |
3y
2
b
2
------------
\ .
| |
3z
2
c
2
------------
\ .
| |
exp exp exp =
q
r
x y z , , ( )
6 3f
r
Q
abc
r

-----------------------
3x
2
a
2
------------
\ .
| |
3y
2
b
2
------------
\ .
| |
3z
2
c
2
------------
\ .
| |
exp exp exp =
q
f
q
r
Q VI =
a b
c
f
c
r
f
f
f
r
f
f
2
1 c
r
c
f
+ ( )
----------------------------- =
f
r
2
1 c
f
c
r
+ ( )
----------------------------- =
z
c
f
x
y
a
b
c
r
259
CHAPTER 6
weld flux. The dimensions are mandatory when the standard disc or ellipsoidal models are used and are optional when
the UWELDFLUX user subroutine is used. The weld pool dimensions are used in three different ways by the program:
for defining the weld flux rate,
for defining a bounding box for filler element detection, and
for defining box dimensions when a box adaptive meshing criterion is used.
The weld pool dimensions can be optionally varied with the help of tables. The tables can be functions of time or arc
length. The arc length is measured along the associated weld path, from the first point of the path to the current position
of the weld flux.
The weld flux rate in Equations (6-14) and (6-15) can be further scaled by a factor. This scale factor can be
automatically calculated by the program or optionally set by the user. For 3-D problems, the automated scale factor is
set to 1. For 2-D problems (both planar and axisymmetric), the automated scale factor is calculated by equating the
integral of the flux rate over the entire x-y plane and the out-of plane thickness to the applied power .
For Equations (6-14) and (6-15), the scale factor can be shown to be
(6-17)
It should be noted that the automated scale factor only ensures that the applied fluxes in the 2-D case matches the 3-
D case. In general, the temperature solution comparisons between 2-D and 3-D also depend on the velocity of the weld
source, the thermal properties, other thermal and mechanical boundary conditions, and heating effects before and after
the heat source has crossed the 2-D plane.
Weld Path
The path followed by the welding heat source and the orientation of the weld arc are specified by the WELD PATH
model definition option. This information is used to determine a local coordinate system for the moving source, as
shown in Figure 6-26. The weld path vector is taken as the local z axis. The arc orientation vector is taken as the local
y axis. The tangent vector that is perpendicular to both y and z axes is taken as the local x axis.
It is possible to specify the weld path and orientation information through the Marc Input, Text Input or the
UWELDPATH user subroutine.
s
t Q
s q x y 0 , , ( )tdxdy
} }
Q =
s

3
---
r
t
- =
s

3
---
c
r
c
f
+ ( )
2t
--------------------- =

Figure 6-26 Local Coordinate System for Moving Source obtained by Translation and Rotation of Global Coordinate
System
Marc Input
Two options are available for specifying the weld path Ordered List of Nodes or Point Coordinates of Polyline
Curves. The weld path vector at point N is defined as the vector from point N to point N+1. For a closed weld path,
the first and last node should be identical, or, the first and last point should have the same coordinates. An update path
option is provided for the nodes option for a coupled analysis which allows the path to be updated based on the current
displacements of the nodes along the path. This option is used in conjunction with the FOLLOWER FOR parameter.
When the latter flag is used, all distributed fluxes are evaluated on the updated geometry and the path followed by the
weld source is also updated in this case.
A variety of options are available for specifying the arc orientation: Ordered List of Nodes, Point coordinates of
Polyline Curves, Vector Components, or Euler Angles. When nodes or points are used, the arc vector at point N is
defined as the vector from the weld path point at N to the arc orientation point at N. It is further possible to rotate the
arc vector through an angle specified in degrees about the weld path axis. When vector components are used, they can
be specified at distinct points and can be optionally interpolated as a function of arc length (length of the path from
the first point) or of position coordinates along the path. When Euler angles are used, a unit vector in the global X
direction is rotated through the global X-axis, Y-axis, and Z-axis by the specified angles to yield the arc orientation
vector. The rotation angle, arc orientation vectors or Euler angles can be further varied along the weld path using tables
as a function of arc length or point coordinates. For a point that is in between two given path points, Marc can
automatically interpolate to find out the current arc orientation.
Text Input
In many industrial applications, weld paths for robots are specified via text files. A generic text file capability is
provided in Marc in order to allow the specification of the weld path and the weld orientation. The weld path is
specified through point coordinates in the columns 1-3 of the text file. The weld path vector at point N is defined as
the vector from point N to point N+1. The arc orientation, defined in columns 4 - 6 of the text file for each point N,
can be defined via point coordinates, vector components or Euler angles. The entry in each column is a real number
of width 10. The columns can be in free or fixed format, with commas being used to separate the columns in the free
format mode.
X
Y
Z
y (weld orientation)
x (weld width)
(weld path) z
261
CHAPTER 6
User Routine Input
The weld path can also be defined using the UWELDPATH user subroutine. The final position of the associated weld
flux, the weld vector at that point, and the orientation vector at that point have to be defined in the subroutine. Refer
to Marc Volume D: User Subroutines and Special Routines for more details.
In some cases, it is possible that the given/computed local y axis is not perpendicular to the local z axis. It is however
important for purpose of weld flux computation to have a set of mutually perpendicular local axes. Since weld flux
dimensions are typically provided in directions along and perpendicular to the weld path, the weld path vector (z axis)
is taken as fixed and an equivalent set of perpendicular x and y axes are computed from the given information.
The actual position of a weld heat source can be offset by a user specified amount from the associated weld path. The
offsets are specified in the local x and y directions. This offset allows the same weld path to be used for multiple weld
heat sources and is particularly useful for multi-pass welding where the weld heat source repeatedly moves in a similar
pattern, but offset from the original path by a small amount.
Weld Filler
The dynamic creation of filler elements and the generation of associated boundary conditions is defined by the WELD
FILL model definition option. Two methods are supported for the creation of filler elements:
the quiet element method and
the deactivated element method.
In the quiet element method, filler elements are initially used with scaled down material properties. The scale factor
is 10
-5
by default. The coefficient of thermal expansion is set to zero and all other material properties (except yield
stress, specific heat, and thermal mass density) are scaled down. The scaling is applied to reference temperature
material properties and temperature dependencies are ignored. When temperature dependent material properties have
been entered through tables, the reference temperature properties are evaluated at the lowest temperature (i.e, the first
table value). Weld fluxes and other point/distributed thermal loads on the quiet elements are ignored. When the filler
elements are physically created by the moving heat source, the thermal properties are fully restored while the filler
elements are still quiet on the mechanical side. If the thermal activation time is zero, the mechanical properties are also
fully restored in the next increment. If the thermal activation time is nonzero, the filler elements remain mechanically
quiet till the thermal activation time duration is passed. During this period, all stresses and strains in the elements are
reset to zero. This disjoint thermal and mechanical treatment of the weld filler prevents the build up of mechanical
strains and stresses during the temperature ramp up period. The quiet element method allows the elements to move
with the model. This is particularly useful for accommodating large deformations during the welding process.
However, the quiet element method is prone to ill-conditioning due to the large discrepancy in material stiffnesses.
In the deactivated element method, filler elements are initially deactivated in the analysis and are not shown on the
post file. When the elements are physically created by the moving heat source, they are activated in the model and
appear on the post file. The filler elements are only thermally activated initially and remain inactive on the mechanical
side. If the thermal activation time is zero, the filler elements are activated on the mechanical side in the next
increment. If the thermal activation time is nonzero, the filler elements remain mechanically inactive until the thermal
activation time is passed. This disjoint thermal and mechanical activation of the weld filler prevents the build up of
mechanical strains and stresses during the temperature ramp up period. The deactivated element method does not
suffer from any ill-conditioning problems, but the interior nodes of the deactivated elements do not move with the rest
of the model. This can cause distorted filler elements when used in a large-deformation welding analysis.

The initial temperatures for the deposited filler elements can be directly specified on the WELD FILL model definition
option. Alternately, heat input from the weld torch can be modeled as distributed fluxes acting on for the base metal
and deposited weld filler through the WELD FLUX model definition option. When the temperature option is used,
special thermal boundary conditions are dynamically specified on the filler elements as they are created. These melting
point temperatures are directly specified on the nodes of the created filler elements as long as the latter remain in the
weld pool. Once the heat source moves on, the nodal temperature boundary conditions are removed and the filler
elements are allowed to cool. By default, the melting point temperatures are applied instantaneously. This sudden
boundary condition can cause convergence difficulties, especially with the AUTO STEP time stepping algorithm
which bases the time step on the allowable temperature increase. A small ramp time (thermal activation time) can be
optionally provided by the user to alleviate this problem. The temperature of the filler elements is increased from the
current value to the melting point value over the ramp time specified by the user. If the ramp time is left at 0, the
melting point temperature is applied instantaneously.
The identification of the filler elements to be created in any particular increment is given by the following algorithm:
Let , , be the global position of the weld heat source origin at the beginning of the increment. Let ,
, be the global position of the weld heat source origin at the end of the increment. The global coordinates of
the final position and of the filler element nodal points are transformed to local coordinates using the local welding
coordinate system based on the weld path and arc orientation. Let , , be the local coordinates of the final
source position and , , be the local coordinates of the nth node of a filler element. A bounding box in
the local coordinate system is then used to determine if the nodal point falls within the weld pool. Let the dimensions
of the bounding box be defined as , , and , where is the dimension in the weld width direction,
is the dimension in the weld depth direction, is the dimension in the forward weld path direction and is
the dimension in the rear weld path direction. If not specified by the user, is taken as 1.5 times the weld pool width,
is taken as 2 times the weld pool width, is taken as the forward weld pool length and is taken as the rear
weld pool length.
Nodal point n is said to be inside the weld pool if the following relations are satisfied:
(6-18)
If all nodes of a particular element are inside the weld pool, then the element is activated. It should be noted that in
case the weld heat source moves across multiple segments of the weld path in a given increment, a number of sub-
steps are used within the increment and the above algorithm is used for a linear segment in each substep.
Filler elements can be set up as a homogeneous mesh with the base structure or as separate contact bodies. In the
former case, thermal boundary conditions are automatically created at the filler-base interface. In the latter case, an
appropriate contact heat transfer coefficient needs to be specified to ensure that the heat from the filler is transferred
to the base structure.
X
gi
Y
gi
Z
gi
X
gf
Y
gf
Z
gf
X
l f
Y
l f
Z
l f
X
l n
Y
l n
Z
l n
X
b
Y
b
Z
bf
Z
br
X
b
Y
b
Z
bf
Z
br
X
b
Y
b
Z
bf
Z
br
X
ln
X
b
0 < Y
ln
Y
b
0 <
Z
ln
Z
br
0 < + Z
ln
Z
lf
Z
bf
0 <
263
CHAPTER 6
Additional items to facilitate welding analysis in Marc are:
Time-Stepping and Convergence Checking
Fixed stepping or adaptive stepping procedures can be used for welding problems. The fixed stepping scheme that
should be used is TRANSIENT NON AUTO. The adaptive stepping scheme that should be used is AUTO STEP.
For fixed stepping, the only temperature control that can be used is the temperature error in estimate. For adaptive
stepping, the temperature controls that can be used include the temperature error in estimate and the allowable
temperature change per increment. The default temperature change per increment is 20. Time-step cutback is used in
the AUTO STEP algorithm if the temperature controls are not satisfied. The TRANSIENT time stepping scheme is
obsolete and is not recommended for welding problems since time step cutback is not available with that scheme when
the time step is reduced. When the time step is reduced, time step cutback is necessary to ensure that the weld motion,
filler element activation, etc. work properly.
Adaptive Meshing
Two of the most useful adaptive meshing criteria for welding are:
Temperature Gradient Based criterion and
the Node in Box criterion.
The general Box criterion requires the UADAPBOX user subroutine. However, for welding, the box definition has been
automated in the program. The current location of the weld source and the weld pool dimensions are used to define
the box. Nodes that fall in the box and the associated elements are adapted. An unrefine capability is also provided so
that when the box moves away, the subdivided element mesh become coarse. It is important that the user provides weld
pool dimensions through the WELD FLUX model definition option for the box criterion to work properly. It is also
important to note that if the weld filler elements are identified as candidates for adaptive meshing, a suitable upper
bound for the number of weld filler elements should be provided on the WELDING parameter.
Material Properties
Nonlinear temperature dependence for mechanical and thermal material properties is automatically taken into account
by the program. Latent heat of solidification can be modeled by specifying the solidus/liquidus temperatures or by
varying the specific heat as a highly nonlinear function of temperature. An appropriate temperature error in estimate
control should be used to ensure that the material properties at the current temperatures are used. It should be noted
that a general solid-solid phase transformation capability is not available in the current version of Marc.
Parallel Computation
The welding capability is fully supported in DDM. Each data file should contain the complete weld path. When nodes
are used to define the weld path/weld orientation, this may require nodes belonging to other domains to be written out
to the domain file. Also, if a weld flux belonging to domain A refers to a weld filler F, the elements of which are in
domain B, then weld filler F with an empty set of elements needs to be available in domain A.
Radiation
There are six approaches to solve radiation problems in Marc with different levels of sophistication. They include:
1. Viewfactor calculation by direct adaptive integral method.
2. View factor calculation by Monte Carlo method.
3. Viewfactor calculation by Pixel Based Modified Hemi-cube method.

4. Radiation to Space using the FILMS model definition option.
5. Radiation to Space using any of the CONTACT or THERMAL CONTACT options.
6. Radiation into the body using the QVECT option.
There are several aspects in performing a radiation viewfactor calculation including:
1. Defining the edges or faces involved in the viewfactor calculation and determining if the region (cavity) is open
or closed.
2. Calculation of the viewfactors.
3. If the region is open defining the environment temperature. This temperature may be constant or varying with
time.
4. Definition of the surface emissivity and absorptivity, including temperature dependence and frequency
dependence (spectral behavior). By default, the absorptivity is equal to the emissivity.
5. Redefinition of the viewfactors due to large deformation or other phenomena.
6. Redefinition of the viewfactors due to either local or global adaptive meshing.
7. Radiation between surfaces results in increasing the size of the operator (stiffness) matrix, which results in
greater memory requirements and increased computational times.
All of these aspects are discussed in the following paragraphs.
In the radiation calculations in Marc, there are several assumptions made:
Each surface is a diffuse emitter and reflector; i.e., the thermal behavior is independent of the orientation of the
radiation.
Each surface is black; i.e., is a perfect absorber for all incident radiation.
The surfaces are isothermal.
The third assumption requires either that an adaptive procedure is used to insure accuracy or that the finite elements
are sufficiently small for each surface to be assumed to be isothermal.
Using modern mesh generation techniques, there is a tendency to over-refine the finite element mesh, so the need for
these adaptive techniques may be less significant.
In theory, the viewfactors would form a symmetric matrix, the size of which is dependent upon the number of radiating
surfaces. If one has a closed cavity, in theory, the summation of all viewfactors emitting from a surface should equal
one. If desired, the numerically evaluated viewfactors can be scaled such that the sum is one.
Computational Approaches
There are several methods available for the viewfactor calculation in Marc, namely Direct Adaptive Integration,
Monte Carlo, and Pixel Based Modified Hemi-Cube Method. The choice is made on the RADIATION parameter. While
265
CHAPTER 6
this is the historical order that they have been implemented, the Hemi-Cube method runs fastest, gives the most
accurate viewfactors, and is the most flexible, and hence, is discussed first.
1. Defining the Radiating Surfaces
Depending upon which technique is used for calculating the viewfactors, this can be summarized as follows:
a. Pixel Based Modified Hemi-Cube
The CAVITY DEFINITION option is used to define the edges (2-D) or faces (3-D). Using Mentat, this is in
the MODELING TOOLS CAVITIES menu. Shells may be included in the model, in which case, one needs to
indicate if it is the top or bottom or both surfaces. Symmetry surfaces may be defined. If physically there are
multiple cavities in the problem, they should be placed in separate cavities to reduce the computational time.
b. Monte Carlo
The definition of the edges (2-D) or faces (3-D) is done via Mentat or Patran; see BOUNDARY
CONDITIONS>THERMAL>EDGE RADIATION OR FACE RADIATION menu. The calculation of the
viewfactors is also performed in Mentat or Patran via the BOUNDARY
CONDITIONS>THERMAL>COMPUTE RADIATION menu.
c. Direct Integration
In Marc, this is only available for axisymmetric solid elements. The geometry is defined using the
RADIATING CAVITY option by specifying the node numbers around the cavity. It is not supported by Mentat.
2. Radiation Viewfactors
In many analyses, the radiative transfer of heat between surfaces plays a significant role. To properly model
this effect, it is necessary to compute the proportion of one surface which is visible from a second surface
known as the formfactor or viewfactor. The viewfactor, defined by a fourth order integral, presents many
difficulties in its computation. Primary among these is the large amount of computing power needed, especially
when shadowing effects are included.
The radiative flow of hear from surface 1 to surface 2 is given by:
(6-19)
in which, is the viewfactor and is calculated as:
Pixel Based Modified Hemi-Cube Method
The foundation of radiation is based upon radiation being received from multiple directions and/or being emitted in
multiple directions. This leads to the concept of a hemisphere, where one considers an element being projected onto
the hemisphere as shown in Figure 6-27.
This geometric projection can be easily calculated for 2-D or axisymmetric, but is problematic in 3-D. To overcome
this problem of exact projections, one can consider dividing the hemisphere into pixels as shown in Figure 6-28, and
the simple count the pixels. This procedure is easier to implement in 3-D, though it is still expensive. As an
alternative one can approximate the hemisphere with a Hemi-Cube as shown in Figure 6-29.
q
12
F
12
T
1
4
T
2
4
( ) =
F
12
F
12
1
A
1
------ -
1

2
cos cos
r
2
------------------------------ A
2
d A
1
d
A
2
}
A
1
}
=

Figure 6-27 Element Projected onto Hemisphere
Figure 6-28 Hemisphere Divided into Pixels
Figure 6-29 Hemisphere with Hemi-Cube
Using the Hemi-Cube method is less expensive, but is still difficult around the edges, and especially the corners in 3-
D. One way to reduce this problem is to change the dimensions of the box, such that there is a smaller possibility that
leakage out of the sides is important and then simply neglect this. This is shown in Figure 6-30, where a Hemi-Box
now replaces the Hemi-Cube. To retain accuracy, one must select a good dimension of the box, and as the space is now
267
CHAPTER 6
biased, an equally spaced pixel grid will lead to inaccuracies. To overcome this problem, a Hemi-Plane will be used
with non-equally spaced pixels, such that the weight of each pixel is the same. This is shown in Figure 6-31
Figure 6-30 Hemisphere with Hemi-Box
Figure 6-31 Hemisphere with Hemi-Plane
The basic algorithm for the viewfactor program is as follows:
a. Read in file created by the analysis program containing the geometry.
b. For axisymmetric geometries, radiating faces are swept over a user controlled angle.
c. In 3-D, subdivide quadrilateral faces into triangles.
d. Create a map of nonuniform pixels and associate a weight with them.
e. Convert coordinates to local system.
f. Loop over receiving surface, lines in 2-D, triangles in 3-D.

g. In 2-D, project line unto line; in 3-D, project triangle onto plane.
h. Identify pixels (number and location).
i. Sum the pixel weights.
j. Output the view factor for this surface.
When symmetry planes are present, the receiving faces are doubled for each symmetry plane. For cyclic symmetry,
the receiving faces are doubled n times.
This method does not guarantee a symmetric Radiation Exchange Matrix because all are calculated
separately and independently based upon the above algorithm. There exists a possibility to make the radiation
Exchange Matrix symmetric. For large models, this may be time consuming. The method is based upon an iterative
averaging and column normalization, and for large problems, requires a large amount of memory or disk space.
Monte Carlo Method
In this method, the idea is to randomly emit rays from the surface in question. The percentage of these rays which hit
another surface is the formfactor between the surfaces. The Monte Carlo method computes N formfactors at one time,
providing linear scaling. In fact, the larger the number of surfaces, the faster the formfactors are computed compared
to the Direct Adaptive Integration and Adaptive Contour Integration. Hence, this method is adopted for the viewfactor
calculations.
Once the viewfactors are calculated by Mentat; the resulting output file containing the viewfactors can be read in with
the inclusion of the RADIATION parameter and the VIEW FACTOR model definition option.
Some of the features of the viewfactor calculation in Marc are:
a. You are not required to specify blocking elements. This is embedded into the algorithm completely and,
hence, done automatically. This is specially useful in three-dimensional analysis since, for complex
geometries, it is impractical to predict what surfaces are blocking other surfaces.
b. The cost of calculation is nearly linearly proportional to the number of elements which means that, for big
problems, the cost does not increase significantly.
c. The methodology in Marc guarantees that the sum of viewfactors is always 1.0. This is in contrast to the
direct integration approach which has the requirement of normalization to avoid artificial heat gain and loss,
because the sum of the viewfactors for any one surface is unlikely to add up to 1.0 due to errors in the
computation of individual viewfactors.
d. Shadowing effects (due to two surfaces being hidden from one another by other surfaces) can be modeled.
For a surface to participate in the computations, it must participate in the following operations:
a. Ray Emission: The surface should be able to randomly emit a ray from its surface. The origin of the ray
should be randomly distributed over the area of the surface (see Figure 6-32). The direction of the ray should
be distributed according to the cosine of the angle between the ray and the normal to the surface at the origin
of the ray, thereby emitting more rays normal to the surface than tangent to it.
b. Ray Intersection: Given an origin, direction, and length of a ray, the surface should be able to determine
whether it is hit by that ray, and if hit, the length of the ray at the point of intersection.
Determination of viewfactors involves consideration of several desired properties like speed, accuracy, shadowing,
translucence, absorption, nonuniform emission, reciprocity, and efficiency. In light of the properties listed above, the
Monte Carlo approach is adapted in conjunction with the ray tracing and boxing algorithms.
A
j
F
i j
( ) F
i j
269
CHAPTER 6
Figure 6-33 depicts the formation of shadows which essentially involves the computation of incident light by tracing
rays from the light source to the point of incidence of the eye rays. These rays do not reflect or refract. Shadows are
formed when light source rays are obstructed, either partially or fully, from reaching an object. Refraction and
reflection of light from light sources is not computed.
Figure 6-32 Random Ray Emission in Monte Carlo Method
Figure 6-33 Shadowing Effects in Ray Tracing
Finally, to compute the viewfactors, it is not necessary to check the intersection of incident rays with all objects under
consideration. Such a method would be prohibitively computationally expensive and preclude a large scale three
dimensional analysis. An effective technique for fast calculation of intersections is employed.
The method relies on sorting the objects before any intersections are computed. The information computed is used to
eliminate most of the intersection computations. The technique requires that each object to be sorted have a bounding
box which entirely encloses it.
The sorting relies on creation of a binary tree of bounding boxes. Thus, a bounding box is computed for all objects.
The objects are then sorted by the coordinate of the longest dimension of the bounding box. This list of objects is then
divided into two sets, each having an equal number of objects. This process is repeated recursively until each set
contains no more than a given maximum number of objects. This recursive sorting process is depicted in Figure 6-34.
1
2
Point Light
Eye Ray
Light Source Rays
Shadow

Ray intersections are performed by searching the binary bounding box tree which involves determining whether the
ray hits the bounding box of all the objects. If the intersection with one bounding box is not found, then another
bounding box is considered. In the event an intersection does exist, each object in the node is intersected with the ray
if the intersection is a bottom node of the tree. Otherwise, the bounding boxes of both subtrees are intersected with the
search in the closest subtree conducted first. The process continues until all rays are exhausted.
Figure 6-34 Fast Intersection Technique for Sorting Objects
2 1
6
9
8
5 4 3
7
271
CHAPTER 6
Direct Adaptive Integration
This approach computes viewfactors by directly computing a fourth order integral between every pair of surfaces. This
means that there are N squared integrations to be performed, a quadratic scaling.
Besides being quadratic, to achieve reasonable accuracy, this approach requires a huge number computations for
surfaces which are close together; a situation which frequently occurs.
Using the direct integration approach, Marc calculates the viewfactors automatically for you in each cavity of an
axisymmetric body involving radiative heat transfer. This capability is only available for axisymmetric bodies.
You must subdivide the radiative boundary in this heat transfer problem into one or more unconnected cavities. For
each cavity, you define the outline of the cavity in terms of an ordered sequence of nodes. Usually, the nodes coincide
with the nodes of the finite element mesh. You can add extra nodes, provided you also give the appropriate boundary
conditions.
The nodes must be given in counterclockwise order with respect to an axis orthogonal to the plane of the figure and
pointing to the viewer (see Figure 6-35). If the cavity is not closed, the program adds the last side by connecting the
last node with the first one. This side is treated as a black body as far as radiation is concerned; its temperature is taken
as the average between the temperatures of the adjacent nodes.
Figure 6-35 Radiating Cavity (Approach II: Valid for Axisymmetric Capability Only)
Marc internally computes the viewfactor between every side of the cavity and all other sides. The matrix with the
viewfactors can be stored into a file, and read in again during a subsequent analysis, thus avoiding a new computation.
During a transient heat transfer analysis, for every time step, the program estimates the temperature reached at the end
of the step. From the estimated temperature, the emissivity (temperature dependent) is computed. In addition, the
radiating heat fluxes are computed. The temperatures at the end of the step are computed by solving the finite element
equations.
At every node, the difference between estimated and computed temperature is obtained. If the tolerance allowed by
the CONTROL model definition option is exceeded, iterations within the time step take place. Otherwise, the
computation of the step is concluded.
The cavity is defined by its boundary defined by a list of nodes ordered counterclockwise.
Visibility
V
1 2
3
4
5
6
7
8
9
10
11
R
Z

Insulated boundary condition (for example, symmetry boundaries) requires that the sum of the heat fluxes at a node
be zero. This requirement is satisfied automatically. Therefore, no input is required for this type of boundary condition.
As previously mentioned, in a heat transfer analysis of axisymmetric body involving radiative boundary conditions,
Marc automatically calculates viewfactors for radiation. A description of the viewfactor calculation follows.
The amount of radiation exchanged between two surfaces will depend upon what fraction of the radiation from each
surface impinges the other surfaces. Referring to Figure 6-35, the radiation propagating from surface i to surface j will
be:
(6-20)
It is noted that the viewfactor, , is solely geometrical in nature. From the definition of , we see that
(6-21)
We are now ready to derive the heat transfer radiation equation. Steady-state (equilibrium) energy conservation
requires:
(6-22)
Two independent expressions for can be formed.
a. The incoming radiation on a surface must equal the radiation emitted by all other surfaces which strike this
surface,
(6-23)
where is the number of surfaces involved in the computations.
b. The other expression for is
(6-24)
where
is the emissive power
is the emissivity
is the reflectance.
q
i j
J
i
A
i
i

j
A
j
cos cos
r
2
----------------------------------
\ .
| |
J
i
A
i
F
i j
= =
F
i j
F
i j
A
j
F
j i
A
i
F
i j
=
q
i
A
i
J
i
G
i
( ) =
G
i
A
i
G
i
J
j
j 1 =
N
A
j
F
j i
J
j
A
i
F
i j
j 1 =
N
= =
G
i
J
j
j 1 =
N
F
i j
=
or
N
G
i
G
i
J
i

i
E
n
i
i
-----------------------
J
i

i
E
n
i
1
i
----------------------- = =
E
n
273
CHAPTER 6
Substituting Equation (6-24) into Equation (6-22) and rearranging it gives
(6-25)
This expression for can be inserted into Equation (6-23). After regrouping terms, you get the governing
equation for gray body diffuse radiation problems
(6-26)
For the problem of black bodies, that is, = 1, we have,
(6-27)
This equation states the obvious; net radiation heat flow from a black surface is the difference between
radiation given off and received; that is, there is no reflection.
Figure 6-36 Viewfactor Definition
The heat flux radiating from to is computed as
(6-28)
where is the power radiating from , the first term within parentheses is the projection of normal to
the connecting line, and the second term is the solid angle under which is seen from the center of
defining the viewfactor:
(6-29)
J
i
E
n
i
1
i
A
i
i
-------------- q
i
=
J
i
A
i
i j
1
j
( )A
i
F
i j
A
j
j
---------------------------------------------------- q
j
{ } A
i
i j
A
i
F
i j
[ ] E
n
i
)
`

=
q
j
{ } A
i
i j
A
i
F
i j
[ ] E
n
i
)
`

=
A
i
i
r
A
j
j
A
i
A
j
q
i j
J
i
A
i

i
cos ( )
A
j

j
cos
r
2
--------------------
\ .
| |
=
J
i
A
i
A
i
A
j
A
i
F
i j
i

j
cos cos
r
2
---------------------------- A
j
=
q
i j
J
i
A
i
F
i j
=

3. Environment Temperature
If the sum of the viewfactors from an emitting face is not equal to one, then there is a radiation contribution to
the environment. Hence, If the cavity is closed, there is no radiation to the environment and the environment
temperature is not used.
Including a PRINT-30 parameter, one will obtain the sum of the viewfactors from each emitting surface. For a
closed, cavity, comparing this to one will be a measure of the accuracy of the viewfactor calculation.
The environment temperature may either be defined by creating an extra node and associating it with a cavity
on the CAVITY DEFINITION model definition option. The temperature of the node is defined using FIXED
TEMPERATURE model definition option. As an alternative one can specify the environment temperature on
the RAD-CAVITY model definition option.
b. Monte Carlo
A constant environment temperature is defined in Mentat or Patran. This information is passed into Marc in
the vfs file.
c. Direct Integration
The cavity is assumed to be closed; there is no radiation to the environment.
4. Definition of the Surface Emissivity
The emissivity may be entered on an element edge or face. Hence, a particular element such as a brick could
have different emissivities because of surface conditions. The emissivity could be temperature and
frequency dependent. This is defined using the EMISSIVITY model definition option and the TABLE model
definition option if required. Is is also possible to define the emissivity via the ISOTROPIC,
ORTHOTROPIC, or other material options.
The emissivity of materials as a function of wavelength is discontinuous in nature due to the discrete nature
of particle physics.
Frequency dependent emissivity (spectral) is defined by three types of bands: zero value, narrow, and wide.
The evaluation is also dependent on whether the band is at low or high frequency as shown in Figure 6-37.
The value of emissivity at the endpoints of each band may be a function of temperature. Based upon the table
input procedure, it is possible to enter the emissivity as a function of temperature and wavelength. The
emissivity is assumed to have a linear variation within each band.
275
CHAPTER 6
Figure 6-37 Band Frequency
Besides entering the emissivity, if wavelength dependent data is used then one also needs to enter Plancks
second constant and the speed of light. This is done via the PARAMETERS model definition option. The unit
of wavelength used in the table should be consistent with the unit of length used to define the speed of light.
Wavelength dependent emissivity is based upon:
Narrow bands - use average reciprocal wavelength
Wide bands at short wavelength - use exponential expansion
Wide band at long wavelength - use power series in reciprocal wavelength
b. Monte Carlo or Direct Integration
The emissivity is defined via the ISOTROPIC, ORTHOTROPIC, or other material options. Temperature
dependent emissivity is entered through the TEMPERATURE EFFECTS or TABLE model definition options.
5. Redefinition of Viewfactors due to Large Deformation
If the distance between the updated coordinates and the coordinates at which the last viewfactor calculation was
performed is greater than user-specified value, then the viewfactors will be re-evaluated. Note that the motion
of nodes on the cavity is only an approximate measure of the inaccuracies of the previously generated
viewfactors. The theoretically correct way would be to continuously re-evaluate the viewfactors, and then
measure the difference between the two sets of viewfactors. This, of course, would defeat the objective of
minimizing calculations. The critical distance to re-evaluate is specified on the RAD-CAVITY model definition
option. This method is only available if the LARGE STRAIN parameter is used. It is also possible to specify that
the viewfactors should be re-evaluated based upon increment numbers.
Viewfactors are only recalculated if the Pixel Based Modified Hemi-Cube method is used. In the current
release, this is only available if there is one cavity in the model.

6. Redefinition of Viewfactors due to Adaptive Meshing
If the geometry of the cavity has been defined using geometric entities (curves and surfaces) in the CAVITY
DEFINITION model definition option, then when local adaptive meshing (2-D or 3-D) or global adaptive
meshing (2-D) occurs, the viewfactors will be re-evaluated. This is based upon that after remeshing the new
edges or faces are automatically attached to the curves or surfaces. The program will determine the new list of
element edges or faces in the cavity and recalculate the viewfactors.
This capability is only available with the Pixel Based Modified Hemi-Cube method.
7. Computational Costs
Regardless of the method used to calculate the viewfactors, there are two consequences:
a. The number of viewfactors becomes large, growing quadratically with the number of radiating surfaces.
b. The subsequent calculation in the analysis program is dependent upon the number of viewfactors calculated.
The number of viewfactors is reduced because of shadowing effects, but depending upon the procedure used
to calculate the viewfactors, this may lead to higher computational costs in calculating them. Because self-
shadowing or third surface shadowing is likely, sparse storage techniques are beneficial in storing the view
factors. As the data structures required for the view factor calculations are different than what is required for
traditional finite element or finite difference calculations, the computation is often put into stand-alone
programs.
Using finite difference, contour integral, Gaussian integration or direct integration, the number of operations
increases quadratically with the number of radiating surfaces. All of these procedures (combined with adaptive
approaches or refined meshes) can produce accurate results, but at a clear computational cost. With the Monte
Carlo method, the computational cost increases linearly with the number of radiating surfaces and increases
linearly with the number of emitting rays. This clearly is beneficial for large models. The problem with the
Monte Carlo method is that a large number of rays may be necessary to achieve accurate results (for example,
if thin cavities exist).
The advantage of the Pixel Based Modified Hemi-Cube method is that the computational costs are linearly
dependent on the number of radiating surfaces and linearly dependent upon the number of pixels. The accuracy
is dependent upon the number of pixels chosen, and the size/shape of the cube. A very large number of pixels
has the ramification that the amount of memory required is increased and the amount of calculations is
increased.
The number of viewfactors increases the cost of the analysis calculation, because the viewfactor file is larger,
and because the bandwidth of the system increases. There are two ways to control the computational costs:
a. One can set a tolerance and any viewfactor less than this value will be ignored. This results in radiation being
neglected.
b. One can set a second tolerance and viewfactors below this number will be treated explicitly as opposed to
implicitly. This means that the radiation contribution will be on the right-hand side and not in the operator
matrix. This may result in more iterations, but lead to a significant reduction in the size of the operator
matrix. This leads to a smaller bandwidth and a reduction in the computational costs to solve the system of
equations.
277
CHAPTER 6
Nonstructural and Coupled Procedure Library
6SS\
Graphical Visualization
Using the Pixel Based Modified Hemi-Cube method, is possible to visualize the relative magnitude of the viewfactors
associated with an emitting surface. This will create a post file, jid_cxx_vfs.t19, where xx is the cavity id. Each
increment will represent the viewfactors from one surface. Creating a contour plot with Mentat, one will observe a
triangle (2-D) or pyramid (3-D) indicating which face is emitting and the values of the viewfactors to the other faces.
Data, Data Flow, and Data Constraints
The geometric information is created in Marc and written to a file, jid_cxx_t18, where jid is the job id
and xx is the cavity number.
The stand-alone program then reads this file and calculates the view factors. This is then written to a file,
jid_cxx.vfs.
The analysis program then reads this file and performs the thermal analysis.
If graphical visualization of the viewfactors is requested, this will be written to file, jid_cxx_vfs.t19.
b. Limitations
A radiating cavity must be fully contained within one domain if DDM is used.
There is currently a limitation of 99 cavities.
Conrad Gap
For the thermal contact gap element, in the gap open condition, two surface temperatures and at the centroid
of the surfaces of a thermal contact element are obtained by interpolation from the nodal temperatures. These two
surface temperatures are used for the computation of an equivalent conductivity for the radiation/convection link. The
expression of the equivalent thermal conductivity is:
(6-30)
where is the emissivity, is the Stefan-Boltzmann constant, is the length of the element (distance between a and
b), , are absolute temperatures at a and b converted from and ; and is the film coefficient. The film
coefficient can be a function of the temperature by referencing a table.
The equivalent thermal conductivity for the thermal contact element is assumed to be in the gap direction. The
thermal conductivities in other two local directions are all set to zero. A coordinate transformation from the local to
the global coordinate system allows the generation of the thermal conductivity matrix of the thermal contact element
in the global system for assembly.
Similarly, in the gap closed condition, tying constraints are automatically generated by the program for thermal contact
elements. The constraint equation for each pair of nodes can be expressed as:
(6-31)
T
a
T
b
k
1
k
1
L T
ka
T
kb
+ ( ) T
ka
2
T
kb
2
+ ( ) H L + =
L
T
ka
T
kb
T
a
T
b
H
k
1
T
I
T
J
+ if T
gap
T
cl ose
> ( )

where
= nodal temperatures at nodes I and J
= gap closure temperature.
Channel
For the fluid channel element, the one-dimensional, steady-state, convective heat transfer in the fluid channel can be
expressed as:
(6-32)
where is mass flow rate, is specific heat, is fluid temperature, is solid temperature, is streamline
coordinate, is circumference of channel, is film coefficient, and is inlet temperature. The film coefficient
may be a function of the temperature, and the inlet mass flow rate and temperature may be a function of time.
Similarly, the conductive heat transfer in the solid region is governed by the following equation:
(6-33)
subjected to given initial condition and fixed temperature and/or flux boundary conditions. At the interface between
the fluid and solid, the heat flux estimated from convective heat transfer is
(6-34)
In Equation (6-33), is the heat capacity matrix, is the conductivity matrix and is the heat flux vector. Equations
(6-33) and (6-34) are coupled equations. The coupling is due to the unknown solid temperature appearing in
Equation (6-32) and unknown fluid temperature in Equation (6-34) for the solution of Equation (6-33).
The solutions for Equations (6-32) and (6-33) are obtained from the introduction of a backward difference for the
discretization of time variable in Equation (6-33) and of streamline distance in Equation (6-32). Let
(6-35)
we obtain
(6-36)
T
gap
1
2
--- T
I
T
J
+ ( ) =
T
I
T
J
,
T
close
m
c
T
f
s
-------- h T
f
T
s
( ) + 0 =
T
f
0 ( ) T
i nl et
=
m
c T
f
T
s
s
h T
inlet
CT
s
KT
s
+ Q =
q h T
s
T
f
( ) =
C K Q
T
s
T
f
T
s
T
s
i
T
s
i 1
[ ] t ( ) =
1
t
----- C K +
\ .
| |
T
s
i
Q
i
1
t
----- CT
s
i 1
+ =
279
CHAPTER 6
where = time-step in transient analysis. Similarly, let
(6-37)
we obtain
(6-38)
where is the streamline increment,
(6-39)
Output
Marc prints out both the nodal temperatures and the temperatures at the element centroid when the CENTROID
parameter is used, or at the integration points if the ALL POINTS parameter is invoked. You can also indicate on the
HEAT parameter for the program to print out the temperature gradients and the resulting nodal fluxes.
To create a file of element and nodal point temperatures, use the POST model definition option. This file can be used
as temperature input for performing a thermal stress analysis. This file is processed using the CHANGE STATE option
in the subsequent thermal stress analysis. This post file can also interface with Mentat or Patran plot temperature as a
function of time.
Heat Transfer with Convection
Marc has the capability to perform heat transfer with convection if the velocity field is known. The numerical solutions
of the convection-diffusion equation have been developed in recent years. The streamline-upwind Petro-Galerkin
(SUPG) method has been implemented into the Marc heat transfer capability.
The elements which are available are described in Table 6-1.
To activate the convection contribution, use the HEAT parameter and set the fifth field to 2. Due to the nonsymmetric
nature of the convection term, the nonsymmetric solver is used automatically. Specify the nodal velocity vectors using
the VELOCITY option. To change velocity, use VELOCITY CHANGE. If nonuniform velocity vectors are required, the
UVELOC user subroutine is used. This capability can be used in conjunction with the Rigid-Plastic Flow section in
Chapter 5 of this volume to perform a coupled analysis, in which the velocity fields are obtained.
Table 6-1 Heat Transfer Convection Elements
36, 65 2-, 3-node link
37, 39, 41, 69, 131 3-, 4-, 6-, 8-node planar
38, 40, 42, 70, 87, 88, 122, 132 3-, 4-, 8-node axisymmetric
43, 44, 71, 123, 133 8-, 20-node hexahedral
135. 133 4-, 10-node tetrahedral
t
dT
f
ds
--------- T
f
T
f
j 1
[ ] s ( ) =
T
f
j
s T
f
j 1
+ [ ] 1 s + ( ) =
s
h m
c ( ) ; = hT
s
j 1
m
c ( ) = and

The general convection-diffusion equation is:
(6-40)
The perturbation weighting functions are introduced as:
is the standard interpolation function in Equation (6-1). The upwinding parameter, , is the weighting used to
eliminate artificial diffusion of the solution; while the beta term, , is to avoid numerical dispersion. is the
magnitude of local velocity vectors. is the temperature, is the diffusion tensor. is the source term and is
the time increment.
The optimal choice for and are:
(6-41)
where is the local Peclet number in the local element and is the local Courant number:
Peclet = density specific heat characteristic length magnitude of the fluid velocity/conductivity
Peclet =
and
where is the time increment.
The characteristic length is defined in [Ref. 11] where is required for numerical stability. When , the
is set to be zero and a large time step is recommended to avoid numerical dispersion.
Diffusion
Marc has a capability to perform diffusion analyses using two approaches. One is to use the capability using the PORE
parameter which was developed for poro-elasticity/soil analysis, and the one described here which is activated using
the DIFFUSION parameter. The basic equations governing the behavior are the mass balance equation and Darcys law
which are supplemented by a compressibility model for the fluid/gas. Darcys law can be derived from an averaged
momentum balance for viscous flows at low Reynolds numbers in porous media.
Note:
The interpolation function N is not the time-space functions defined in [Ref. 6], so that most Marc heat
transfer elements can be used. The convection contribution of heat transfer shell elements is limited due to
the definitions of the perturbation weighting function and the interpolation function.
*
c
T
t
------- v T +
\ .
| |
T ( ) Q + =
W N
h
2 v
---------- v N ( )
\ .
| |

h
4 v
---------- t v N ( )
\ .
| |
+ + =
N
v
T Q t
Pecl et 2 ( ) 2 Peclet ( ) coth =
C 3 2 Pecl et ( ) * C ( ) =
Peclet C
*c*h* v k
C v * t ( ) h = t ( )
h C 1 C 1 >
281
CHAPTER 6
The capability is available for either steady-state or transient behavior. The capability is available for two- or three-
dimensional analysis using the standard heat transfer continuum elements.
Marc computes and prints the following information at the element integration points: pressure, gradient of pressure,
and the mass flow rate. The nodal point data consists of the pressure and the equivalent nodal mass flux.
The material permeability should be entered on either the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model
definition options. Dependency on the pressure, or the temperature should be defined by referencing a table. The
porosity or the void ratio should be entered using the INITIAL POROSITY or INITIAL VOID RATIO model definition
options. In a pure diffusion analysis, those values do not change with time. The fluid/gas data required is the fluid
viscosity, the fluid mass density, and the fluid bulk modulus. This data is also entered in the ISOTROPIC,
ORTHOTROPIC, or ANISOTROPIC model definition options.
It should be noted that for a purely incompressible fluid, the rate term is not significant, and a steady-state solution is
immediately obtained.
The boundary conditions associated with diffusion analysis consist of prescribed pressures defined with the FIXED
PRESSURE model definition option where there is always only one degree of freedom per node. For transient
analyses, initial pressure should also be defined using the INITIAL PRESSURE model definition option.
Specify nodal point mass fluxes in the POINT MASS model definition option. Specify distributed mass fluxes in the
DIST MASS model definition option. Nonuniform or time dependent distributed mass fluxes may be defined by the
FLUX user subroutine or by referencing a table.
Diffusion analysis can be used in conjunction with the contact capabilities. Contact bodies are defined through the
THERMAL CONTACT option.
The diffusion capability is based upon an implementation of the mass balance equation for low Reynolds number flow
of a liquid or gaseous fluid through a rigid porous medium. An averaged momentum balance for such a flow results
in Darcys law relating the average flow velocity of the fluid to the pressure gradient. This allows the flow problem to
be formulated in terms of the unknown pressures alone, greatly simplifying the formulation and also greatly reducing
the cost of the solution. Usually, the fluid is assumed to be incompressible, but the formulation here is extended to
include compressibility by allowing the density to be a function of pressure.
The balance relations are derived from considering a fixed control volume of the rigid porous material, where the pores
are filled with the flowing fluid, so mass transport is taking place over the control volume boundaries. The total volume
of a volume element is the sum of the volume taken by the pores filled with fluid and the remaining volume
of the solid material :
The ratio of the pore volume to the total volume defines the porosity , which is easily seen to be bounded by
and is a characteristic parameter of the material:
dV dV
g
dV
s
dV dV
g
dV
s
+ =
0 1
dV
g
dV
---------- =

The porosity may vary spatially, but it is constant over time since the solid material is assumed to be rigid. A related
quantity that is often used is the void ratio , which is defined as the ratio of the pore volume to the solid volume and
is an unbounded quantity :
The two are dimensionless quantities and they are related by:
Since the solid material is assumed to be completely rigid its mass inside the control volume is constant and does not
explicitly enter the mass balance equation. The formulation here is for a single fluid with density which is assumed
to flow through the volume with a percolation or filtration velocity . This percolation velocity is an averaged flow
velocity, averaged with respect to entire volume, and not the real flow velocity of the fluid, since the fluid only
occupies the pores and not the entire volume. The two velocities are related as:
where denotes the real velocity, which is understood as an averaged velocity in the pores of the volume. This
relation easily follows from comparing the average momentum over an entire volume element with the average
momentum over its pore volume :
The mass balance relation states that the rate of change of the fluid mass inside the control volume equals the rate of
fluid mass flow that takes place over the control volume boundary:
Here represents the outward unit normal to the control volume boundary. Employing the divergence theorem this
leads to the following partial differential equation:
where
is the divergence operator
The rate term in above differential equation represents the local time derivative of the fluid density.
e
0 e <
e
dV
g
dV
s
---------- =
e
1 e +
------------ =
g
u
g
u
g
u
g
=
u
g
dV
dV
g
g
u
g
dV
g
u
g
dV
g

g
u
g
dV = =
d
dt
-----
g
V d
V
}

g
u
g
n A d
A
}
=
n
g
t
--------- di v
g
u
g
( ) + 0 =
di v
283
CHAPTER 6
Darcys law relates the percolation velocity to the pressure gradient as:
where
The units here are mentioned in S.I. units, but, of course, other units can be used as well. Darcys law can be viewed
as an averaged momentum balance for low Reynolds number flows through a porous medium. The formulation here
is for an isotropic material, but orthotropic or anisotropic behavior for the permeability is supported as well.
The fluid is assumed to be compressible and its density may be expressed as a function of pressure, ,
from which the density change rate can be expressed in terms of the pressure change rate. It is customary to express
the density change rate in the following way:
where
Combining these relations results in the following partial differential equation that must be met in the interior of the
volume in terms of the unknown pressure:
The boundary conditions are either a prescribed pressure on some parts of the exterior boundary of the volume
(denoted by ) or a prescribed mass flow rate on other parts of the exterior boundary of the volume (denoted by
):
Here, the asterisked terms denote the prescribed quantities. The union of the two regions represents the entire exterior
boundary, their intersection is empty. If nothing is specified on some part of the boundary, the natural boundary
is the absolute or intrinsic permeability
[m
2
]
is the dynamic viscosity
[Ns/m
2
]
pis the pressure
[N/m
2
]
is the gradient operator
is the bulk modulus of the fluid, which is defined as:
on :
prescribed pressure
[N/m
2
]
on :
prescribed mass flow rate
[kg/sm
2
]
u
---p =
g

g
p ( ) =
g
t
---------

g
p
Kt
------------ =
K
K
g
d
g
dp
--------- =
g
K
---------
p
t
------ di v
g
--- p
\ .
| |
0 =
A
p
A
m
p p* = A
p
---
p
n
------ m* = A
m

condition with automatically applies meaning there is no mass exchange with the environment on this part
of the boundary.
The problem description is completed by specifying the initial pressure distribution:
To arrive at the finite element equations of the problem the governing differential equation with the accompanying
boundary conditions is cast in a weak form:
Here, the weighting functions are denoted by . By the Galerkin method, they are interpolated spatially in the same
way as the unknown pressures and they are referred to as the virtual pressures. The virtual pressure field is zero on
those boundary areas where the pressure is prescribed, so the area integral only extends over the boundary area where
the mass flow rate is prescribed. Using partial integration and employing the divergence theorem this can be brought
to its final form:
This represents a nonlinear transient equation for the unknown pressure field. By choosing appropriate spatial
interpolations for this pressure field a nonlinear system of ordinary differential equations is obtained:
where
and the dot over the pressure vector denotes time differentiation. By further choosing appropriate temporal
interpolations, this becomes a system of nonlinear algebraic equations. As a time integration method, Marc uses the
backward Euler method which is the same method that is being used in heat transfer analysis. Similar controls for
automatic time stepping, updating of pressure dependent matrices, and for governing the iteration behavior within an
increment are used as for heat transfer analysis.
If the variation of fluid density with pressure is small, the density values in above finite element equations may be held
at their initial values and can be treated as constants. If the other material parameters ( , , and ) are also constant,
the system of equations becomes linear. If the fluid can be assumed to be incompressible ( ), the transient term
vanishes and a steady state solution is immediately obtained.
is the vector of unknown nodal pressures
is the diffusion capacity matrix
is the diffusion conductivity matrix
is the vector of external nodal mass fluxes
m
*
0 =
p x
t , ( 0 ) p
0
x
( ) = =
g
K
---------
p
t
------ iv
g
--- p
\ .
| |
d p V m*
g
--- p n
\ .
| |
p A d
A
m
}
d
V
}
0 =
p
p
g
K
---------
p
t
------ p V
g
--- p p V d
V
}
+ d
V
}
m*p A d
A
m
}
=
M
g
p
K
g
p + m
g
=
p
M
g
K
g
m
g
K
K
285
CHAPTER 6
The fluid density ( ) and the other fluid material parameters ( , , and ) can all be functions of pressure making
the system of equations nonlinear and these pressure dependences can be entered through tables. If the fluid density
is entered through a table, the bulk modulus is evaluated by the functional relation given earlier through a numerical
differentiation. This is done irrespective of the bulk modulus value given in the input, unless the density variation in
the table with pressure is very small, in which case the bulk modulus from the input also gets evaluated.
Hydrodynamic Bearing
Marc has a hydrodynamic bearing analysis capability, which enables you to solve lubrication problems. This
capability makes it possible to model a broad range of practical bearing geometries and to calculate various bearing
characteristics such as load carrying capability, stiffness, and damping properties. It can also be used to
analyze elasto-hydrodynamic problems.
The lubricant flow in hydrodynamic bearing is governed by the Reynolds equation. The bearing analysis capability
has been implemented into Marc to determine the pressure distribution and mass flow in bearing systems.
Marc is capable of solving steady-state lubrication problems; the incremental procedure analyzes a sequence of
different lubricant film profiles. Marc also can be used to solve coupled elasto-hydrodynamic problems. This analysis
requires a step-by-step solution for both the lubrication and the stress problems using separate runs. Because the finite
element meshes for each problem are different, the program does not contain an automated coupling feature.
Only one-dimensional or two-dimensional lubricant flow needs to be modeled, since no pressure gradient exists across
the film height. This modeling is done with the available heat transfer elements. The library elements listed in
Table 6-2 can be used for this purpose.
Marc computes and prints the following elemental quantities: lubricant thickness, pressure, pressure gradient
components, and mass flux components. Each of these is printed at the element integration point.
The nodal point data consists of pressures, equivalent nodal mass flux at fixed boundary points, or residuals at points
where no boundary conditions are applied. In addition, the program automatically integrates the calculated pressure
distribution over the entire region to obtain consistent equivalent nodal forces. This integration is only performed in
regions where the pressure exceeds the cavitation pressure.
Table 6-2 Hydrodynamic Bearing Elements
Element Description
36 2-node, three dimensional link
37 3-node, planar triangle
39 4-node, bilinear quadrilateral
41 8-node, planar biquadratic quadrilateral
65 3-node, three-dimensional link
69 8-node, biquadratic quadrilateral with reduced integration
121 4-node bilinear quadrilateral with reduced integration
131 6-node triangle
g
K

The output includes the load carrying capacity (the total force on the bearing). This capacity is calculated by a vectorial
summation of the nodal reaction forces. In addition, the bearing moment components with respect to the origin of the
finite element mesh can be calculated and printed.
To activate the bearing analysis option, use the BEARING parameter. If the analysis requires modeling of flow
restrictors, also include the RESTRICTOR parameter.
The values of the viscosity, mass density, and cavitation pressure must be defined in the ISOTROPIC option. Specify
temperature-dependent viscosity values via TEMPERATURE EFFECTS or TABLE. If thermal effects are included, the
STATE VARS parameter is also required. In hydrodynamic bearing analyses, temperature is the second state variable.
Pressure is the first state variable.
The fluid thickness field can be strongly position-dependent. A flexible specification of the film profile is allowed by
using either the nodal thickness or elemental thickness option. Define nodal thickness values in the THICKNESS
option. You may also redefine the specified values via the UTHICK user subroutine. Elemental values of lubricant
thicknesses can be defined in the GEOMETRY option.
Marc also enables the treatment of grooves. Constant groove depth magnitudes can be specified in the GEOMETRY
option. If the groove depth is position-dependent, the contribution to the thickness field can be defined in the UGROOV
user subroutine.
The relative velocity of the moving surfaces is defined on a nodal basis in the VELOCITY option. In addition, you can
redefine the specified nodal velocity components in the UVELOC user subroutine.
Specify prescribed nodal pressure values in the FIXED PRESSURE option. Define restrictor type boundary conditions
in the RESTRICTOR option. To specify nonuniform restrictor coefficients, use the URESTR user subroutine.
Input nodal point mass fluxes using the POINT MASS option. Specify distributed mass fluxes in the DIST MASS
option. If nonuniform fluxes are necessary, apply this via the FLUX user subroutine. The TABLE option may also be
used to define spatially varying boundary conditions.
Define variations of the previously specified lubrication film thickness field through the THICKNS CHANGE option.
The program adds this variation to the current thickness values and solves the lubrication problem.
Activate the calculation of bearing characteristics (that is, damping and stiffness properties) through the DAMPING
COMPONENTS or STIFFNS COMPONENTS options. Marc evaluates these properties based on the specified change
in film thickness. This evaluation requires the formation of a new right-hand-side, together with a matrix back
substitution. This is performed within so-called subincrements. The bearing force components calculated within these
subincrements represent the bearing characteristics (that is, the change in load carrying capacity for the specified
thickness change or thickness rate). The previously specified total thickness is not updated within subincrements. The
calculated bearing characteristics are passed through to the UTHICK user subroutine. This allows you to define an
incremental thickness change as a function of the previously calculated damping and/or stiffness properties. This
procedure can be applied when analyzing the dynamic behavior of a bearing structure. Mechanical problems can often
be represented by simple mass-damper-spring systems if the bearing structure is nondeformable. A thickness
increment can be derived based on the current damping and stiffness properties by investigating the mechanical
equilibrium at each point in time.
The bearing analysis capability deals with only steady-state solution and does not include the analysis of transient
lubrication phenomena. Note that the incrementation procedure is only meant to analyze a sequence of film profiles.
No nonlinearities are involved; each increment is solved in a single step without iteration.
287
CHAPTER 6
To calculate the reaction forces that act on the bearing structure, Marc requires information about the spatial
orientation of the lubricant. This information is not contained in the finite element model because of the planar
representation of the lubricant. Therefore, it is necessary to define the direction cosines of the unit normal vector that
is perpendicular to the lubricant on a nodal basis in the UBEAR user subroutine. The resulting nodal reaction forces
are printed.
Marc requires a step-by-step solution of both the lubrication problem and the stress problem in separate runs. The
thickness changes need to be defined within the lubrication analysis based upon the displacements calculated in the
stress analysis. The stress analysis post file and the UTHICK user subroutine can be used for this purpose. The tractions
to be applied in the stress analysis can be read from the bearing analysis post file in the FORCDT user subroutine.
The flow of a lubricant between two surfaces that move relative to each other is governed by the Reynolds equation
(6-42)
where:
The following assumptions are involved in the derivation of this equation:
The lubricant is a Newtonian fluid; that is, the viscosity is constant.
There is no pressure gradient across the film height.
There is laminar flow.
Inertial effects are negligible.
The lubricant is incompressible; that is, mass density is constant.
Thermal effects are absent.
By introducing the film constant
(6-43)
Equation (6-43) can be written as
(6-44)
is lubricant pressure
is mass density
is film thickness
is viscosity
is time
is the relative velocity vector between moving surfaces
is the mass flux per unit area added to the lubricant
h
3
12
---------- p
\ .
| |
h ( )
t
---------------
1
2
--- hu ( ) M + 0 =
p
t
u
M
h
3
12
---------- =
p ( ) M
r
+ 0 =

where is the reduced mass flux given by
(6-45)
In case of a stationary bearing, the transient term in Equation (6-45) vanishes.
Three different kinds of boundary conditions can be distinguished for the lubrication problem: prescribed pressure on
boundary, prescribed mass flux normal to the boundary, and mass flux proportional to pressure.
Prescribed pressure on boundary is specified as
(6-46)
where is the value of the prescribed pressure.
Prescribed mass flux normal to the boundary has the form
(6-47)
where is the reduced inward mass flux. Here, refers to the inward normal on the boundary, and and are
the inward components of total mass flux and relative velocity, respectively.
If a restrictor is used (as in hydrostatic bearings), the total mass flux is a linear function of the pressure on the boundary.
This condition is specified as
(6-48)
or, written in a slightly different form
(6-49)
where c is the restriction coefficient and
(6-50)
is the reduced pressure.
The differential Equation (6-42), together with the boundary conditions (Equations (6-47), (6-48), and (6-50))
completely describe the lubrication problem. This is analogous to a heat conduction problem as shown in Table 6-3.
Table 6-3 Comparison of Lubrication and Heat Conduction
Lubrication Heat Conduction
Pressure Temperature
Film constant Conductivity
M
r
M
r
M
h ( )
t
---------------
1
2
--- hu ( ) =
p p =
p
p
n
------ m
n
1
2
--- hu
n
m
n
r
= =
m
n
r
n m
n
u
n
m
n

p
n
------
1
2
--- hu
n
+ c p p ( ) = =
p
n
------ c p
r
p ( ) =
p
r
p
hu
n
2c
------------ =
p T
k
289
CHAPTER 6
Electrostatic Analysis
Marc has the capability to perform electrostatic analysis. This allows the program to evaluate the electric fields in a
body or media, where electrical charges are present. This can be solved for one-, two-, or three-dimensional
fields. Semi-infinite elements can be used to represent an infinite domain. The electrostatic problem is governed by
the Poisson equation for a scalar potential. This analysis is purely linear and has been implemented in Marc
analogously to the steady state heat transfer problem. The available elements are described in Table 6-4 below.
Reduced body mass flux
Body heat flux
Reduced boundary mass flux Boundary heat flux
Restriction coefficient Film coefficient
Reduced reference pressure Reference temperature
Table 6-4 Element Types for Electrostatic Analysis
37, 39, 131, 41 3-, 4-, 6-, 8-node planar
69 8-node reduced integration planar
101, 103 6-, 9-node semi-infinite planar
38, 40, 132, 42 3-, 4-, 6-, 8-node axisymmetric
70 8-node reduced integration axisymmetric
43, 44 8-, 20-node 3-dimensional brick
71 20-node reduced integration brick
105, 106 12-, 27-node semi-infinite brick
123 8-node reduced integration brick
135, 133 4-, 10-node tetrahedral
50, 85, 86 3-, 4-, 8-node shell
87, 88 2-, 3-node axisymmetric shell
137, 203 6-, 15-node pentrahedral
Table 6-3 Comparison of Lubrication and Heat Conduction
Lubrication Heat Conduction
M
r
Q
m
n
q
n
c h
p
r
T
r

Marc computes and prints the following quantities: electric potential field vector ( ) and electric displacement vector
( ) at the element integration points. The nodal point data consists of the potential and the charge .
To activate the electrostatic option, use the ELECTRO parameter. The value of the isotropic permittivity constant is
given in the ISOTROPIC option, orthotropic constants can be specified using the ORTHOTROPIC option. Optionally,
the UEPS user subroutine can be used.
Specify nodal constraints using the FIXED POTENTIAL option. Input nodal charges using the POINT CHARGE option.
Specify distributed charges by using the DIST CHARGES option. If nonuniform charges are required, the FLUX user
subroutine can be used for distributed charges and the FORCDT user subroutine for point charges. The TABLE option
may also be used to define spatially varying boundary conditions.
The electrostatic capability deals with linear, steady-state problems only. The STEADY STATE option is used to begin
the analysis. The resultant quantities can be stored on the post file for processing with Marc Mentat.
The Maxwell equations to govern electrostatics are written in terms of the electric displacement vector and the
electric field vector such that
(6-51)
and
(6-52)
where is a given volume charge density.
The constitutive law is typically given in a form as:
(6-53)
where is the dielectric constant.
Introducing the scalar potential such that
(6-54)
which satisfied the constraint Equation (6-52) exactly.
Denoting the virtual scalar potential by , the variational formulation is
(6-55)
The natural boundary condition is applied through the surface integral in terms of the normal electric displacement
using:
(6-56)
E
D Q
D
E
D =
E 0 =
D E =
E =
( )dV
V
}
dV
V
}
n ( )dA
}
+ =
n D n D
n
= =
291
CHAPTER 6
Consider a material interface , separating two materials 1 and 2, and as is continuous over the material interface,
the tangential electric field constraint is automatically satisfied.
(6-57)
If charges are present on the interface, these are applied as distributed loads as follows:
(6-58)
Using the usual finite element interpolation functions and their derivatives , we obtain
(6-59)
(6-60)
and finally
(6-61)
Capacitance
A capacitor is a device that can store electrostatic charge and hence electrostatic potential energy. This energy can be
thought as the energy stored in the electric field created by the charge on the capacitor.Capacitance is always associated
with conductors that can store charge. A problem can contain a number of conductors each of which will store charge.
This gives rise to self and cross capacitance due to interaction between the conductors. This is expressed by the
capacitance matrix.
All conductors in the problem are assumed to be bodies and defined as a set of elements. This is done using the
THERMAL CONTACT model definition option. No two conductors can touch each other or overlap. It is required to
model the inside of the conductor bodies. A sub-set of the conductor bodies can be considered for capacitance
computation. This subset is specified using the EMCAPAC history definition in the THERMAL CONTACT model
definition option. Marc performs capacitance computation only if the above definitions are specified in the input file.
The electrostatic analysis is used for computation of the capacitance matrix. The computation of the capacitance
matrix results from the usual electrostatic analysis with some constraints on boundary conditions:
The boundary of the problem must satisfy far-field conditions and the electric potential on this boundary must
be zero.
At problem boundaries where far field is not satisfied, the homogenous Neumann boundary condition must
apply.
The electrostatic charge boundary condition must not be used on the problem boundary.
Any boundary condition applied on an individual conductor body is ignored.
12

n E
1
E
2
( ) 0 = on
12
n D
1
D
2
( )
s
= on
12
N
N = N =
K
T
dV
V
}
=
F N
T
NdV
V
}
N
T
ND
n
dA
}
N
T
N
s
dA
12
}
+ + =
K F =

When dissimilar meshes are used to model the problem, then the conductors and dielectrics are distinct contact bodies.
In this case it is best that the conductors are the contact bodies which are being touched and the dielectrics are the
touching contact bodies.
Consider a single conducting body placed in an infinite homogenous dielectric medium. This conductor can be
charged by putting a charge on it or applying an electric potential. The charge redistributes on the conductor such that
the electric field inside the conductor is zero. This indicates that these charges reside on the outside surface of the
conductor and there is no charge inside the conductor. A normal electric field is present on the surface of the conductor.
The value of the normal electric field is equal to:
(6-62)
where:
The normal electric displacement is
(6-63)
The total charge on the conductor is:
(6-64)
where defines the surface of the conductor.
The capacitance is then defined as:
(6-65)
where is the constant potential on the conductor.
Capacitance Matrix
For multiple conductors the capacitance computation is expressed as a capacitance matrix. For multiple conductors,
the capacitance matrix is:
is the surface charge density
is the permittivity of the medium surrounding the conductor
E
n
--- =
D
n
E
n
= =
Q dS
S
}
=
S
C
C
Q
V
---- =
V
n
293
CHAPTER 6
(6-66)
where
To find the values of , the analysis has to be repeated times.
To find for any column and , the procedure is as follows:
1. Specify volt and .
2. For this run the values of are computed and the capacitance values of column are
calculated.
Magnetostatic Analysis
Marc has the capability to perform magnetostatic analysis. This allows Marc to calculate the magnetic field in a media
subjected to steady electrical currents. This can be solved for two- or three-dimensional fields. Semi-infinite elements
can be used to represent an infinite domain.
The magnetostatic analysis for two-dimensional analysis is solved using a scalar potential, while for three-dimensional
problems, a full vector potential is used. The magnetic permeability can be a function of the magnetic field, hence,
creating a nonlinear problem. Only steady-state analyses are performed. The available elements which are described
in Table 6-5.
is the self capacitance of conductor , if
is the mutual capacitance between conductor and , if
Table 6-5 Elements Types for Magnetostatic Analysis
37, 39, 131, 41 3-, 4-, 6-, 8-node planar
102, 104 6-, 9-node semi-infinite axisymmetric
109 8-node brick
110 12-node semi-infinite brick
Q
1
Q
2
.
.
Q
n
C
11
C
12
. . C
1n
C
21
C
22
. . C
2n
. . . . .
. . .
. .
C
n1
C
n2
. , C
nn
V
1
V
2
.
,
V
n
=
C
i j
i i j =
C
i j
i j i j
C
i j
n
C
i j
j i 1 2 n , , =
V
j
1 = V
1
V
2
V
i j
V
n
0 = = = = = =
Q
i
i 1 2n , = ( ) j

Marc computes and prints magnetic induction ( ), the magnetic field vector ( ), and the vector potential at the
element integration points. The nodal point data consist of the potential ( ) and the current ( ).
To activate the magnetostatic option, use the MAGNETO parameter. The value of the isotropic permeability () is
given on the ISOTROPIC option; orthotropic constants can be specified using the ORTHOTROPIC option. Optionally,
the UEPS user subroutine can be used. Often, it is easier to specify (1/), which is also available through these options.
A nonlinear permeability can be defined using the B-H relation.
Specify nodal constraints using the FIXED POTENTIAL option. Input nodal currents using the POINT CURRENT
option. Specify distributed currents by using the DIST CURRENT option. Permanent magnets can be introduced by
using the PERMANENT option for isotropic materials, or by entering a remanence vector via the B-H RELATION
option for orthotropic materials. In addition, the FLUX user subroutine can be used for nonuniform distributed current
and the FORCDT user subroutine for fixed nodal potentials or point current. For convenience line element 183 can be
used to define an external loading. The current on this element will be converted to point currents on the nodes,
pointing in the direction of this line element. The TABLE option may also be used to define spatially varying boundary
conditions.
The magnetostatic capability is linear unless a nonlinear B-H relation is defined. In such problems, convergence is
reached when the residual current satisfies the tolerance defined in the CONTROL option. The STEADY STATE option
is used to begin the analysis. The resultant quantities can be stored on the post file for processing with Mentat.
The Maxwell equations for magnetostatics are written in terms of the magnetic flux density vector B such that:
(6-67)
and
(6-68)
where J is the current density vectors.
For magnetic materials, the following relation between , , and , the magnetization vector, holds:
(6-69)
181 4-node tetrahedron
183 2-node 3-D link
204 6-node pentahedral
206 20-node brick
Table 6-5 Elements Types for Magnetostatic Analysis
B H
A J
H J =
B 0 =
B H M
B
0
H M + ( ) =
295
CHAPTER 6
with being the permeability of vacuum. Denoting the magnetic susceptibility by and the permanent magnet
vector by , we have
(6-70)
which can be substituted into Equation (6-69) to yield:
(6-71)
in which is the permeability, given by:
(6-72)
and B
r
is the remanence, given by:
(6-73)
Notice that is usually called the relative permeability .
Figure 6-38 Nonlinear B-H Relationship
For isotropic linear material, and are scalar constants. If the material is orthotropic, and change into
tensors. For real ferromagnetic material, and are never constant. Instead, they depend on the strength of the
magnetic field. Usually this type of material nonlinearity is characterized by a so-called magnetization curve or B-H
relation specifying the magnitude of (a component of) B as a function of (a component of) H.
In Marc, the magnetization curve can be entered via the B-H RELATION option. For isotropic material, only one set of
data points, representing the magnitude of the magnetic induction, , as a function of , the magnitude H, needs
to be given. For orthotropic material, multiple curves are needed with each curve relating a component of B to the
corresponding component of H.
When the table driven input is used, different ways of entering a magnetization curve are available. Then the
ISOTROPIC or ORTHOTROPIC material option contains either the permeability, the inverse permeability, the H-B
relation, or the B-H relation. For ORTHOTROPIC, this can be different for each component. For the H-B relation and
the B-H relation, a table has to be given where for the H-B relation, B is the independent variable, and for the B-H
0

m
M
0
M
m
H M
0
+ =
B H B
r
+ =

0
1
m
+ ( ) =
B
r

0
M
0
=
1
m
+ ( )
r
B
B
r
H
m
M
0

m
M
m
B H

relation, H is the independent variable. Permeability and inverse permeability can also be controlled by a table where
the independent variable can be H, B, or any other variable. A table can be either a set of data points or a function.
From Equations (6-71) and (6-73), it can be seen that for orthotropic materials, a permanent magnetization or
remanence can be entered through the B-H RELATION option, by putting in a nonzero B value for H = 0. For isotropic
material, this does not work since the direction of the remanence vector is still indeterminate. Therefore, in the
isotropic case, the only possibility is to supply the magnetization vector through the PERMANENT option. Any offset
of the -curve, implying is disregarded in this case. For orthotropic material, it is not
allowed to use the PERMANENT option. In this case, the magnetization can exclusively be specified through the B-H
RELATION option or using the table driven input.
It is emphasized that the magnetization curve specified in the B-H relation or H-B relation must be monotonic and
uniquely defined.
Introducing the vector potential A such that:
(6-74)
which automatically satisfies the constraint, Equation (6-68), and we then have the final form:
(6-75)
The vector potential ( ) is not uniquely defined by Equation (6-74). In 2-D magnetostatic simulations, this
indeterminacy is removed by the reduction of to a scalar quantity. In 3-D situations, Marc uses the Coulomb gauge
for this purpose:
(6-76)
Denote the virtual potential by ; then, the variational formulation is:
(6-77)
where is the outward normal to at the boundary .
In the three-dimensional case, the Coulomb gauge, Equation (6-76), is enforced with a penalty formulation. The
resulting term added to the variational formulation, Equation (6-77) reads:
(6-78)
The default value used for is:
(6-79)
B H ( ) B 0 at H ( 0 ) =
A B =

1
A ( ) J
1
B
r
+ =
A
A
A 0 =
W
1
W ( ) A ( )dV
V
}
W JdV W
1
B
r
( )dV W H n ( )d
}
+
V
}
+
V
}
=
n V
1
W ( ) A ( )dV
V
}
W JdV W
1
B
r
( )dV W H n ( )d
}
+
V
}
+
V
}
=
r W ( ) A ( ) V d
V
}
+
r
r 10
4
1
=
297
CHAPTER 6
Using the usual finite element interpolation functions , the discrete curl operator , and the weighting function
.
(6-80)
Leads to the resulting system of algebraic equations
(6-81)
where
(6-82)
(6-83)
The Coulomb gauge is based upon the principle of conservation of electric charge which in its steady state form reads:
(6-84)
In Marc, it is up to you to specify the current distribution. When doing so, it is recommended to ensure that this
distribution satisfies Equation (6-83). Otherwise, condition Equation (6-76) could be violated. As a consequence, the
results could become less reliable.
From the third term on the right-hand side of Equation (6-83), it becomes clear that the natural boundary condition in
this magnetostatic formulation is , the tangential component of the magnetic field intensity. Consequently, if no
other condition is specified by you, by default a zero tangential magnetic field intensity at the boundaries is assumed.
Besides, when there are no currents present on a material interface separating two materials 1 and 2, the tangential
magnetic field intensity is assumed to be continuous:
(6-85)
With and , the magnetic field intensities in material 1 and 2, respectively. A discontinuous tangential magnetic
field intensity can be modeled by assigning an appropriate distributed shear current density to the interface. This
(surface) current density is related to and by:
(6-86)
N G
W N =
G
N y
N x
=
for two-dimensional problems
G
N y N z
N z N x
N x N y
=
for three-dimensional problems
Ku F =
K G
T
1
G V d
V
}
=
F N
T
J V G
T
1
B
r
V N
T
H n ( )
}
+ d
V
}
+ d d
V
}
=
J 0 =
H n
12
n H
1
H
2
( ) 0 on
12
=
H
1
H
2
J H
1
H
2
n H
1
H
2
( ) J on
12
=

Edge Effects at Interface Between Materials of Different Magnetic Permeabilities
At the interface between materials of different permeabilities, there is an abrupt change of permeability value. In
absence of a surface current at the interface, the following holds
and
(6-87)
where
A surface sheet current can flow at the interface only if one of the materials is a perfect conductor ( ). Many
magnetic materials are conducting too but are not perfect conductors.
For any material we have:
(6-88)
where
Hence,
(6-89)
Now,
(6-90)
where
The conditions (equations 6-87, (6-89), and (6-90) are satisfied if the normal component of the magnetic vector
potential A is allowed to be discontinuous along the interface normal. The tangential component of magnetic vector
potential A is continuous.
For 2-D problems, only the scalar Z component of the magnetic vector potential A is considered. This component is
always tangent to any material interfaces. Hence, the edge effect is automatically satisfied for these problems. The
edge effects are required only for three-dimensional magnetostatic problems.
is the magnetic field intensity.
is the magnetic flux density.
corresponds to the normal component
corresponds to the tangential component at interface
1 and 2 corresponds to either side of interface
is the permeability of the material.
is the magnetic vector potential.
B
n1
B
n2
= H
t 1
H
t 2
=
H
B
n
t

B H =
B
t 1
1
--------
B
t 2
2
-------- =
B A =
A
299
CHAPTER 6
In order to obtain accurate results for the magnetic vector potential A, the edge effect requires the imposition of the
Coulombs gauge. The Coulombs gauge is imposed using a penalty factor. The value of the correct value of the
penalty factor depends on the following factors:
1. Type of element: tetrahedral or hexahedral
2. Order of the element: first or second order.
3. Finite element mesh density and on the uniformity or non uniformity of the mesh. A uniform mesh requires a
lesser value.
The value of penalty factor is quite stable only for a linear hexahedral element and this value is 1.0E-04. For all other
elements, this value lies in different ranges and it is difficult to know the correct value for any given problem. It is,
hence, advisable to use linear hexahedral element with a default penalty value of 1.0E-04.
This interface conditions is implemented in Marc through the THERMAL CONTACT option, where materials on either
side of the interface have to be defined as electromagnetic contact bodies. There is now a choice between glued and
touching contact (CONTACT TABLE or CONTACT TABLE with TABLES). With glued contact, all components of the
magnetic vector potential are tied and this is recommended when two dissimilar meshes containing the same material
properties need to be connected. With touching contact, only the tangential components of the magnetic vector
potential are tied and thus satisfying the above interface conditions. This is recommended at interfaces with different
material properties. The glued or touching contact interface condition should be given on the CONTACT TABLE
option. Note that for convenience, Mentat has a special tool to create a disconnected mesh from a uniform mesh which
is called matching boundaries.
Electrical Windings and Coil Current Boundary Condition:
In many electrical problems, a magnetic circuit is excited by electrical windings. A winding is a set of multi-turn coils.
A set of multi-turn coils end at two termination points called terminals. Each coil has the same cross section with
respect to shape and size. Usually, a coil has a rectangular or circular shape. Other shapes are rarely used. Each coil
has a thin insulation coating. The multi-turns are bound together by some insulation material. The set of multi-turn
itself defines a cross section. This is the winding cross section. A winding cross section is usually rectangular,
sometimes circular. A winding cross section contains the coil cross section and required insulation. The current in each
coil is multiplied by the number of turns (this is called Ampere-turns). It is then assumed that this Ampere-turns flows
through the full winding cross-section.
Electrical windings are specified in the following.
1. Windings:
a. The winding path is defined by a contiguous set of nodes or a contiguous set of polylines.
b. Orientation of the winding cross section.
c. Winding cross-section type: rectangular or circular.
d. Length and breadth for rectangular type and radius for circular type.
e. Number of turns of the coil.
f. Coil cross section type: rectangular or circular.
g. Length and breadth for rectangular type and radius for circular type.
h. Direction of the current with respect to the defined path.
i. Associated electromagnetic contact body identification number. This is used for calculation of inductance
of the winding.

2. Coil current boundary condition
This defines the actual current in each coil which requires the following data.
a. The winding to which this boundary condition applies.
b. The magnitude of current in each coil. For harmonic analysis, the coil current has a real/imaginary or
magnitude/phase component.
c. The geometric type and the list of these geometric entities.
The complete winding data is entered in the windings modeling tools. The coil current and the winding association is
entered in the COIL CURRENT boundary condition. To activate this option, use the EMWINDING and COIL CURRENT
model definition options.
Inductance Matrix Computation
An inductor is a device that can store magnetic potential energy. This energy can be thought as the energy stored in
the magnetic field created by the current flowing in the windings.
Self-inductance
The self inductance of a coil is defined as:
(6-91)
where
Mutual-inductance
Two coils, and , are of the same dimensions and are glued to each other, but separated by an insulator. Coils
and

have and turns, respectively. A current

flows through , while the current in coil

is
zero.

produces a magnetic field intensity (H) and flux density (B) and magnetic energy is stored in this field. The
magnetic field intensity (H) depends on the current , the coil dimensions and the material surrounding the coil. The
magnetic flux produced by coil

and also passing through the coil

is obtained from
(6-92)
is the flux linking the coil
is the current in the coil that produces the flux
is the number of coil turns
L
N
I
--------- =
I
N
C
1
C
2
C
1
C
2
N
1
N
2
I
1
C
1
I
2
C
2
I
1
I
1
C
1
C
2
1

2
B dS
} }
= =
coil area
301
CHAPTER 6
The self inductance of each coil is given by:
(6-93)
Then mutual-inductance of coil with respect to coil is given by
(6-94)
If the roles of the two coils are reversed, then
(6-95)
where

is the mutual-inductance of coil with respect to coil
In this case, the self and mutual inductance are related by
Hence,
and (6-96)
and since
(6-97)
If the two coils are loosely coupled; i.e., there is leakage flux, then
and (6-98)
(6-99)
L
11
N
1
1
I
1
-------------- =
L
22
N
2
2
I
2
-------------- =
M
21
C
2
C
1
M
21
N
2
2
I
1
-------------- =
M
12
N
1
1
I
2
-------------- =
M
12
C
1
C
2
L
11
M
21
L
22
M
12
= = =
L
11
M
12
----------
I
2
I
1
---- =
L
22
M
21
----------
I
1
I
2
---- =
M
12
M
21
=
M
12
L
11
L
22
=
L
11
N
1
1
I
1
-------------- = M
12
N
1
k
1
1
I
2
-------------------- =
L
11
M
12
----------
I
2
k
1
I
2
---------- =
k
1
1 <

Similarly, for the second coil,
(6-100)
and since
(6-101)
is called the coupling factor of the coils.
The Inductance Matrix:
For n coils , , , , each coil has self inductance and mutual inductance with the remaining coils. For
multiple conductors, it is better to express inductance as an inductance matrix. For n multiple conductors, the
inductance matrix is:
(6-102)
where
Note that the inductance matrix is symmetric and
.
The terms are also written as . The magnetic
stored energy due to a currents in n coils , , , is:
(6-103)
The total magnetic stored energy can also be written in terms of and as,
(6-104)
is the self inductance of conductor i, if
is the mutual inductance between conductor i and j, if .
L
22
M
21
----------
I
1
k
2
I
2
---------- =
k
2
1 <
M
12
M
21
=
M
12
k L
11
L
22
+
=
k
2
k
1
k
2
= 0 k 1 < <
k
C
1
C
2
C
3
C
n
2
.
.
n
L
11
L
12
. . L
n
L
21
L
22
. . L
2n
. . . . .
. . . . .
L
n1
L
n2
. . L
nn
I
1
I
2
.
.
I
n
=
L
i j
i j =
L
i j
i j
L
i j
L
j i
= L
i j
M
i j
C
1
C
2
C
3
C
n
W
t
1
2
--- L [
11
I
1
2
L
22
I
1
2
L
33
I
1
2
L
nn
I
n
2
2M
12
I
1
I
2
2M
23
I
2
I
3
2M
34
I
3
+ + + + + + + =
2M
13
I
1
I
3
] + +
B H
W
t
1
2
--- B H V d
}
V
} }
=
303
CHAPTER 6
where is the magnetic induction, is the magnetic field, is the total domain, and is the total stored magnetic
energy
For inductance calculations the two expressions for the stored magnetic energy are equivalent, and Marc uses this to
compute the inductance matrix.
The inductance matrix computation requires the windings feature. The current flowing through each inductor has to
be defined by the EMWINDING model definition and COIL CURRENT boundary condition. To achieve this, the
inductors must be defined as THERMAL CONTACT (electromagnetic contact) bodies. Each winding has to refer to one
of these contact bodies using the EMWINDING model definition option. The inductors for which the inductance matrix
will be computed are selected in the EMINDUC history definition option.
Differential Inductance Computation
The differential inductance corresponds to the change in inductance due to perturbation in the exciting currents. For
the set of n coils described above, the flux linkage of any coil is the surface integral of the magnetic flux density
multiplied by the number of turns . The change in inductance can occur with current variation for nonlinear
magnetic materials. Then, the inductance matrix can be described by the flux linkage as follows:
where
The terminal coil voltage vector is defined by:
where is called the differential inductance matrix. The equation shows that the induced voltage is directly
proportional to the differential inductance matrix and the time derivative of the coil current.
Losses in Ferromagnetic Lamination Cores
Materials consisting of multiple layers of ferromagnetic material separated by isolators (laminations) are always used
in all power industry devices; i.e., alternators, synchronous generators, DC generators and motors, transformers,
induction motors, relays and switch gear. Eddy currents and hysteresis is produced due to time varying magnetic flux
density (B) in laminations. For steady state operations, the time variation is assumed sinusoidal.
These laminations are used to drastically reduce eddy current and hysteresis losses in ferromagnetic materials. These
losses reduce the efficiency of the device, cause undesirable heating and require additional devices like cooling etc.
and it is required to predict them accurately. Accurate prediction goes a long way in reducing manufacturing and
prototyping costs.
is the vector of flux linkage in coils.
is the coil current vector.
is the flux linkage vector when the coil currents are zero and accounts for permanent magnet.
B H V W
C
1
N
1
[ ] L I [ ] ( { } I [ ]
0
[ ] + =
[ ]
I [ ]
0
[ ]
V [ ]
[ ]
t
-------------
d L [ ]
d I [ ]
------------ I [ ] L [ ] +
\ .
| |
I [ ]
t
---------- L
d
[ ] I [ ] { }
I [ ]
t
---------- = = =
L
d
[ ]

Core Losses
Core loss in a magnetic material occurs when the material is subjected to a time varying magnetic flux. Traditionally,
core loss, , consists of two components: hysteresis loss, , and eddy current loss, . According to the Steinmetz
equation, measurement and calculation of core losses are normally made with sinusoidal flux density of varying
magnitude and frequency. These measurements and calculations are based on the standard coil and frequency modeled
by a two term function of the form
, (6-105)
where is the frequency of the external magnetic field, is the flux density, , , and are the coefficients, which
depend on the lamination material, thickness, conductivity, as well as other factors. However, this formula is only
applicable under the assumption that the maximum magnetic flux density of 1.0 Tesla is not exceeded and the
hysteresis loop is under the static situation, which is not practical in electrical machines. When the magnetic flux
density is over 1.0 Tesla or the field frequency becomes high, there is a big discrepancy between the calculation based
on (6.97) and experimental results.
Core loss plays an increasingly important role both in the improvement of the quality of electrical steels at the
production stage and in the optimization of their operating conditions. According to the statistical loss theory, the core
loss under sinusoidal flux condition is given by
(6-106)
where is referred to as the excess loss and is related to the material thickness, cross-sectional area, conductivity
and a parameter which describes the material microstructure. The coefficients , , and

vary slightly with the
magnetic flux density and frequency and this variation must be considered to obtain better accuracy. The
coefficients a, b, and c depend on the frequency and this must be considered in the loss calculations.
In the present procedure, the loss coefficients are evaluated from the original loss data supplied by manufacturers. The
coefficients are obtained mathematically using polynomial curves.
In practice, the lamination steel manufacturers do not present the loss coefficients mentioned above. They only provide
the loss curves and tables in watts per kilogram (W/kg) or watts per pound (W/lb.) versus the flux density or frequency
to indicate the combination of hysteresis loss, eddy loss, and excess loss. The classical estimation of core loss (6-105)
is systematically lower than the measured values. Ratios of two or more have been observed. Thus, for the analysis
and calculation of core loss, the accuracy of the computed loss coefficients is essential.
To compute lamination losses the materials in question are defined as THERMAL CONTACT (electromagnetic contact)
bodies. Each such contact body is a stack of lamination held together by bolts, clamps etc. The lamination loss curve
is provided by the manufacturer. The loss curve is actually a set of curves for different operating frequencies. There
must be at least three curves in this set in order to compute lamination losses in Marc. Each curve in this set shows the
variation of the loss with the magnetic induction B. To achieve good accuracy a sufficient number of data points must
be extracted from the curves. The data points must also cover all values of magnetic induction B for the problem in
P
c
P
h
P
e
P
e
P
h
P
e
+ =
k
h
fB
n
k
e
f
2
B
2
+ =
f B k
h
k
e
n
P
e
P
h
P
e
P
a
+ + =
k
h
fB
a bB cB
2
+ + ( )
k
e
f
2
B
2
k
a
f
1.5
B
1.5
+ + =
P
a
k
a
k
h
k
e
k
a
B f
305
CHAPTER 6
question. The data points are entered by the TABLE model definition option. For each lamination body, the loss curve
table should be given on the THERMAL CONTACT option. It is given as the table reference of the scaling factor of the
loss curve. Needless to say that this option only works with the table driven input. The default multiplying factor for
the loss curve is 1.
To activate this feature and to set the operating frequency use the EMLAMIN history definition option in the history
definition section.
Magnetodynamic Analysis
For static or time-invariant electric or magnetic sources, the first two Maxwells equations become decoupled and
result in two types of analyses: Electrostatic and Magnetostatic. If there is a time-varying electric current or magnetic
source, the first two Maxwells equations are coupled. This coupling is partial when the frequency of time variation is
typically less than 1 to 100 MHz.or the wavelength of electromagnetic wave propagation is larger than the typical
dimension of the electrical device. In general, a frequency less than 1 to 100 MHz is considered Low frequency and
then the first two Maxwells equation are partially coupled. The analysis in this Low frequency range is called
magnetodynamics. In magnetodynamic analyses, the displacement current is very small and is neglected. In essence,
the dielectric permittivity has a low value similar to that of air or most commonly occurring materials.
Marc has the capability to perform both transient (dynamic) and harmonic coupled magnetodynamic analysis. This
allows Marc to calculate the electrical and magnetic fields subjected to external excitation. This can be solved for both
two- or three-dimensional fields. A vector potential for the magnetic field is augmented with a scalar potential for the
electrical field. If a transient analysis is performed, the magnetic permeability can be a function of the magnetic field;
hence, a nonlinear problem. The elements available for magnetodynamic analysis are described in Table 6-6.
Table 6-6 Element Types for Magnetodynamic Analysis
Element
Type Description
111 4-node planar
112 4-node axisymmetric
113 8-node brick
228 3-node planar
230 4-node tetrahedral
231 6-node planar
234 8-node quadrilateral

Marc prints the magnetic flux density ( ), the magnetic field vector ( ), electric flux density ( ), and the electrical
field intensity at the integration points. In a harmonic analysis, these have real and imaginary components. The nodal
point data consists of the vector potential , the scalar potential , the charge , and current .
To activate the magnetodynamic option, use the EL-MA parameter. The values of the isotropic permittivity ( ),
permeability ( ), and conductivity ( ) are given in the ISOTROPIC option. Orthotropic constants can be specified
using the ORTHOTROPIC option. Optionally, the UEPS, UMU, and USIGMA user subroutines can be used. A nonlinear
permeability can be defined using the B-H relation.
Specify nodal constraints using the FIXED POTENTIAL option. Input nodal currents and charge using the POINT
CURRENT option. Specify distributed currents by using the DIST CURRENT option and distributed charges by using
the DIST CHARGE option. Nonuniform distributed currents and charges can also be specified by the FORCEM user
subroutine.
The magnetodynamic capability is linear, unless a nonlinear B-H relation is defined. In such problems, convergence
is reached when the residual satisfies the tolerance defined in the CONTROL option. The transient capability is only
available with a fixed time step; use the DYNAMIC CHANGE option to activate this option. The resultant quantities
can be stored on the post file for processing with Marc Mentat.
In magnetodynamic analysis, you can enter the current and/or the charge. In a harmonic analysis, you can enter both
the real and imaginary components or the magnitude and the phase if the table driven input format is used.
235 8-node axisymmetric quadrilateral
236 20-node brick
Table 6-7 Input Options for Magnetodynamicc Analysis
Load Description
Input Options
Model Definition History Definition User Subroutine
Nodal Current
Nodal Charge
POINT CURRENT
POINT CHARGE
POINT CURRENT
POINT CHARGE
FORCDT
Distributed Current DIST CURRENT DIST CURRENT FORCEM
Distributed Charge DIST CHARGE DIST CHARGE FORCEM
Table 6-6 Element Types for Magnetodynamic Analysis
Element
Type Description
B H D
A V Q I

307
CHAPTER 6
Technical Formulation
Magnetodynamic analysis is based upon the well-known Maxwells equations. This has been implemented in Marc
for both transient and harmonic analyses.
Transient Formulation
The Maxwells equations are:
(6-107)
(6-108)
(6-109)
(6-110)
where the constitutive relations are
and
and
Additionally, we have the conservation of charge: where is the charge density. We assume that the
magnetization vector is given by:
is the electric field intensity
is the electric flux density
is the magnetic field intensity
is the magnetic flux density
is the current density
is the magnetization
is the permittivity
is the permeability
is the conductivity
is the permeability of free space.
E B
+ 0 =
H E
E 0 =
E
E + ( ) 0 =
B 0 =
D E =
B
0
H M + ( ) =
J E =
E
D
H
B
J
M
J + 0 =

(6-111)
where is the strength of the permanent magnet and is the susceptibility. The magnetic field can be defined as:
(6-112)
in which is the permeability, given by:
(6-113)
and is the remanence, given by:
(6-114)
Notice that is usually called the relative permeability .
A vector magnetic potential and a scalar potential are introduced, such that
(6-115)
(6-116)
Note that since only the curl of is required, an arbitrary specification of the divergence can be made. The Coulomb
gauge is then introduced.
(6-117)
This is implemented using a penalty condition. It is important to note that depends not only on the scalar potential,
but also upon the vector potential. Hence, interpretation of as the usual voltage can lead to erroneous results.
Substituting into Maxwells equations results in:
(6-118)
(6-119)
It has been assumed that ; in that, the permittivity has a zero time derivative.
Two time stepping schemes have been implemented in Marc for transient magnetodynamics. The default scheme uses
the Newmark-Beta algorithm. This discretizes the second-order hyperbolic equations given in Equations (6-118) and
(6-119) and is preferred in mid-to-high frequency situations where the permittivity-based terms in these equations
have a significant influence in the transient response. For low-frequency systems where the permittivity-based terms
can be insignificant, the Newmark-Beta scheme is sometimes found to produce spurious oscillations in the solutions
for the potential, E and D.
In such situations, a second scheme that uses the Backward-Euler algorithm to discretize first-order equations obtained
by dropping the permittivity-based terms in Equations (6-118) and (6-119) is preferred.
For the Newmark-Beta scheme, the general form is given by:
M H M
0
+ =
M
0

B H B
r
+ =

0
1
m
+ ( ) =
B
r
B
r

0
M
0
=
1
m
+ ( )
r
A V
B A =
E V A
+ ( ) =
A
A 0 =
E
V

1
A
0
M
0
( ) [ ] V
+ ( ) V A
+ ( ) + + 0 =
V
+ ( ) V A
+ ( ) + [ ] 0 =
0 =
309
CHAPTER 6
(6-120)
(6-121)
The particular form of the dynamic equations corresponding to the trapezoidal rule:
(6-122)
results in a symmetric formulation, which is unconditionally stable for linear systems.
For the Backward-Euler scheme, the governing equations are rewritten as
(6-123)
(6-124)
The derivative term in Equations (6-123) and (6-124) is discretized as
(6-125)
In the current formulation, a fixed time step procedure must be used. The time step is defined through the DYNAMIC
CHANGE option.
A
n 1 +
A
n
tA
n
1
2
---
\ .
| |
t
2
A
n
t
2
A
n 1 +
+ + + =
A
n 1 +
A
n
1 ( )t A
n
tA
n 1 +
+ + =
1
2
--- =
1
4
--- =

1
A
0
M
0
( ) [ ] V A
+ ( ) + 0 =
V A
+ ( ) [ ] 0 =
A
n 1 +
A
n 1 +
A
n
t
----------------------------- =

6
Harmonic Formulation
In harmonic analysis, it is assumed that the excitation is a sinusoidal function, and the resultant also has a sinusoidal
behavior. This results in the solution of a complex system of equations. In this case, Maxwells equations become:
(6-126)
(6-127)
(6-128)
(6-129)
where is the excitation angular frequency and .
Additionally, we have the conservation of charge:
(6-130)
where all symbols are the same as in the discussion above regarding transient behavior. Again, a vector potential
and a scalar potential are introduced. In this case, these are complex potentials. Substituting into the Maxwells
equations results in:
(6-131)
(6-132)
where
With a little manipulation, a symmetric complex formulation may be obtained. The excitation frequency is prescribed
using the HARMONIC option.
Note that the capability to extract the natural frequencies of a complex system by modal analysis does not exist in
Marc. The harmonic formulation is assumed to be linear; therefore, you should not include the B-H RELATION option.
Piezoelectric Analysis
The piezoelectric effect is the coupling of stress and electric field in a material. An electric field in the material causes
the material to strain and vice versa. Marc has a fully coupled implementation of piezoelectric analysis, thus
simultaneously solving for the nodal displacements and electric potential. The elements available for piezoelectric
analysis are described in Table 6-8. They can be used in static, transient dynamic, harmonic, and eigenvalue analysis.
The analysis can be geometrically nonlinear but is materially linear. The piezoelectric elements have their equivalent
heat transfer elements, so that they can also be used in a coupled thermal-piezoelectric analysis. Such a coupled
analysis is weakly coupled, and solved using a staggered approach. When piezoelectric elements are used in a contact
analysis with a node of the piezoelectric element touching a segment of another piezoelectric element, then a
multipoint constraint relation is set up for the nodal displacements as well as for the electric potential.
E i B + 0 =
H iE E 0 =
D 0 =
B 0 =
i 1 =
i J + 0 =
A
V

1
A
o
M
o
( ) [ ]
V iA + ( ) + 0 =

V iA + ( ) [ ] 0 =
i + =
311
CHAPTER 6
Marc prints the stresses ( ), the strains ( ), the electric displacement ( ) and the electric field intensity ( ) at the
integration points. The nodal data points consists of the displacements , forces , the potential , and the charge .
To perform a piezoelectric analysis, use the PIEZO parameter. The values of the isotropic, orthotropic and anisotropic
mechanical properties are given in the ISOTROPIC, ORTHOTROPIC, and ANISOTROPIC model definition option,
respectively. The electric constants, and the constants defining the coupling between the mechanical and electric part
can be specified with the PIEZOELECTRIC model definition option.
Specify nodal constraints using the FIXED DISP option or FIXED POTENTIAL option. Input nodal loads using the
POINT LOAD option or nodal charges using the POINT CHARGE option. Specify distributed loads with the DIST
LOADS option and distributed charges with the DIST CHARGE option. Fixed nodal displacements and potentials, or
nodal forces and charges can also be specified by the FORCDT user subroutine, nonuniform distributed loads can be
specified by the FORCEM user subroutine, and nonuniform distributed charges can be specified by the FLUX user
subroutine. The TABLE option may also be used to specify spatially and temporarily varying boundary conditions.
The mechanical equilibrium equation for the piezoelectric effect is:
(6-133)
and the electrostatic equilibrium equation is (see also Equation (6-55):
(6-134)
where
Table 6-8 Element Types for Piezoelectric Analysis
160 4-node plane stress
161 4-node plane strain
163 8-node brick
is the stress tensor
is the strain tensor
is the traction at a point on the surface
is the displacement
D E
u f Q
: V t u A d
}
f u V d
V
}
+ = d
V
}
D E V D n A d
}
=
V
V d
V
}
+ d
V
}
t
u

The constitutive equations to govern piezoelectricity are written for the mechanical behavior as:
(6-135)
and for the electrostatic behavior as:
(6-136)
where
The superscripts and represent coefficients measured at constant electric field, and constant strain, respectively.
The term with gives the electro-mechanical coupling in the two consitutive Equations (6-135) and (6-136).
We approximate the displacements and electrical potential within a finite element as
and
where contains the shape functions and and contain the nodal degrees of freedom. The body forces and
charges, as well as the distributed loads and distributed charges, are interpolated in a similar manner.
The strains and the electric field are given as:
(6-137)
and
(6-138)
where and contain the gradients of .
is the body force per unit volume
is the electric displacement vector
is the electric field vector
is the volume charge density
is the virtual electric field corresponding to the virtual potential .
is the elastic stiffness
is the piezoelectric matrix (stress based)
is the permittivity
f
D
E
E
x
=

L
E
: e E =
D e
T
:
E + =
L
e
E
e
u NU =
N =
N U
B
u
U =
E B
=
B
u
B
N
313
CHAPTER 6
Denoting the virtual displacement by and the virtual potential by the variational formulations can be obtained
by substituting Equations (6-135) and (6-137) into Equation (6-133):
(6-139)
and similarly by substituting Equations (6-136) and (6-138) into Equation (6-134):
(6-140)
The final set of equations in matrix form is then:
(6-141)
Strain Based Piezoelectric Coupling
It is also possible to apply strain based coefficients for the piezoelectric coupling matrix. Then the consitutive
equations are for the mechanical behavior:
(6-142)
and for the electrostatic behavior:
(6-143)
where
When and are given, Marc converts this into stress based properties, where and .
is the elastic compliance
is the piezoelectric matrix (strain based)
is the permittivity (strain based)
U
T
V U
T
B
u
T
V t U A d
}
f U V d
V
}
+ = d
V
}
= d
V
}
U
T
B
u
T
LB
u
U V U
T
B
u
T
eB
V tU A d
}
f U V d
V
}
+ = d
V
}
+ d
V
}
U
T
K
uu
U U
T
K
u
U
T
F
u
= +
E
T
D V
T
B
T
D V d
V
}
= d
V
}
Dn A d
}

V
V d
V
}
+ =
T
B
T
eB
u
U V
T
B
T
B
V Dn A d
}

V
V d
V
}
+ = d
V
}
+ d
V
}
T
K
u
U
T
K
+
T
=
K
uu
K
u
K
u
K
F
u
=
C: d E + =
D d:
E + =
C
d
e L:d =
e
T
d =

Acoustic Analysis
Marc has the capability to perform acoustic analysis in a rigid as well as a deformable cavity. This allows the program
to calculate the fundamental frequencies of the cavity, as well as the pressure distribution in the cavity. This can be
solved for two- or three-dimensional fields.
Rigid Cavity Acoustic Analysis
The acoustic problem with rigid reflecting boundaries is a purely linear problem analogous to dynamic analysis, but
using the heat transfer elements.
The elements which are available for acoustic analysis of a rigid cavity are described in Table 6-9.
Marc computes and prints the following quantities: pressure and pressure gradient at the integration points. The nodal
point data consists of the pressure and the source.
The wave equation in an inviscid fluid can be expressed in terms of the pressure as:
(6-144)
where is the sonic velocity:
(6-145)
where is the bulk modulus and is the density.
Table 6-9 Element Types for Rigid Cavity Acoustic Analysis
37, 39, 131, 41 3-, 4-, 6-, 8-node planar
p
2
p
t
2
--------- c
2
2
p =
c
c K =
K
315
CHAPTER 6
Equation (6-144) can be rewritten as:
(6-146)
Where the source terms are neglected, note that this is analogous to the dynamic equation of motion.
The modeling of rigid reflecting boundaries can be done as follows.
Mathematically, a reflecting boundary is described by:
(6-147)
where is the pressure gradient normal to the reflecting surface.
This is analogous to an insulated boundary in heat transfer. Hence, a reflecting boundary can be modeled by a set of
nodes at the outer surface of the area which are not connected to another part of the medium. A reflecting plate in the
middle of an acoustic medium can be modeled by double nodes at the same location.
To activate the acoustic option, use the ACOUSTIC parameter. The number of modes to be extracted should also be
included on this parameter. The bulk modulus and the density of the fluid are given in the ISOTROPIC option.
Specify nodal constraints using the FIXED PRESSURE option. Input nodal sources using the POINT SOURCE option.
Specify distributed sources using the DIST SOURCES option. If nonuniform sources are required, apply these via the
FLUX user subroutine or use the TABLE option.
To obtain the fundamental frequencies, use the MODAL SHAPE option after the END OPTION. The nodes can be used
in a transient analysis by invoking the DYNAMIC CHANGE option. The point and distributed sources could be a
function of time.
Fluid Mechanics
Marc has the capability to perform fluid flow analysis. Marc solves the Navier-Stokes equations in the fluid under the
restrictions that the fluid is considered to be nonreactive, incompressible, single phases, and laminar. The capabilities
in Marc can be applied to four different problems listed below:
Fluid behavior only
Fluid-thermal coupled behavior
Fluid-solid coupled behavior
Fluid-thermal-solid coupled behavior
Note:
Where there are no boundary conditions applied, there is a zero-valued eigenvalue, corresponding to a
constant pressure mode. Hence, you need to have a nonzero initial shift point.
K
2
p
2
p
t
2
--------- 0 =
p
n
------ 0 =
p
n
------

Mass Conservation
The principle of mass conservation for a single-phase fluid can be expressed in differential form as:
(6-148)
where is the mass density, is the Eulerian fluid velocity, and is the time.
Momentum Conservation
The principle of conservation of linear momentum results in:
(6-149)
where is the stress tensor and is the body force per unit mass.
In general, the stress tensor can be written as the sum of the hydrostatic stress and the deviatoric stress as.
(6-150)
(6-151)
In fluid mechanics, the fluid pressure is introduced as the negative hydrostatic pressure:
(6-152)
Energy Conservation
For incompressible fluids, the principle of conservation of thermal energy is expressed by:
(6-153)
where is the temperature, is the specific heat at constant pressure, is the thermal flux, and is the internal
heat generated.
Equation of State
The equation of state for a homogeneous, single-phase gas is:
(6-154)
m is the mass of the fluid, is the molecular weight of the gas, is the universal gas constant, is the volume that
the gas occupies.
D
Dt
-------- v + 0 =
v t
v
t
------ v v +
\ .
| |
f + =
f
t r ( )
3
-------------
1
3
---
i j
i j
=
i j
1
3
---
i j
i j

i j
t r ( )
3
-------------
i j
= =
p
tr
3
-------- =
C
p
T
t
------- v
i
T
x
i
-------- +
\ .
| |
q
i
x
i
-------- H + =
T C
p
q
i
H
pV
mRT
M
------------- =
M R V
317
CHAPTER 6
The gas constant is often expressed as for each substance. The equation of state is then written as:
(6-155)
In Marc, it is assumed that the fluid is incompressible. In such cases, the density is constant
Constitutive Relations
The shear strain rate tensor is defined:
(6-156)
For viscous incompressible fluids, it is assumed that there is an expression:
(6-157)
where is the dynamic viscosity.
If the viscosity is not a function of the strain rate, the material is considered Newtonian. Viscous fluid flow is usually
characterized by the Reynolds number :
(6-158)
where is a typical velocity and is the characteristic length.
Viscosity for common fluids like air and water is fairly constant over a broad range of temperatures. However, many
materials have viscosity which is strongly dependent on the shear strain rate. These materials include glass, concrete,
oil, paint, and food products. There are several non-Newtonian fluid models in Marc to describe the viscosity as
described below.
Piecewise Non-Newtonian Flow
is the equivalent strain rate = (6-159)
, where is a piecewise linear function.
Bingham Fluid
A Bingham fluid behaves as a rigid fluid if the stress is below a certain level, labeled the yield stress, and behaves
in a nonlinear manner at higher stress levels.
if (6-160)
if (6-161)
R R M =
P
RT
-------- =

o
=
i j
1
2
---
v
i
x
j
--------
v
j
x
i
-------- +
\ .
| |
=
i j
d
2
i j

i j
= =
Re
Re
Lv
---------- =
v L
2
3
---
i j
i j
\ .
| |
1 2
=

( ) =
i j
d
i j
g
i j

+ = g
i j
0 = g <

The effective viscosity is and
The materials that can be simulated with this model are cement, slurries, and pastes.
Power Law Fluid
The fluid is represented as:
(6-162)
where is a nondimensional constant.
This model is useful for simulating flow of rubber solutions, adhesives, and biological fluids.
Carreau Model
This model alleviates the difficulties associated with the power law model and accounts for the lower and upper
limiting viscosities for an extreme value of the equivalent shear strain rate.
(6-163)
where and are the viscosity at time equals zero and infinity, respectively.
The thermal flux is governed by the Fourier law.
(6-164)
Finite Element Formulation
In the procedures developed, the fluid-thermal coupling can be treated as weak or strong. In the weak formulation, the
solution of the fluid and thermal equilibrium equations are solved in a staggered manner. In the strongly coupled
approach, a simultaneous solution is obtained.
This section cannot cover all of the details of the finite element formulation. It does discuss the final form of the
linearized set of equations and some of the consequences. Beginning with degrees of freedom in a system namely,
v, p, and T for the mixed method we utilize the traditional finite element interpolation functions to relate the values
within the element to the nodal values.
Because the pressure stabilizing Petrov-Galerkin (PSPG) method is employed in Marc, equal order interpolation
functions can be used for the velocity and the pressure. The method of weighted residuals is used to solve the coupled
Navier-Stokes equations. Based upon the conservation laws of momentum (mass and energy), we obtain first order
differential equations in the form:
(6-165)
(6-166)

o
g
+ =
1
2
---
i j
i j
\ .
| |
1
2
---
=
i j
d
o
K
( )
n 1
i j
=
K

( ) 1
2
2
+ ( )
n 1 ( ) 2
+ =
q
i
k
i j
T
x
j
-------- =
Mv
A v ( )v K T v , ( )v Cp B T ( )T + + + F t ( ) =
C
T
v 0 =
319
CHAPTER 6
(6-167)
where the first equation is obtained from momentum, the second from mass, and the third from energy conservation,
respectively. This is expressed in matrix form as:
(6-168)
if the penalty method is employed
(6-169)
resulting in
(6-170)
The introduction of the streamline upwinding technique (SUPG) developed by Brooks and Hughes is a major
improvement for the stability of the fluid equations. This procedure controls the velocity oscillations induced by the
advection terms. Effectively, this procedure adds artificial viscosity to the true viscosity. The second stabilization
method, PSPG, allows equal order velocity and pressure interpolation functions to be used without inducing
oscillations in the pressure field. The PSPG term is not included when using the penalty formulation. The details of
these stabilization terms are not provided here, but note that the magnitude of the contributions are dependent upon
the element size, viscosity, and the time step. Including these contributions, the semi-discrete set of equations takes
the form:
(6-171)
The terms is included for SUPG and and for PSPG. The other terms are neglected, leading to:
(6-172)
These submatrices can be interpreted as follows:
represents advection of momentum
represents advection of energy
NT
D v ( )T L T ( )T + + Q v t , ( ) =
M 0 0
0 N 0
0 0 0
v
)

`

A K + B C
0 D L + 0
C
T
0 0
v
T
p
)

`

+
F
Q
0
)

`

=
p M
P
1
C
T
v =
M 0
0 N
v
)
`

A K CM
P
1
C
T
+ + B
0 D L +
v
T )
`

+
F
Q )
`

=
M M
+ 0 0
M
N 0
0 0 0
v
)

`

A K K
+ + B C C
+ ( )
K
D L + 0
C
T
0 C
v
T
p
)

`

+
F F
+
Q Q
+
0 )

`

=
K
d
C
e
G
e
M 0 0
0 N 0
0 0 0
v
)

`

A K K
+ + B C
0 D L + 0
C
T
0 C
v
T
p
)

`

+
F
Q
0
)

`

=
A
D

Note that typically , , , , , and are dependent upon the temperature. If a non-Newtonian fluid is used,
, , , and are dependent upon the rate of strain. If the fluid is subjected to large motions, can also depend
on the total displacement.
In terms of physical parameter, excluding geometry, observe:
where:
represents diffusion of momentum or viscosity matrix
represents diffusion of energy or conductance matrix
represents mass
represents heat capacitance
represents buoyancy
represents SPUG stabilization matrix
represents gradient matrix
represents divergence matrix
represents PSPG stabilization matrix
represents externally applied forces
represents externally applied fluxes
is the initial density
is the specific heat
is the thermal conductivity
K
L
M
N
B
K
C
C
T
C
e
F
Q
D K B L Q F
A K D G F
A A
o
( ) =
D D
o
C
p
, ( ) D
o
C
p
T , , ( ) = =
K K ( ) K T , ( ) = =
L L K ( ) L K T , ( ) = =
M M
o
( ) =
N N
o
C
p
, ( ) =
B B
0
g B , , ( ) =
o
C
p
K
321
CHAPTER 6
Penalty Method
The penalty method is an alternative method to satisfy the incompressibility constraints. The objective is to add
another term to the operator matrix (viscosity matrix) such that incompressibility is satisfied. Effectively, is
replaced with where:
(6-173)
where is a large number and the penalty matrix. can be written as:
(6-174)
where and are functions of the geometry and shape functions only. The value of is typically between
to .
The penalty method should not be used for three-node planar elements, or four-node tetrahedral elements.
It is possible to simplify the equations governing fluid flow by assuming that the time derivatives of the velocity and
temperature are zero. This is achieved by using the STEADY STATE option. For thermally coupled, non-Newtonian
flows, you still have a highly nonlinear system and multiple iterations are required.
Transient Analysis
The equations discussed above are still differential equations because of the presence of time derivatives of velocity
and temperature. The conversion of the equations to fully discrete linear equations requires an assumption of the
behavior during the increment. The approximations inevitably result in problems with accuracy, artificial damping,
and/or stability problems.
Marc uses the first order backward Euler procedure such that:
(6-175)
This is substituted into the equations to yield:
(6-176)
is the gravity
is the coefficient of thermal expansion
g
B
K
K
c
K
c
K K
p
+ =
K
p
K
p
K
p
CM
p
1
C
T
=
C M
p
10
5
10
9
v
n 1 +
v
n 1 +
v
n
( ) t =
A K K
M t + + + B C
0 D L N t + + 0
C
T
0 C
v
T
p
)

`

F M t v
n

Q Q
N t T
n
+
0 )

`

=

Marc allows either fixed time step or adaptive time step procedure.
The equations of fluid flow are highly nonlinear even for Newtonian behavior because of the inclusion of the advection
terms. The Newton-Raphson and direct substitution procedures are available to solve these problems.
Solid Analysis
Solid can be modeled in two ways:
The first is to model them as a fluid with a very large viscosity. In such cases, an Eulerian procedure is used
throughout the model. The constitutive laws are limited to the fluid relationships.
The second is to model them as true solids. In such cases, a Lagrangian or updated Lagrangian approach is
used in the solid.
The complete Marc constitutive routines are available for representing the solid behavior.
An important consideration is the interface between the fluid and solid. Large structural deformation or fluid motion
are not supported in the coupled Fluid-Solid capability. For small changes in the structural deformation, the nodes at
the interface are updated. The pressure is transmitted between the fluid and the solid.
Solution of Coupled Problems in Fluids
The coupled fluid-thermal problem can be solved using either a tightly coupled procedure as shown above or in a
staggered manner. Problems such as free convection are inherently coupled and are best solved using the tightly
coupled procedure.
The fluid-solid interaction problem is solved in a weakly coupled manner. Every attempt has been made to maximize
the computational efficiency. It should be noted that, while the staggered procedure can result in more global iterations,
there are several motivating factors for solving the system in this manner.
In solving fluid problems, especially in three dimensions, very large systems of equations are obtained. Any
procedure that reduces the number of equations (for example, excluding the solid) is potentially beneficial.
The fluid flow solution always requires the solution of a nonsymmetric system. In a weakly staggered analysis,
the structural problem can still be solved with a symmetric solver.
When using an iterative solver, the solution of the strongly coupled system will result in an ill-conditioned
system, which results in poor convergence. This is because the terms associated with the fluid, thermal, and
structural operator are of several different orders of magnitude.
The problem can be divided into two regions (fluid and solid) to accommodate these requirements. In each region, a
different solver can be invoked; different nodal optimizers can be used; and a different memory allocation can be
performed. Often, the fluid uses an out-of-core nonsymmetric solver while the structure uses an in-core symmetric
solver.
Degrees of Freedom
The degrees of freedom in a fluid analysis are the velocities or, for the mixed method, the velocity and the pressure.
When the pressure is not included explicitly as a degree of freedom, the incompressibility constraint is imposed using
a penalty approach.
323
CHAPTER 6
If a strongly coupled fluid-thermal problem is solved, the degrees of freedom are either the velocities and temperatures
or the velocities, pressure, and temperature. For a three dimensional problem, the number of degrees of freedom could
be 5.
The degrees of freedom associated with the solid are the conventional displacements or the displacements and
temperatures.
Element Types
The fluid region can be represented using the conventional displacement elements in Marc. When using the mixed
method, the pressure has the same order and interpolation functions as the velocity. In a coupled fluid-thermal analysis,
the temperature also has the same order as the velocity. Hence, the element types available are:
Planar
Axisymmetric
Three Dimensional
The three-node triangular elements or the four-node tetrahedral element give poor results when used with the penalty
formulation.
Type
4-node isoparametric bilinear 11
4-node isoparametric bilinear reduced integration 115
6-node isoparametric triangle 125
8-node isoparametric biquadratic 27
8-node isoparametric biquadratic reduced integration 54
Type
4-node isoparametric bilinear 10
4-node isoparametric bilinear reduced integration 116
6-node isoparametric triangle 126
8-node isoparametric biquadratic 28
8-node isoparametric biquadratic reduced integration 55
Type
8-node trilinear brick 7
8-node trilinear brick with reduced integration 117
20-node brick 21
20-node brick with reduced integration 57

The shape functions used are identical to those used for structural analyses and can be found in any finite element
textbook.
All of the mesh generation capabilities in Mentat can be used to generate the fluid mesh. Furthermore, Mentat can be
used to visualize the results in a manner consistent with structural analyses.
Marcs implementation of the Navier-Stokes equations utilizes the natural boundary condition.
At nodal points, you can prescribe the time dependent values of the velocity and temperature. This can be achieved by
using the FIXED VELOCITY, FIXED TEMPERATURE, VELOCITY CHANGE, and/or TEMP CHANGE options. The
FORCDT user subroutine can be used for time-dependent behavior or the TABLE option.
Gravity and centrifugal loads can also be applied.
For coupled fluid-solid interaction problems, the pressure of the fluid is automatically applied to the structure. The
resultant deformation of the structure is applied to the fluid boundaries. In the current release, only small deformations
of the solid are permitted.
The viscosity, mass density, conductivity, and specific head are defined through the ISOTROPIC option. The STRAIN
RATE or TABLE option is used to define non-Newtonian viscous behavior. The penalty parameter can be entered
through the PARAMETERS option.
Coupled Analyses
The definition of coupled systems includes the multiple domains and independent or dependent variables describing
different physical systems. In the situation with multiple domains, the solution for both domains is obtained
simultaneously. Similarly, the dependent variables cannot be condensed out of the equilibrium equations explicitly.
Coupled systems can be classified into two categories:
1. Interface variables coupling: In this class of problems, the coupling occurs through the interfaces of the domain.
The domains can be physically different (for example, fluid-solid interaction) or physically the same but with
different discretization (for example, mesh partition with explicit/implicit procedures in different domains).
Figure 6-39 Fluid-structure Interaction (Physically Different Domains)
Note:
In fluid analysis, data for POINT LOAD and DIST LOADS should be prescribed as total rather than
incremental quantity (as used in mechanical analysis). Similarly, POINT FLUX and DIST FLUXES for heat
transfer analysis are also given as total quantity. This specification is to be used consistently for the fluid
and/or heat transfer portion of analysis in coupled fluid-solid, fluid-thermal, and fluid-thermal-solid.
325
CHAPTER 6
2. Field variables coupling: In this, the domain can be the same or different. The coupling occurs through the
governing differential equations describing different physical phenomenon; for example, coupled
thermo-mechanical problems.
Figure 6-40 Metal Extrusion with Plastic Flow Coupled with Thermal Field
Marc can solve the following types of coupled problems: fluid-solid, fluid-thermal, fluid-solid-thermal interaction,
piezoelectric, electrostatic-structural, thermo-electrical (Joule heating), thermo-mechanical, electrical-thermal-
mechanical (Joule Mechanical), fluid-soil (pore pressure), magnetodynamic, and magnetodynamic-thermal. The type
of coupling is summarized in Table 6-11.
Table 6-10 Summary of Coupled Procedures
Interaction Category Coupling
Thermal- Mechanical Field Weak
Fluid-Solid
added mass approach
general approach
Interface
Interface
Weak
Weak
Fluid-Thermal
channel approach
general approach
Interface
Field
Weak
Strong or Weak
Fluid-Thermal-Solid Field/Interface Strong or Weak/Weak
Piezoelectric Field Strong
Electrostatic-Structural Field Weak
Magnetostatic-Structural Field Weak
Thermal-Electric (Joule) Field Weak
Electrical-Thermal- Mechanical Field Weak
Magnetostatic-Thermal Field Weak
Fluid-Pore Field Strong
Magnetodynamic Field Strong
Magnetodynamic-Thermal Field Weak

6
There are two approaches for solving fluid-solid interaction problems. In the first approach, the fluid is assumed to be
inviscid and incompressible. The effect of the fluid is to augment the mass matrix of a structure. Modal shapes can be
obtained for a fluid/solid system; the modal superposition procedure predicts the dynamic behavior of the coupled
fluid/solid system. This prediction is based on extracted modal shapes. This method is discussed later.
The general fluid-solid capability models nonlinear transient behavior and is discussed in the fluid mechanics section.
There are two approaches for solving fluid-thermal interaction problems. The first is for a fluid constricted to move
through thin areas and is implemented using the CHANNEL option. This is discussed earlier in this chapter. The second
approach solves the coupled Navier-Stokes equations and is discussed in the fluid mechanics section.
The fluid-thermal-mechanical capability is described in more detail in the fluid mechanics section.
In the coupled thermo-electrical problem, the coupling takes place through the temperature-dependent electrical
conductivity in the electrical problem and the internal heat generation caused by electrical flow in the thermal problem.
The program solves for the voltage and temperature distribution.
Similarly, the coupling between the thermal and mechanical problems takes place through the temperature-dependent
material properties in the mechanical (stress) problem and the internal heat generation in the mechanical problem
caused by plastic work, which serves as input for the heat transfer problem. The temperature distribution and
displacements are obtained.
In each of the coupled problems described above, two analyses are performed in each load/time increment. Iterations
can also be carried out within each increment to improve the convergence of the coupled thermo-electrical and
thermo-mechanical solutions.
In the coupled fluid-soil model, the fluid is assumed to be inviscid and incompressible. The effect of the fluid is to
augment the stress in the soil material to satisfy equilibrium, and to influence the soils material behavior.
In the magnetodynamic analysis, the partially coupled Maxwells equations are solved.
In the latter two analyses, the equations (fluid flow/structural or electrical/magnetic) are solved simultaneously.
Thermal Mechanically Coupled Analysis
Many operations performed in the metal forming industry (such as casting, extrusion, sheet rolling, and stamping) can
require a coupled thermo-mechanical analysis. The observed physical phenomena must be modeled by a coupled
analysis if the following conditions pertain:
The body undergoes large deformations such that there is a change in the boundary conditions associated with
the heat transfer problem.
Deformation converts mechanical work into heat through an irreversible process which is large relative to
other heat sources.
In either case, a change in the temperature distribution contributes to the deformation of the body through thermal
strains and influences the material properties.
Marc has a capability that allows you to perform mechanically coupled analysis. This capability is available for all
stress elements and for small displacement, total Lagrangian, updated Lagrangian, or rigid plastic analysis.
The COUPLE parameter is used to invoke this option. When defining the mesh, if you specify the element as a stress
type through the CONNECTIVITY option, Marc generates an associated heat transfer element, if possible. The region
327
CHAPTER 6
having an associated heat transfer element has coupled behavior. If you specify the element as a heat transfer type
through the CONNECTIVITY option, that region is considered rigid. Only heat transfer is performed in that region.
Marc uses a staggered solution procedure in coupled thermo-mechanical analysis. It first performs a heat transfer
analysis, then a stress analysis. Use the CONTROL option to enter the control tolerances used in the analysis.
Depending on the type of stress analysis performed, Marc can perform three different types of coupled analysis:
Quasi-static coupled analysis: This comprises of a transient heat transfer pass and a static mechanical pass.
Fixed stepping can be specified using TRANSIENT NON AUTO and adaptive stepping can use TRANSIENT or
AUTO STEP. AUTO STEP is preferred over the TRANSIENT option.
Creep coupled analysis: This comprises of a transient heat transfer pass and a creep mechanical pass. Fixed
stepping can be specified using CREEP INCREMENT and adaptive stepping can use AUTO CREEP or AUTO
STEP
Dynamic coupled analysis: This comprises of a transient heat transfer pass and a dynamic mechanical pass.
Fixed stepping can be specified using DYNAMIC CHANGE and adaptive stepping can use AUTO STEP.
Load control and time step control can be specified in either of two ways:
A fixed time step/load size can be specified by using the TRANSIENT NON AUTO option, the CREEP
INCREMENT option, or the DYNAMIC CHANGE option. In these cases, mechanical loads and deformations are
incremental quantities that are applied to each step. Fluxes are total quantities. Time variations for the
mechanical and thermal loads can be specified using appropriate s.
An adaptive time step/load size can be specified by using the AUTO CREEP option or the AUTO STEP option.
In this case, mechanical loads and deformations are entered as the total quantities that are applied over the load
set. Fluxes are total quantities. By default, mechanical loads are linearly increased while thermal loads are
applied instantaneously.
Exercise caution when applying boundary conditions. Use the FIXED DISP, FIXED TEMPERATURE, DISP
CHANGE, TEMP CHANGE, or TABLE options for mechanical or thermal boundary conditions, respectively.
There are two primary causes of coupling. First, coupling occurs when deformations result in a change in the
associated heat transfer problem. Such a change can be due to either large deformation or contact. Large deformation
effects are coupled into the heat transfer problem only if the LARGE STRAIN or UPDATE parameter is invoked. The
gap element in Marc (Type 12) has been modified so that if no contact occurs, the gap element acts as a perfect
insulator. When contact does occur, the gap element acts as a perfect conductor.
The second cause of coupling is heat generated due to inelastic deformation. The irreversibility of plastic flow causes
an increase in the amount of entropy in the body. This can be expressed as:
(6-177)
where is the fraction of the rate of plastic work dissipated into heat. Farren and Taylor measure f as
approximately 0.9 for most metals. Using the mechanical equivalent of heat ( ), the rate of specific volumetric flux
is:
(6-178)
TS
fW
p
=
fW
p
M
R MfW
p
=

Use the CONVERT model definition option to define .
Of course, all mechanical and thermal material properties can be temperature dependent. The governing matrix
equations can be expressed as:
(6-179)
(6-180)
In Equation (6-180), represents the amount of heat generated due to plastic work and

represents the amount
of heat generated due to friction. The specific heat matrix and conductivity matrix can be evaluated in the
current configuration if the updated Lagrange option is used.
Coupled Acoustic-Structural Analysis
In a coupled acoustic-structural analysis, both the acoustic medium and the structure are modeled. In this way, the
effect of the acoustic medium on the dynamic response of the structure and of the structure on the dynamic response
of the acoustic medium can be taken into account. Such a coupled analysis is especially important when the natural
frequencies of the acoustic medium and the structure are in the same range. In Marc, only a harmonic coupled
acoustic-structural analysis can be performed. Since the interface between the acoustic medium and the structure is
determined automatically by Marc based on the CONTACT option, setting up the finite element model is relatively
easy since the meshes do not need to be identical at the interface.
The acoustic medium will be called the fluid, although it might also be a gas, and is considered to be inviscid and
compressible. The equilibrium equation is given by:
(6-181)
in which is the pressure, is the fluid density, is the displacement vector and a superposed dot indicates
differentiation with respect to time.
Using the bulk modulus of the fluid , the constitutive behavior can be written as:
(6-182)
The Equations (6-181) and (6-182) can be combined to:
(6-183)
It is also possible to add a damping term to Equation (6-181), with the resistivity or fluid drag:
(6-184)
Note: All terms except can be temperature dependent.
Mf
Mu
Du
K T u t , , ( )u + + F =
C
T
T ( )T
K
T
T ( )T + Q Q
I
Q
F
+ + =
Q
I
Q
F
C
T
K
T
M
p
f
u
+ 0 =
p
f
u
K
f
p K
f
u =
1
K
f
------ p
f
----- p
2
0 =
ru
r
p ru
f
u
+ + 0 =
329
CHAPTER 6
Since we restrict the discussion to harmonic analyses, we can write , with the excitation frequency and
the imaginary unit, so that Equation (6-184) reduces to:
(6-185)
in which the complex density is given by:
(6-186)
The weak form of Equation (6-185) is obtained in a standard way by introducing the variational field and
integrating over the fluid volume :
(6-187)
Applying the Greens divergence theorem to this expression yields:
(6-188)
in which represents the boundary of the fluid with an inward normal .
Along the boundary, various conditions can occur:
The pressure is prescribed on the boundary .
The normal acceleration is prescribed to be on the accelerating boundary .
The normal acceleration of the fluid equals the normal acceleration of the solid, , on the fluid-
structure interface .
Nonreflecting or reactive boundary conditions are introduced using a spring-dashpot analogy on as
. In this way, a spring and a dashpot are placed in series between the acoustic medium
and its boundary, with the spring and the dashpot parameter, both per unit area. Some classical
conditions can be modeled by this expression:
1. Free surfaces waves in a gravity field: , with the gravity acceleration;
2. The plane wave radiation boundary condition: , with the wave velocity;
3. Using the complex admittance of the boundary, with the impedance, the normal velocity can
be related to the pressure by , so that a given can be modeled by
setting and .
u
i u =
i
1
K
f
------ p
f
----- p
2
0 =
f

f
ir
----
\ .
| |
=
p
V
p
1
K
f
------ p
f
----- p
2
\ .
| |
V d
V
}
0 =
p
1
K
f
------ p
f
----- p p +
\ .
| |
V d
V
}
1
f
----- p
p
n
----- A d
}
+ 0 =
n
p
p
a
n

a
u
f n
u
sn
=
f s
f n
1
---
p
n
----- p
c
I
p
k
I
+ =
k
I
c
I
p
n
----- p
g = g
p
n
----- p
c = c
1 z ( ) z ( )
u
f n
p z ( ) 1 c
I
i k
I
+ ( )p = = z ( )
1 c
I
1 k
I

Upon discretizing the fluid and the solid using finite elements, the fluid pressure and the displacements of the solid
can be approximated in a standard way, so that Equation (6-188) yields:
(6-189)
where and contain the pressure and displacement degrees of freedom, contains the interpolation functions and
is a transformation matrix to relate the displacements in normal direction to the global displacement components.
Since Equation (6-189) must be valid for arbitrary admissible values of , evaluating the various integrals and using
, and , results in:
(6-190)
In order to end up with a set of coupled equations, we need to consider the solid as well. The effect of the fluid on the
solid originates from the fluid pressure at the fluid-solid interface. This can be easily included by evaluating the virtual
work corresponding to the fluid pressure:
(6-191)
Notice that the minus sign has been used to express that is positive in the outward direction of the solid. Including
Equation (6-191), the solid behavior for a harmonic analysis is governed by:
(6-192)
with the stiffness matrix, which can include stress-stiffening, is the damping matrix, the mass matrix, and
the external load vector, except for the fluid pressure.
Premultiplying Equation (6-190) by and combining the result with Equation (6-192) gives the desired coupled
complex equation system:
(6-193)
in which and .
p
T
N
p 1
c
I
---- N
pT
p
N
p 1
k
I
---- N
pT
p
+
\ .
| |
A d
f n
}
+ 0 =
p
T
N
p 1
K
f
------ N
pT
p
p
T
N
p
( )
1
f
----- N
pT
( )p +
)
`

V d
V
}
p
T
N
p
a
n
A p
T
N
p
N
uT
Gu
A d
f s
}
d
a
}
p u N
G
p
p
ip = p
2
p = u
2
u =
K
f
i C
f

2
M
f
+
)
`

p
2
S
f s
u + F
f
=
W
p
u
ns
p A d
f s
}
u
T
G
T
N
u
N
pT
p A d
f s
}
u
T
S
f s
T
p = = =
u
ns
K
s
iC
s

2
M
s
+
)
`

u S
f s
T
p + F
s
=
K
s
C
s
M
s
F
s
2
A
f
S
f s
S
f s
T
A
s
p
u
2
F
f
F
s
=
A
f
K
f
iC
f
+ = A
s
K
s
i C
s
+ =
331
CHAPTER 6
The procedure to perform a coupled acoustic-structural is as follows. The acoustic medium and the structure are
modeled separately; the acoustic structure using heat transfer elements with acoustic material properties and the
structure using conventional stress elements. The elements representing the acoustic medium are assigned to an
acoustic contact body and the elements representing the solid to a deformable contact body. If the acoustic medium is
not surrounded completely by a deformable structure and one wants to, for example, model radiation boundary
conditions along this part of the acoustic medium, then one can define rigid contact bodies in this area. Like
deformable bodies, rigid bodies can be used to define the spring-dashpot analogy defined before. A harmonic analysis
can be performed based on initial contact, so using the undeformed configuration of the contact bodies, but also after
a preload of the deformable contact bodies. In the latter case, it might be necessary to remesh the acoustic and/or
deformable bodies before the harmonic analysis can be performed, since during the static loading of the deformable
bodies the deformation of the acoustic bodies is not taken into account.
The elements which are available for coupled-structural acoustic analysis are described in Table 6-11.
Marc computes and prints the following quantities: pressure and pressure gradient at the integration points. The nodal
point data consists of the pressure and the source.
The ACOUSTIC parameter is used to indicate that a coupled acoustic-structural analysis is performed. In addition to
the CONTACT option, the ACOUSTIC and REGION model definition options are used to define the material properties
of the acoustic medium and to set which elements correspond to the solid and the fluid region. Typical boundary
conditions for the acoustic medium can be entered using the FIXED PRESSURE model definition and PRESS
CHANGE history definition option. The DIST SOURCES and POINT SOURCE model definition and DIST SOURCES
and POINT SOURCE history definition options are used to define the load on the acoustic medium. The HARMONIC
history definition option is used to define the excitation frequencies. Spatially and frequency dependent boundary
conditions may be defined by referencing the TABLE option.
Fluid/Solid Interaction Added Mass Approach
The fluid/solid interaction procedure investigates structures that are either immersed in, or contain a fluid. Examples
of problems that make use of this feature are vibration of dams, ship hulls, and tanks containing liquids.
Marc is capable of predicting the dynamic behavior of a structural system that is subject to the pressure loading of
fluid. The fluid is assumed to be inviscid and incompressible; for example, water.
In Marc, the fluid is modeled with heat transfer elements (potential theory) and the structure is modeled with normal
stress or displacement elements. The element choice must ensure that the interface between the structural and fluid
Table 6-11 Element Types for Coupled-Structural Acoustic Analysis
37, 39 3-, 4-node planar
38, 40 3-, 4-node axisymmetric

models has compatible interpolation; that is, both solid and fluid elements are either first order or second order. The
TYING option can be used to achieve compatibility if necessary. To identify the interfaces between the fluid and the
structure, the FLUID SOLID model definition set is necessary.
During increment zero, Marc calculates the stiffness matrix for the structure and the mass matrix for the structure
augmented by the fluid effect. Marc then extracts the eigenvalues of the coupled system using the MODAL SHAPE
option. The modal superposition procedure can then be used to predict the time response of the coupled system.
DYNAMIC CHANGE can be used to perform modal superposition.
To input properties of solids and the mass density of the fluid elements, use the ISOTROPIC model definition option.
The calculation of the structural mass augmentation requires triangularization of the fluid potential matrix: this matrix
is singular, unless the fluid pressure is fixed at least at one point with the FIXED DISP option.
In a fluid/solid interaction problem, the equations of motion can be expressed as
(6-194)
The pressure vector p can be calculated from:
(6-195)
The matrices in Equations (6-194) and (6-195) are defined as
(6-196)
(6-197)
(6-198)
(6-199)
where
and
are mass densities of the fluid and solid, respectively
is the acceleration vector
is the displacement vector
M
s
a Ku +
1
f
----- S
T
p =
Sa Hp =
M
s
N
i
N
i

s
V d
V
}
=
K
i j
D
i j kl

kl
V d
V
}
=
S R
f
n
i
N
i
A d
A
T
}
=
H
R
x
i
----------
R
x
i
---------- V d
V
f
}
=
f

s
a
u
333
CHAPTER 6
In the present case, the fluid is assumed to be incompressible and inviscid. Only infinitesimal displacements are
considered during the fluid vibration, so that the Eulerian and material coordinates coincide.
Substituting Equation (6-195) into Equation (6-194), we obtain:
(6-200)
This equation now allows the modes and frequencies of the solid structure immersed in the fluid to be obtained by
conventional eigenvalue methods.
Coupled Electrostatic-Structural Analysis
In a coupled electrostatic-structural analysis, both the electrostatic field and the structure are modeled. In this analysis
type, the Coulomb force (the force between charged bodies) links the electrostatic part to the structural part, and, in
turn, the deformation influences the electrostatic field. This is different from a piezoelectric analysis where a change
in the electrostatic field causes the deformation which causes a change in the electrostatic field (movement of charge),
but Coulomb forces are not taken into account. The electrostatic-structural coupling is a weak coupling, where in the
first pass the electrostatic field is computed, then the corresponding Coulomb forces are transferred to the structural
pass, and finally the structural pass is evaluated. In a next cycle, the newly deformed state is used in the electrostatic
field calculation. Since the electrostatic solution is a steady-state solution, a time dependent problem will be
considered as quasi static for the electrostatic part. In general, in an electrostatic analysis, we can distinguish
conductors and insulators. For example, in a capacitor, the plates carrying a charge are the conductors, and the
dielectric or air between the plates is the insulator. When a potential difference is applied between the plates, the
surfaces of these plates are charged. In Marc, the calculation of the charge on the different bodies is performed using
the concept of contact bodies. In the case of the capacitor, the plates and the dielectric are separate contact bodies. The
charge is then computed at the interface of two contact bodies. Note that it is necessary that the insulating contact body
is touching the conducting contact body. In other words, the contact bodies on which a Coulomb force is acting are
the ones being touched.
is the strain displacement relation
is the material constitutive relation
is the displacement interpolation function
is the pressure interpolation function
is the outward normal to the surface with fluid pressure
is the surface on which the fluid pressure acts
i j
D
i j kl
N
i
R
n
i
p
A
T
M
s
S
T
H
1
S
f
+ ( )a Ku + 0 =
or
Ma Ku + 0 =

Coulomb Force
Two methods are available to compute the Coulomb force in a coupled electrostatic-structural analysis. The default
method uses the following equation:
(6-201)
where is the charge at the node where the Coulomb force is calculated, and is the electric field at the insulator
side of the interface. Note that in principle the Coulomb force gets half the contribution from both sides of the
interface, but it is assumed that one side is the conductor where , so this contribution is neglected. This method
works well if the charged bodies are relatively close to each other (within 20 to 40 element edge lengths) such as the
capacitor shown in Figure 6-41. When the gradient in the electric field becomes too large, the computation of the
Coulomb force becomes less accurate. This is especially true when two charged bodies are located far away from each
other. In such cases, the contribution to the Coulomb force from other bodies is small relative to the self contribution.
For this situation, the following equation is available to compute the Coulomb force
(6-202)
where and are the charge on two nodes, the distance between the two charged nodes, and the permittivity
of the medium between the two nodes. This means that for each node carrying a charge the interaction with all the
other nodes carrying a charge is calculated. However, the relevant nodes carrying a charge are at the surface of contact
bodies, which reduces the number of calculations. The total charge computed on the surface of a contact body will
balance the external and reaction charge on other nodes in the contact body. This procedure is activated by the
ELECTRO,2 parameter.
Figure 6-41 Capacitor with Two Closely Spaced Charged Bodies
Equation (6-202) is valid in a 3-D analysis. For an axisymmetric analysis, the charge carrying nodes are extended to
circles. Equation (6-202) is used for a number of points on the circle. For a planar analysis, the nodal charge is a line
charge and the following equation then holds for the Coulomb force:
(6-203)
F
Coul omb
qE
2
------- =
q E
E 0
F
i
1
4
0
------------
Q
i
Q
j
r
i j
2
------------
j 1 i j , =
n
=
Q
i
Q
j
r
i j

0
F
i
1
2
0
------------
Q
i
Q
j
r
i j
------------
j 1 i j , =
n
=
335
CHAPTER 6
Analysis Procedure
The procedure to perform a coupled electrostatic-structural analysis is as follows. The different bodies (conductors
and insulators) that can be distinguished must be selected as contact bodies. The way the contact bodies are touching
is important for the computation of the Coulomb Force. The contact bodies on which a Coulomb force is acting are
the ones being touched, so in general a conductor is being touched, and air or a dielectric is the touching contact body.
Both electrostatic and structural element types are available. If an electrostatic element type is selected, then it is not
active in the structural pass. This means that this body does not deform, unless remeshing is carried out. When a
structural element is selected the corresponding electrostatic element will be automatically used in the electrostatic
pass.
The ELECTRO parameter together with the STRUCTURAL parameter is used to indicate the coupled electrostatic-
structural analysis. In addition the to the CONTACT option, the ISOTROPIC or ORTHOTROPIC model definition
option is used to define the material properties for both the electrostatic and structural pass. Specify nodal constraints
using the FIXED DISP or FIXED POTENTIAL option. Input nodal loads using the POINT LOAD option or nodal charges
using the POINT CHARGE option. Specify distributed loads with the DIST LOADS option and distributed charges with
the DIST CHARGE option. Fixed nodal displacements and potentials, or nodal forces and charges can also be specified
by the FORCDT user subroutine, nonuniform distributed loads can be specified by the FORCEM user subroutine, and
nonuniform distributed charges can be specified by the FLUX user subroutine. For the history definition, the AUTO
LOAD and AUTO STEP options can be used.
Coupled Magnetostatic-Structural Analysis
In a coupled magnetostatic-structural analysis, both the magnetostatic field and the structure are modeled. In this
analysis type, the Lorentz force links the magnetostatic part to the structural part, and in turn, the deformation
influences the magnetostatic field. Two methods to obtain the Lorentz force are available, the Virtual Work Method
(VWM), and the Maxwell Stress Tensor Method (MST). The magnetostatic-structural coupling is a weak coupling,
where in the first pass the magnetostatic field is computed, the Lorentz force is obtained and transferred to the
structural pass, and finally the structural pass is evaluated. In a next cycle, the newly deformed state is used in the
magnetostatic field calculation. Since the magnetostatic solution is a steady-state solution, a time dependent problem
will be considered and quasi static for the magnetostatic part.
Virtual Work Method
With the virtual work method the Lorentz force is derived from the variation of the magnetic energy, caused by small
displacements of a considered body. The magnetic energy is obtained from
, (6-204)
with the magnetic energy, the magnetic induction, the magnetic field intensity and the total volume of the
system. The Lorentz force is then
. (6-205)
A group of elements needs to be selected onto which the Lorentz force is calculated. This group of elements is then
displaced in the x-, y-, and for 3-D also in the z-direction. In Marc, this group of elements must be defined as a contact
W
1
2
--- B H V d
}
V
} }
=
W B H V
F
L
dr dW =

body. For correct results, this body must be surrounded by air; to improve accuracy, Marc finds a number of layers of
elements surrounding the body, and these elements together with the body get the delta displacements to obtain the
Lorentz force. The computed force is a total force on the particular body, and is applied in the structural pass as a global
load on this body.
Maxwell Stress Tensor
The general expression for the Maxwell Stress Tensor is
, (6-206)
with the magnetic induction, and the identity tensor. The force equation at any regular point then becomes,
. (6-207)
The total force is the integration of this force on an arbitrary closed surface as,
, (6-208)
with any closed surface lying entirely in vacuum bounding the region , and the normal to the surface. Note
that, although nodal forces are calculated, only the summation of these forces is a representation of the force on the
specific body or region. In Marc, the region must be defined by a contact body, where the surface of this body will be
the surface onto which the nodal forces are calculated.
The accuracy of this method depends on the choice of the surface and the accuracy of the magnetic field . Best
results are obtained when the body onto which the force is calculated is surrounded by a number of layers of elements
having material properties of vacuum/air.
Analysis Procedure
The procedure to perform a coupled magnetostatic-structural analysis is as follows. Setup the model as usual,
structural elements are the default. For structural elements, the corresponding magnetostatic elements are
automatically selected in the magnetostatic pass. If, for certain regions, like the outer surrounding air needed for the
magnetic field, magnetostatic elements are chosen; then these elements are not active in the structural pass. This means
that this region does not deform, so elements at the interface could get heavily distorted or even go inside out. This
can be overcome by remeshing those regions. Either the VWM or the MST method must be selected. The VWM
method only works for one body; for the MST, method multiple bodies can be selected.
The MAGNETO parameter together with the STRUCTURAL parameter is used to indicate the coupled magnetostatic-
structural analysis. In addition to the CONTACT option to define the different bodies, the ISOTROPIC or
ORTHOTROPIC model definition option is used to define the material properties for both the magnetostatic and
structural pass. Specify nodal constraints using the FIXED DISP or FIXED POTENTIAL option. Input nodal loads using
the POINT LOAD option or nodal currents using the POINT CURRENT option. Specify distributed loads with the DIST
LOADS option and distributed currents with the DIST CURRENT option. To create a coil, the COIL CURRENT and
EMWINDING can be used to define its geometry, the location, and the magnitude of the current in this coil. Fixed nodal
displacements and potentials or nodal forces and currents can also be specified by the FORCDT user subroutine;
T
1
0
------ BB
1
2
--- B
2
I
\ .
| |
=
B I
f
L
T =
F
n T ( )dS
S
}
=
S n
S
S B
337
CHAPTER 6
nonuniform distributed loads or nonuniform distributed currents can be specified by the FORCEM user subroutine. For
the history definition, the AUTO LOAD and AUTO STEP options can be used. Use the FORCE MAGNETOSTATIC
option to specify which method for the force calculation should be used.
Coupled Thermal-Electrical Analysis (Joule Heating)
The coupled thermal-electrical analysis procedure can be used to analyze electric heating problems. The coupling
between the electrical problem and the thermal problem in a Joule heating analysis is due to the fact that the resistance
in the electric problem is dependent on temperatures, and the internal heat generation in the thermal problem is a
function of the electrical flow.
Marc analyzes coupled thermo-electrical (Joule heating) problems. Use the JOULE parameter to initiate the coupled
thermo-electrical analysis. This capability includes the analysis of the electrical problem, the associated thermal
problem, and the coupling between these two problems.
The electrical problem is a steady-state analysis and can involve current and/or voltage boundary conditions as well
as temperature-dependent resistivity.
The thermal analysis is generally a transient analysis with temperature-dependent thermal properties and
time/temperature-dependent boundary conditions.
Use the ISOTROPIC, ORTHOTROPIC, TEMPERATURE EFFECTS, ORTHO TEMP, or TABLE model definition
options to input reference values of thermal conductivity, specific heat, mass density, and electrical resistivity, as well
as their variations with temperatures. The mass density must remain constant throughout the analysis.
Use the FIXED VOLTAGE model definition option for nodal voltage boundary conditions, and the POINT CURRENT
and/or DIST CURRENT model definition options for current boundary conditions. No initial condition is required for
the electrical problem, since a steady-state solution is obtained.
In the thermal problem, you can use the INITIAL TEMP, FILMS, POINT FLUX, DIST FLUXES, and FIXED
TEMPERATURE model definition options to prescribe the initial conditions and boundary conditions. Use the FILM
and FLUX user subroutines or tables for complex convective and flux boundary conditions. To enter the unit
conversion factor between the electrical and thermal problems, use the JOULE model definition option. Marc uses this
conversion factor to compute heat flux generated from the current flow in the structure.
Use the STEADY STATE, TRANSIENT, POINT CURRENT, DIST CURRENT, VOLTAGE CHANGE, POINT FLUX,
DIST FLUXES, and TEMP CHANGE history definition options for the incrementation and change of boundary
conditions.
A weak coupling between the electrical and thermal problems is assumed in the coupled thermo-electrical analysis,
such that the distributions of the voltages and the temperatures of the structure can be solved separately within a time
increment. A steady-state solution of the electrical problem (in terms of nodal voltages) is calculated first within each
time step.
The heat generation due to electrical flow is included in the thermal analysis as an additional heat input The
temperature distribution of the structure (obtained from the thermal analysis) is used to evaluate the temperature-
dependent resistivity, which in turn is used for the electrical analysis in the next time increment.

For output, voltage, current density, and heat generation are available as integration point values. Note that current
density must not be confused with Ohmic current. Current density is the electric current per unit area of cross section,
while Ohmic current is the current going through the total area. So the latter is a global quantity.
In the coupled thermo-electrical analysis, the matrix equation of the electrical problem can be expressed as;
(6-209)
and the governing equation of the thermal problem is:
(6-210)
In Equations (6-209) and (6-210):
The coupling between the electrical and thermal problems is through terms and in Equations (6-209) and
(6-210).
The selection of the backward difference scheme for the discretization of the time variable in Equation (6-210) yields
the following expression:
(6-211)
Equation (6-211) is used for the computation of nodal temperatures in each time increment .
is the temperature-dependent electrical conductivity matrix
is the nodal current vector
is the nodal voltage vector
and
are the temperature-dependent heat capacity and thermal conductivity
matrices, respectively
is the nodal temperature vector
is the time derivative of the temperature vector
is the heat flux vector
is the internal heat generation vector caused by the current flow.
K
E
T ( )V I =
C
T
T ( )T
K
T
T ( )T + Q Q
E
+ =
K
E
I
V
C
T
T ( ) K
T
T ( )
T
T
Q
Q
E
K
E
Q
E
1
t
----- C
T
T ( ) [ ] K
T
T ( )T
n
+ Q
n
Q
n
E
1
t
----- C
T
T ( )T
n 1
+ + =
t
339
CHAPTER 6
The internal heat generation vector is computed from:
(6-212)
(6-213)
where and are the current and electrical resistance, respectively.
The controls for the heat transfer option allow input of parameters that govern the convergence solution and accuracy
of heat transfer analysis.
Resistance
A resistor is a device that can dissipate electric energy. For a pure resistor, this usually happens in the form of heat.
This energy per unit time (or power) depends on the current in the resistor and its resistance. Resistance is always
associated with all conductors allowing flow of electric current. A typical model can contain a number of conductors
allowing flow of electric current, which gives rise to a resistance value for each conductor.
All conductors in a model are assumed to be bodies and defined as a set of elements. This is done using the THERMAL
CONTACT model definition option. Two conductors can touch each other, but cannot overlap. If two conductors touch,
it implies flow of current between the two. A subset of the conductor bodies can be considered for a resistance
computation. This subset is specified using the EMRESIS history definition option, which refers to the bodies defined
on the THERMAL CONTACT model definition option. Marc performs a resistance computation only if the EMRESIS
history definition option is specified in the input file.
A Joule heating analysis is used to compute the resistance values. All insulators in the problem are neglected for
modeling. The computation of the resistance values results from the usual Joule heating analysis with some constraints
on boundary conditions: there must be at least one electric potential boundary condition on any node. If no boundary
condition is specified on a portion of the problem boundary, it is assumed to satisfy homogeneous Neumann condition.
The resistivity of each material is specified as per the usual Joule heating analysis. If the resistivity of any material is
much higher than those of other materials, then that material is to be treated as an insulator and should not be included
in the model.
Ohms law for conductors is:
(6-214)
(6-215)
Q
E
Q
E
B
T
q
E
V d
V
}
=
q
E
i
2
R =
i R
R
V
I
---- =
W
E
I
2
R
V
2
R
------- VI = = =

where:
The conductance is given by:
(6-216)
Consider a single conducting body placed in an infinite homogenous insulating medium. The insulating medium can
be assumed to have negligible conductivity, in which case the current through it is negligible. All current flows in the
conductor. The current density is then defined as the normal current passing through a unit cross-sectional area.
This means that:
(6-217)
where is the cross-sectional surface of the conductor.
Conduction current is conserved; hence, is same for any conductor cross section.
The electric field is related to current density in the conductor by:
(6-218)
where is the conductivity of the conductor.
The power in a conductor is also given by:
(6-219)
where is the conductor domain.
The electric potential drop across the conductor is obtained from the finite element analysis. The resistance is then
calculated using Equation (6-212).
Coupled Electrical-Thermal-Mechanical Analysis
Coupled electrical-thermal-mechanical analysis (Joule-mechanical) basically combines electrical-thermal analysis
(Joule heating) with thermal-mechanical analysis. Coupled electrical-thermal-mechanical analysis is handled using a
staggered solution procedure. Using this approach, the electrical problem is solved first for the nodal voltages. Next,
the thermal problem is solved to obtain the nodal temperatures. The mechanical problem is solved last for the nodal
displacements.
is the current through the resistor
is the resistance
is the voltage across it
I
R
V
G
G
1
R
---- =
J
I J dS
}
S
}
=
S
I I
E J
J E =
W
E
W
E
1
2
--- J E v d
}
V
c
} }
1
2
---
1
--- J J v d
}
V
c
} }
= =
V
c
V
c
341
CHAPTER 6
Coupling between the electrical and thermal problems is mainly because of heat generation due to electrical flow
(Joule heating). The thermal and mechanical problems are coupled through thermal strain loads and heat
generation due to inelastic deformation and friction . Additional coupling may be introduced in case of
temperature-dependant electrical conductivity and mechanical stiffness . Nonlinearities may arise in the
thermal problem due to convection, radiation, and temperature-dependant thermal conductivity and specific heat. The
mechanical problem may involve geometric and material nonlinearities. Contact is another source of nonlinearity. If
contact occurs between deformable bodies or deformable and rigid bodies in the mechanical problem, boundary
conditions of the electrical and thermal problems are updated to reflect the new contact conditions.
The matrix equations governing the electrical, thermal and mechanical problems can be expressed as:
where
is the nodal voltage vector
is the nodal temperature vector
is the nodal displacement vector
is the temperature-dependent electrical conductivity matrix
is the nodal current vector
is the temperature-dependent heat capacity matrix
is the temperature-dependent thermal conductivity matrix
is the heat flux vector
is the heat generation due to electrical flow vector
is the heat generation due to inelastic deformation vector
is the heat generation due to friction vector
is the mass matrix
Q
E
F
T
Q
I
Q
F
K
E
K
M
Electrical Thermal Mechanical
F
T
, K
M
Q
I
, Q
F
Q
E
K
E
K
E
T ( )V I =
C
T
T ( )T
K
T
T ( )T + Q Q
E
Q
I
Q
F
+ + + =
Mu
Du
K
M
T u t , , ( )u + + F F
T
+ =
V
T
u
K
E
T ( )
I
C
T
T ( )
K
T
T ( )
Q
Q
E
Q
I
Q
F
M

The JOULE and COUPLE parameters are needed to initiate the coupled electrical-thermal-mechanical analysis. When
defining the mesh through the CONNECTIVITY model definition option, specify the elements as stress type. Marc
internally switches to the associated heat transfer element in the electrical and thermal passes.
The FIXED VOLTAGE model definition option can be used for nodal voltage boundary conditions, and POINT
CURRENT and/or DIST CURRENT model definition options for current boundary conditions. No initial condition is
required for the electrical problem, since a steady-state solution is obtained.
In the thermal problem, INITIAL TEMP, FILMS, POINT FLUX, DIST FLUXES, and FIXED TEMPERATURE model
definition options can be used to prescribe the initial conditions and boundary conditions. The FILM and FLUX user
subroutines can be used for complex convective and flux boundary conditions. To enter the unit conversion factor
between the electrical and thermal problems, use the JOULE model definition option. Marc uses this conversion factor
to compute heat flux generated from the current flow in the structure.
In the mechanical problem, the usual structural model definition options such as FIXED DISP, POINT LOAD, and DIST
LOADS are used to prescribe the boundary conditions.
The history definition options (VOLTAGE CHANGE, POINT CURRENT, DIST CURRENT, TEMP CHANGE, POINT
FLUX, DIST FLUXES, DISP CHANGE, POINT LOAD, and DIST LOADS) for the incrementation and change of
boundary conditions.
JOULE and CONVERT model definition options are used to enter the conversion factor for the electrical-thermal and
the thermal-mechanical problems.
Coupled Magnetostatic-Thermal Analysis
A coupled magnetostatic thermal analysis couples a magnetostatic analysis with a thermal analysis. With this analysis
type, heat generation in magnetostatic devices can be processed and simulated. The implementation in Marc follows
a staggered approach. First, a magnetostatic analysis is performed followed by a thermal analysis. In magnetostatic
devices, the source of heat generation is the applied currents flowing through conductors. The lamination loss
computation is available with the magnetostatic analysis and the heat generated due to these losses is also considered
here. It should be noted that lamination loss occurs due to a sinusoidally time-varying operation and the losses are
expressed as RMS (root-mean-squared) quantities. Hence, the heat generated due to lamination loss gives a correct
picture of the subsequent temperature distribution. The applied currents flowing through conductors are assumed to
be steady time-invariant and will not contain skin effect. The lamination loss is considered here, since the
magnetostatic analysis can handle a non-linear B-H curve, which is not possible in harmonic magnetodynamic
analysis.
The total generated heat is then used in the thermal analysis. Temperature dependency for material data can also be
taken into account, which is especially important for the permeability of metals. The relative permeability drops to one
when the Curie temperature is reached, which can result in a significant change in the generated heat flux.
is the damping matrix
is the temperature, deformation and time-dependent stiffness matrix
is the externally applied force vector
is the force due to thermal strain vector
D
K
M
T u t , , ( )
F
F
T
343
CHAPTER 6
The heat generated in magnetostatic analysis is handled in the same way as Joule heating analysis.
To initiate a coupled magnetostatic thermal analysis, add the MAGNETO and HEAT parameters. The Magnetostatic
pass is a static magnetostatic pass; the thermal part can be either a steady state or a transient analysis. When defining
the mesh, if you specify the element as a magnetostatic type through the CONNECTIVITY option, Marc generates an
associated heat transfer element. The region having an associated heat transfer element has coupled behavior. If you
specify the element as a heat transfer type through the CONNECTIVITY option, that region is only active in the thermal
pass. Use the ISOTROPIC or ORTHOTROPIC model definition options to input values for the thermal conductivity,
specific heat, mass density, and emissivity, as well as permeability and electric conductivity. Temperature dependency
for these material properties can be added when the table driven input option is used.
For the magnetostatic pass, boundary conditions can be applied. Use FIXED POTENTIAL to prescribe the
magnetostatic potential on nodes. To apply nodal currents use POINT CURRENT. Distributed currents can be applied
using DIST CURRENT. To create a coil, the COIL CURRENT and EMWINDING can be used to define its geometry, the
location, and the magnitude of the current in this coil. The FORCDT and FORCEM user subroutines can be used to
manually control the size and/or direction of the different vectors.
For the thermal pass, use the INITIAL TEMP, FIXED TEMPERATURE, DIST FLUXES, and FILMS model definition
options and the TEMP CHANGE and POINT FLUX history definition options to prescribe the initial conditions and
boundary conditions. The FILM and FLUX user subroutines can be used for complex convective and flux boundary
conditions.
The operational frequency for the lamination losses is given using the EMLAMIN history definition option.
The thermal pass can be STEADY STATE or time dependent using the TRANSIENT or AUTO STEP history definition
options.
If noncompatible units are used in the magnetostatic and thermal analysis, a unit conversion factor between the current
density and the generated heat can be used. It has to be ensured that the same factor applies to the lamination losses.
Use the CONVERT model definition option for this.
Coupled Magnetodynamic-Thermal Analysis
A coupled magnetodynamic thermal analysis couples a harmonic magnetodynamic analysis with a thermal analysis.
With this analysis type, induction heating processes can be simulated. The implementation in Marc follows a staggered
approach. First, a harmonic magnetodynamic analysis is performed followed by a thermal analysis. The harmonic
magnetodynamic field generates induced currents in the model. Current is also induced in conductors that carry
applied currents. The net current in these conductors is the sum of the applied and induced current. The net current in
the remaining model is the induced current. These net currents generate heat and a heat flux is computed which is then
used in the thermal analysis. Temperature dependency for material data can also be taken into account, which is
especially important for the permeability of metals. The relative permeability drops to one when the Curie temperature
is reached, which can result in a significant change in the generated heat flux. The depth of the material induction
currents generated is an important factor in the heating process. This so-called skin depth is defined as the depth at
which the magnitude of the induced current density drops to of the magnitude at the surface
, (6-220)
e
1
1
f
------------- =

where is the frequency, the electrical conductivity, and the permeability. Hence, the penetration depth depends
on the frequency. Low frequencies are used for more uniform preheating, and higher frequencies are used for surface
heating.
The heat generated in induction heating analysis is similar to a Joule heating analysis, however now in the harmonic
magnetodynamic analysis the computed current density is a complex vector instead of a scalar. The local heat
generation is (see [Ref. 17])
, (6-221)
with the electric conductivity, and the complex current density.
To initiate a coupled magnetodynamic thermal analysis, add the EL-MA and HEAT parameters. The magnetodynamic
pass is a harmonic pass; the thermal part can be either a steady state or a transient analysis. When defining the mesh,
if you specify the element as an magnetodynamic type through the CONNECTIVITY option, Marc generates an
associated heat transfer element. The region having an associated heat transfer element has coupled behavior. If you
specify the element as a heat transfer type through the CONNECTIVITY option, that region is only active in the thermal
pass. Use the ISOTROPIC or ORTHOTROPIC model definition options to input values for the thermal conductivity,
specific heat, mass density, and emissivity, as well as permeability, permittivity, and electric conductivity. Temperature
dependency for these material properties can be added when the table driven input option is used.
For the magnetodynamic pass, boundary conditions can be applied. Use FIXED POTENTIAL to prescribe the
magnetostatic or electrostatic potential on nodes, and POTENTIAL CHANGE in the history definition option. To apply
nodal currents or nodal charges used POINT CURRENT and POINT CHARGE respectively. Distributed currents or
distributed charges can be applied using DIST CURRENT or DIST CHARGE, respectively. The FORCDF and FORCEM
user subroutines can be used to manually control the size and/or direction of the different vectors.
For the thermal pass INITIAL TEMP, FIXED TEMPERATURE, TEMP CHANGE, DIST FLUXES, POINT FLUX, and
FILMS can be used to prescribe the initial conditions and boundary conditions. The FILM and FLUX user subroutines
can be used for complex convective and flux boundary conditions.
The excitation frequency has to be given for the harmonic magnetodynamic pass using the HARMONIC history
definition option.
The thermal pass can be STEADY STATE or time dependent using the TRANSIENT or AUTO STEP history definition
options.
If noncompatible units are used in the magnetodynamic and thermal analysis, a unit conversion factor between the
current density and the generated heat can be used. Use the CONVERT model definition option for this.
References
1. T. J. R. Hughes, L. P. France, and M. Becestra, A New Finite Element Formulation for Computational Fluid
Dynamics, V. Circumventing the Babuska-Brezzi Conduction, A S Petrov-Galerkin Formulation of the Stokes
Problem Accommodating Equal-Order Interpolations, Comp. Meth. in Applied Mech. and Eng., p. 85-90, Vol.
59, 1986.
2. G. Hauke and T. J. R. Hughes, A unified approach to compressible and incompressible flow, Comput.
Methods Appl. Mech. Eng., p. 389-395, 113, (1994).
f
q
E 1
--- J
2 1
--- J J
= =
J
345
CHAPTER 6
3. T. E. Tezduyar, S. Mittal, S. E. Ray, and R. Shih, Incompressible Flow computations with stabilized bilinear
and linear equal order-interpolation velocity-pressure elements, Comput. Methods. Appl. Mech. Eng., p. 221-
242, 95, (1992).
4. B. Ramaswaymy, T. C. Jue, Some Recent Trends and Developments in Finite Element Analysis for
Incompressible Thermal Flows, Int. J. Num. Meth. Eng., p. 671-707, 3, (1992).
5. A. N. Brooks and T. J. R. Hughes, Streamline upwind/Petrov-Galerkin formulations for convection dominated
flows with particular emphasis on the incompressible Navier-Stokes equations, Comp. Meth. Appl. Eng., 30,
(1982).
6. Codina, R. Finite element formulation for the numerical solution of the convection-diffusion equation 1993.
7. Cornfield, G. C., and Johnson, R. H. Theoretical Predictions of Plastic Flow in Hot Rolling Including the
Effect of Various Temperature Distributions. Journal of Iron and Steel Institute 211, pp. 567-573, 1973.
8. Hsu, M. B. Modeling of Coupled Thermo-Electrical Problems by the Finite Element Method. Third
International Symposium on Numerical Methods for Engineering, Paris, March, 1983.
9. Peeters, F. J. H. Finite Element Analysis of Elasto-Hydrodynamic Lubrication Problems. in Proceedings of
the XIth Int. Finite Element Kongress, edited by IKOSS GmbH. Baden- Baden, Germany, Nov. 15-16, 1982.
10. Yu, C C. and Heinrich, J. C. Petro-Galerkin methods for the time-dependent convective transport equation.
Int. J. Numer. Meth. Engrg., Vol. 23 (1986), 883-901.
11. Yu, C C. and Heinrich, J. C. Petro-Galerkin methods for multidimensional time-dependent convective
transport equation. Int. J. Numer. Meth. Engrg., Vol. 24 (1987), 2201-2215.
12. Zienkiewicz, O. C. The Finite Element Method in Engineering Science. Third Ed. London: McGraw-Hill, 1978.
13. Zienkiewicz, O. C., and Godbole, P. N. A Penalty Function Approach to Problems of Plastic Flow of Metals
with Large Surface Deformations. Journal of Strain Analysis 10, 180-183, 1975.
14. Zienkiewicz, O. C., and Godbole, P. N. Flow of Plastic and viscoPlastic Solids with Special Reference to
Extrusion and Forming Processes. Int. Num. Methods in Eng. 8, 1974.
15. Zienkiewicz, O. C., Loehner, R., Morgan, K., and Nakazawa, S. Finite Elements in Fluid Mechanics A
Decade of Progress, John Wiley & Sons Limited, 1984.
16. J. Goldak, A. Chakravarti, and M. Bibby, A New Finite Element Model for Welding Heat Sources,
Metallurgical Transactions B., Volume 15B, June 1984, pp. 299 - 305
17. S. Clain, J. Rappaz, M. Swierkosz, and R. Touzani, Numerical Modelling of Induction Heating for two-
Dimensional Geometries,Math. Models Methods Appl. Sci., Vol 3 no 6, 805-822, 1993
18. C. Chaboudez, S. Clain, R. Glardon, J. Rappaz, M Swierkosz, and R. Touzani, Numerical Modelling of
Induction Heating of Long Workpieces, IEEE Trans. Magn.,Vol 30, 5026-5037, 1994
19. C. Chaboudez, S. Clain, R. Glardon, D.Mari, J. Rappaz, and M Swierkosz, Numerical Modeling of Induction
Heating of Axisymmetric Geometries, IEEE Trans. Magn.,Vol 33, 739-745, 1997
20. R.A. Rindal, An analysis of the coupled chemically reacting boundary layer and charring ablator, Part IV: An
approach for characterizing charring ablator response with in-depth coking reactions., NASA CR-1065, June
1968.
Chapter 7 Material Library s
7
Material Library
Linear Elastic Material 347
Composite Material 349
Gasket 385
Nonlinear Hypoelastic Material 389
Thermo-Mechanical Shape Memory Model 404
Mechanical Shape Memory Model 413
Elastomer 418
Time-independent Inelastic Behavior 428
Time-dependent Inelastic Behavior 464
Temperature Effects and Coefficient of Thermal Expansion 492
Time-Temperature-Transformation 495
Low Tension Material 498
Soil Model 500
Damage Models 512
Nonstructural Materials 524
References 526
347
CHAPTER 7
Material Library
This chapter describes the material models available in Marc. The models range from simple linear elastic materials
to complex time- and temperature-dependent materials. This chapter provides basic information on the behavior of
various types of engineering materials and specifies the data required by the program for each material. For example,
to characterize the behavior of an isotropic linear elastic material at constant temperatures, you need only specify
Young's modulus and Poisson's ratio. However, much more data is required to simulate the behavior of material that
has either temperature or rate effects. References to more detailed information are cited in this chapter.
Data for the materials is entered into Marc either directly through the input file or by user subroutines. Each section
of this chapter discusses various options for organizing material data for input. Each section also discusses the
constitutive (stress-strain) relation and graphic representation of the models and includes recommendations and
cautions concerning the use of the models.
Linear Elastic Material
Marc is capable of handling problems with either isotropic linear elastic material behavior or anisotropic linear elastic
material behavior.
The linear elastic model is the model most commonly used to represent engineering materials. This model, which has
a linear relationship between stresses and strains, is represented by Hookes Law. Figure 7-1 shows that stress is
proportional to strain in a uniaxial tension test. The ratio of stress to strain is the familiar definition of modulus of
elasticity (Youngs modulus) of the material.
E (modulus of elasticity) = (axial stress)/(axial strain) (7-1)
Figure 7-1 Uniaxial Stress-Strain Relation of Linear Elastic Material
Experiments show that axial elongation is always accompanied by lateral contraction of the bar. The ratio for a linear
elastic material is:
= (lateral contraction)/(axial elongation) (7-2)
This is known as Poissons ratio. Similarly, the shear modulus (modulus of rigidity) is defined as:
(shear modulus) = (shear stress)/(shear strain) (7-3)
S
t
r
e
s
s
Strain
E
1
v
G

It can be shown that for an isotropic material
(7-4)
The shear modulus can be easily calculated if the modulus of elasticity and Poissons ratio are known.
Most linear elastic materials are assumed to be isotropic (their elastic properties are the same in all directions).
Anisotropic material exhibits different elastic properties in different directions. The significant directions of the
material are labeled as preferred directions, and it is easiest to express the material behavior with respect to these
directions.
The stress-strain relationship for an isotropic linear elastic method is expressed as
(7-5)
where is the Lame constant and (the shear modulus) is expressed as
(7-6)
The material behavior can be completely defined by the two material constants and .
Use the ISOTROPIC model definition option for the input of isotropic linear elastic material constants (Youngs
modulus) and (Poissons ratio).
The effects of these parameters on the design can be determined by using the DESIGN SENSITIVITY parameter. The
optimal value of the elastic properties for linear elastic analysis can be determined using the DESIGN OPTIMIZATION
parameter.
The stress-strain relationship for an anisotropic linear elastic material can be expressed as
(7-7)
The values of

(the stress-strain relation) and the preferred directions (if necessary) must be defined for an
anisotropic material. For example, the orthotropic stress-strain relationship for a plane stress element is
(7-8)
There are only four independent constants in Equation (7-8).
To input anisotropic stress-strain relations, use the ORTHOTROPIC or ANISOTROPIC model definition option and the
ANELAS or HOOKLW user subroutine. The ORTHOTROPIC option allows as many as 9 elastic constants to be defined.
The ANISOTROPIC option allows as many as 21 elastic constants to be defined. If the anisotropic material has a
preferred direction, use the ORIENTATION model definition option or the ORIENT user subroutine to input a
transformation matrix.
G E 2 1 v + ( ) ( ) =
G E v
i j

i j
kk
2G
i j
+ =
G
E 1 + ( ) 1 2 ( ) ( ) = and
G E 2 1 + ( ) ( ) =
E v
E
v
i j
C
i j kl
kl
=
C
i j kl
C
1
1
12
21
( )
-------------------------------
E
1

21
E
1
0
12
E
2
E
2
0
0 0 1
12
21
( )G
=
349
CHAPTER 7
Material Library
A Poissons ratio of 0.5, which would be appropriate for an incompressible material, can be used for the following
elements: Herrmann, plane stress, shell, truss, or beam. A Poissons ratio which is close (but not equal) to 0.5 can be
used for constant dilation elements and reduced integration elements in situations which do not include other severe
kinematic constraints. Using a Poissons ratio close to 0.5 for all other elements usually leads to behavior that is too
stiff. A Poissons ratio of 0.5 can also be used with the updated Lagrangian formulation in the multiplicative
decomposition framework using the standard displacement elements. In these elements, the treatment for
incompressibility is transparent to you.
Composite Material
Composite materials are composed of layers of different materials (or layers of the same anisotropic material) with
various layer thicknesses and different orientations. The material in each layer may be either linear or nonlinear.
Tightly bonded layers (layered materials) are often stacked in the thickness direction of beam, plate, shell structures,
or solids.
Figure 7-2 identifies the locations of integration points through the thickness of beam and shell elements with/without
the COMPOSITE option.
Note that when the COMPOSITE option is used, as shown on the left, the layer points are positioned midway through
each layer. When the COMPOSITE option is not used, the layer points are equidistantly spaced between the top and
bottom surfaces. Marc forms a stress-strain law by performing numerical integration through the thickness. If the
COMPOSITE option is used, the trapezoidal method is employed; otherwise, Simpsons rule is used.
Figure 7-2 Integration Points through the Thickness of Beam and Shell Elements
Figure 7-3 shows the location of integration points through the thickness of composite continuum elements. Marc
forms the element stiffness matrix by performing numerical integration based on the standard isoparametric concept.
Figure 7-3 Integration Points through the Thickness of Continuum Composite Elements
*
*
*
*
*
*
*
*
*
Beams or Shells with
Composite Option
Beams or Shells without
Composite Option
*
*
*
* *
*
*
*

Layered Materials
To model layered materials including plates, shells, beams, and solids with Marc, use the COMPOSITE option. In this
option, three quantities are specified on a layer-by-layer basis: material identification number, layer thickness, and ply
angle. The entire set of data (a composite group) is then associated with a list of elements. For each individual layer,
various constitutive laws can be used. The layer thickness can be constant or variable (in the case of variable total
thickness elements), and the ply angle can change from one layer to the next. The orientation of the 0
o
ply angle within
each element is defined in the ORIENTATION option. The ply thickness and the ply angle can be used as design
variables in a design sensitivity analysis. The optimal values can be determined using the design optimization
capability for linear elastic analysis.
The material identification number specified in the COMPOSITE option, is cross-referenced with the material
identification number supplied in the ISOTROPIC, ORTHOTROPIC, ANISOTROPIC, NLELAST, TEMPERATURE
EFFECTS, ORTHO TEMP, WORK HARD, and STRAIN RATE options. The ISOTROPIC, ORTHOTROPIC, and
ANISOTROPIC model definition options allow you to input material constants such as Youngs modulus, Poissons
ratio, shear modulus, etc. The TEMPERATURE EFFECTS and ORTHO TEMP options allow for input of temperature
dependency of these material constants. Material constants for a typical layer are as follows:
The ANELAS, HOOKLW, ANEXP, and ANPLAS user subroutines can be used for the anisotropic behavior of elastic
constants, coefficient of thermal expansion, and yield condition.
In models where there is ply drop-off, it is possible to specify a user ply layer id to simplify postprocessing. In this
way, the top, middle, and bottom layers may all have the same id even if the elements have a different number of layers.
There are eight given classes of strain-stress relations. The class of a particular element depends on the number of
direct (NDI) and shear (NSHEAR) components of stress. Table 7-1 lists the eight classes of elements.
thickness of the i
th
layer
Youngs moduli
Poissons ratios

Shear moduli
density
coefficients of thermal expansion
yield stress
Mat material identifier associated with temperature-dependent properties and workhardening data
t
i
E
xx
E
yy
E
zz
, ,
v
xy
v
yz
v
zx
, ,
G
xy
G
yz
G
zx
, ,
xx

yy
,
y
351
CHAPTER 7
Material Library
Table 7-1 Classes of Stress-Strain Relations
Class 1 NDI = 1,NSHEAR = 0
Beam Elements 5, 8, 13, 16, 23, 46, 47, 48, 52, 64, 77, 79 and Rebar Elements
Axisymmetric Shells 15 and 17
Beam Elements 14, 45, 76, 78
Plane Stress, Plates and Thin Shells 49 and 72
Thick Axisymmetric Shells 1 and 89
{ } 1 E
xx
} { } [ =
xx
yy
)
`
1 E
xx

yx
E
yy

xy
E
xx
1. E
yy
xx
yy
)
`

=
yx

xy
E
yy
E
xx
=
)
`
1 E
xx
0
0 1 G
xy
)
`

=
xx
yy
xy
1 E
xx

yx
E
yy
0
xy
E
xx
1 E
yy
0
0 0 1 G
xy
xx
yy
xy
)

`

=
yx

xy
E
yy
E
xx
( ) =
mm
T
)

`

1 E
mm

m
E
m
E
mm
1 E
0
0 0 1 G
m
mm
T
)

`

=

Plane Strain, Axisymmetric with No Twist, Elements 151-154.
Thick Shell, Elements 22, 75, and 140
Three-Dimensional Brick Elements, Elements 149, 150
Table 7-1 Classes of Stress-Strain Relations (continued)
xx
yy
zz
xy
)

`

1 E
xx

yx
E
yy

zx
E
zz
0
xy
E
xx
1 E
yy

zy
E
zz
0
xz
E
xx

yz
E
yy
1 E
zz
0
0 0 0 1 G
xy
xx
yy
zz
xy
)

`

=
yx

xy
E
yy
E
xx
=
zy

yz
E
zz
E
yy
=
xz

zx
E
xx
E
zz
=
xx
yy
xy
yz
zx
)

`

1 E
xx

yx
E
yy
0 0 0
xy
E
xx
1 E
yy
0 0 0
0 0 1 G
xy
0 0
0 0 0 1 G
yz
0
0 0 0 0 1 G
zx
xx
yy
xy
yz
zx
)

`

=
yx

xy
E
yy
E
xx
=
xx
yy
zz
xy
yz
zx
)

`

1 E
xx

yx
E
yy

zx
E
zz
0 0 0
xy
E
xx
1 E
yy

zy
E
zz
0 0 0
xz
E
xx

yz
E
yy
1. E
zz
0 0 0
0 0 0 1 G
xy
0 0
0 0 0 0 1 G
yz
0
0 0 0 0 0 1 G
zx
xx
yy
zz
xy
yz
zx
)

`

=
353
CHAPTER 7
Material Library
Classical Lamination Theory for Multi-Layered Shells
Basic CLT Theory
The stress tensor at a point inside a shell element is defined as
(7-9)
where is the tangent matrix connecting stress and strain tensors; is the coordinate of the point in
the thickness direction; is the strain tensor on the middle surface of the shell and is the tensor of curvature.
The membrane forces is given by
(7-10)
The bending moments is given by
(7-11)
Substituting Equations (7-9) into Equations (7-10) and (7-11) and use the definitions
(7-12)
(7-13)
(7-14)
we obtain
(7-15)
In Equations (7-12) though (7-15), is the shell thickness and .
In the case where the composite material remains linear elastic, it is possible to choose alternative computational
procedures to improve computational time. This maybe activated using either the SHELL SECT parameter or the
COMPOSITE option.
The first procedure is for linear elastic composite materials where no temperature dependent materials are
present and thermal strains occur. This substantially reduces the computational time and the memory
requirements.
The second procedure is for linear elastic composite material, but either temperature dependent materials or
thermal strains are present. This substantially reduces the computational time.
G z ( ) =
G z

f z d
}
=
m z ( ) z d
}
=
hG
1
G z d
}
=
h
2
G
4
z ( )G z d
}
=
IG
2
z
2
G z d
}
=
f
m
)
`
hG
1
h
2
G
4
h
2
G
4
IG
2
)
`

=
h I h
3
12 =

PSHELL Option
PSHELL option is based on the classical lamination theory (also known as equivalent stiffness method). This option
allows you to define the membrane, bending, transverse shear, and coupling properties of the shell
elements independently.
Shell made of homogeneous materials can be modeled with the PSHELL option by simply using the same material
properties for stiffness calculation of membrane, bending, and transverse shear deformations, and using nothing for
the coupling part if there is no offset of the shell middle surface. However, for homogeneous shell structures, it is more
efficient to use the standard shell technique which is relatively easy and inexpensive, generally more accurate, capable
to deal with nonlinear material behavior.
Shell structures with layered composite materials can be solved using a full integration technique or the PSHELL
option. The full layer integration technique is very general. It can simulate material behavior ranging from a simple
linear material to a very complex material with nonlinearity and failure mechanism. The use of the PSHELL option is
limited to linear-elastic behavior. By use of this option, composite layers are converted into one layer with equivalent
stiffness behaviors. This way, only one integration point is needed across the thickness which makes PSHELL very
useful for shell structures with layered composite materials. It is particularly attractive when the number of composite
layers is large, because analysis of these smeared shell structures uses less computer time and storage space.
The smeared material matrices , , and are defined in Equations (7-12), (7-13), and (7-14) representing
membrane, bending and coupling stiffness, respectively.
in Marc contains both membrane and transverse shear parts. Unless you want to adjust transverse shear stiffness,
there is no need to define new types of materials.
Generally , , and can be described by the ANISOTROPIC option.
Material Preferred Direction
Every element type in Marc has a default orientation (that is, a default coordinate system) within which element stress-
strain calculations take place. This system is also assumed to be the coordinate system of material symmetry. This is
especially important for non-isotropic materials (orthotropic, anisotropic, nlelast, or composite materials). With the
ORIENTATION option, you specify the orientation of the material axes of symmetry (relationship between the element
coordinate system and the global coordinate system, or the 0
o
ply angle line, if composite) in one of seven different
ways:
1. as a specific angle offset from an element edge,
2. as a specific angle offset from the line created by two intersecting planes,
3. as a particular coordinate system specified by user-supplied unit vectors,
4. as specified by the ORIENT2 user subroutine,
5. by referencing a coordinate system defined by the COORD SYSTEM option.
6. by specifying rotations around the local element system,
7. by defining NURBS curves and obtaining the orientation as the tangent of the closest point on a curve.
For the first option (EDGE I-J orientation type), the intersecting plane is defined by the surface normal vector and a
vector parallel to the vector pointing from element node I to element node J. The intersection of this plane with the
G
1
G
2
G
4
G
1
G
1
G
2
G
4
355
CHAPTER 7
Material Library
surface tangent plane defines the 0
o
orientation axis. (See Figure 7-4.) The orientation angle is measured in the tangent
plane positive about the surface normal.
Figure 7-4 Edge I-J Orientation Type
For the second option (global plane orientation type), the intersecting plane is the chosen global coordinate plane. The
intersection of this plane with the surface tangent plane defines the 0
o
orientation axis. (See Figure 7-5.)
The third option (user-defined plane orientation type) makes use of one or two user-defined vectors to define the
intersecting plane. Using a single vector, the intersecting plane is that plane which contains the user vector and the
chosen coordinate axis. Using two user vectors, the intersecting plane is that plane which contains both of them. (See
Figure 7-6.)
Orientation type 3-D ANISO also makes use of two user-defined vectors, but in this case, the first vector defines the
first (1) principal direction and the second vector defines the second (2) principal direction. (See Figure 7-7.)
In the fourth option, the ORIENT2 user subroutine is used for defining the material coordinate system.
In the fifth option, the material orientation is defined by the coordinate system defined with the COORD SYSTEM
option. If shell elements are used, the local x-axis is projected on the shell as shown in Figure 7-8.
n = Normal to Surface Tangent Plane

Element Surface
Z
Y
X
= Orientation Angle (Positive
= Ply angle (if COMPOSITE)
Right-hand Rotation About n)
Node I
Node J
Vector Parallel to Edge
Integration Point
0
0
Direction 1 of
preferred coordinate
system (fiber direction
in the ply)
Ply Angle
I-J Projected onto
Surface Tangent Plane

Figure 7-5 Global ZX Plane Orientation Type
Figure 7-6 User Defined XU Plane Orientation Type

Direction 1
E
l
e
m
e
n
t

S
u
r
f
a
c
e
Surface
Global ZX Plane
n - Normal to Surface Tangent Plane
I
n
t
e
r
s
e
c
t
i
o
n

o
f Z
Y
X
of preferred
coordinate
system
p
r
e
f
e
r
r
e
d

p
l
a
n
e
Tangent
Plane

Direction 1 of
u = user-defined vector
n = Normal to Surface Tangent Plane
Tangent Plane
Surface Tangent
Intersection of
Element Surface
Z
Y
X
global
X
Preferred
Coordinate
System
Plane
Two Planes
357
CHAPTER 7
Material Library
Figure 7-7 3-D ANISO Orientation Type
Figure 7-8 COORD SYSTEM when Shell Elements are used
The sixth option is available for hexahedral elements. The material orientation is defined as rotations around the three
axes of the element coordinate system. See Figure 7-9. The first preferred direction joins the centroids of faces 4-1-5-
8 and 3-2-6-7. A second vector joins faces 1-2-6-5 and 4-3-7-8. The third preferred direction is given by the cross
product of the first preferred direction and this vector. The second preferred direction is given by the cross product of
the third and first preferred directions. This system is then rotated around the three local axes by the three given angles.
Z
Y
X
U
1
= Direction 1 of Preferred Coordinate System
U
2
= Direction 2 of Preferred Coordinate System
U
1
= User Vector 1
U
2
= User Vector 2
U
3
= U
1
x U
2
MCID Coordinate System
x
G2
G3
z
y
material
x
material
G4
y
G1

Figure 7-9 Element Coordinate System Definition of Hexahedral Element
The seventh option method uses projection onto a curve. One or more NURBS curves are used for defining the
preferred system. A list of curves are given as input. These curves must be defined with the CURVES model definition
option and only the NURBS variant is allowed. Using the centroid of the element, the closest point on any of the given
curves is found. The first preferred direction is given by the tangent vector at this point. For 2-D elements, the second
preferred direction is given by the cross product of the global z direction and the first preferred direction. For 3-D
elements, this option is only supported for solid shell elements and solid composite elements. The third preferred
direction is given by the thickness direction and the second preferred direction by the cross product between the third
and first preferred direction. The first preferred direction is recalculated as the cross product between the second and
third preferred directions to insure that we have an orthogonal system.
Post codes 691 and 694 can be used to obtain the first and second orientation vectors of the material coordinate system.
It should be noted that these vectors are element based quantities and only account for the zero degree ply orientation.
For specific layer orientations, these vectors are further modified by the layer angle given through post code 697.
Mentat can be used to plot both the element orientations (zero degree ply angle) and the layer orientations. These
postprocessing features can be used to verify if the initial orientations are setup correctly (especially for the ORIENT2
user subroutine or the COORD SYSTEM option) and to evaluate the evolution of the orientation vectors as the structure
deforms. More details on these orientation vectors are given under the POST option in Marc Volume C: Program
Input.
Material Dependent Failure Criteria
Calculations of user specified failure criteria on a layer by layer basis are available in Marc. The available criteria are:
maximum stress (MX STRESS), maximum strain (MX STRAIN), TSAI-WU, HOFFMAN, HILL, HASHIN,
HASHIN FABRIC, HASHIN TAPE, PUCK, and the UFAIL user subroutine. During each analysis, up to three failure
criteria can be selected; failure indices and strength ratios are calculated and printed for every integration point. The
FAIL DATA model definition option is used for the input of failure criteria data. When the table driven input format is
used, the failure material parameters may reference a table to introduce temperature dependent behavior.
1
2
3
4
5
6
7
8
359
CHAPTER 7
Material Library
A simple description of these failure criteria is given below:
1. Maximum Stress Criterion
At each integration point, Marc calculates six failure indices FI and six strength ratios SR: The six failure
indices are given by:
(7-16)
(7-17)
(7-18)
(7-19)
(7-20)
(7-21)
where
are the maximum allowable stresses in the 1-direction in tension and compression.
are maximum allowable stresses in the 2-direction in tension and compression.
are maximum allowed stresses in the 3-direction in tension and compression.
1
X
t
-------
\ .
|
| |
1
X
c
------ -
\ .
|
| |
if
if
1.
1
0 <
1
0 >
2
Y
t
------
\ .
|
| |
2
Y
c
------
\ .
|
| |
if
if
2.
2
0 >
2
0 <
3
Z
t
------
\ .
| |
3
Z
c
------
\ .
| |
3
0 >
3
0 <
if
if
3.
12
S
12
---------
\ .
| |
4.
23
S
23
--------
\ .
| |
5.
31
S
31
--------
\ .
| |
6.
X
t
X
c
,
Y
t
Y
c
,
Z
t
Z
c
,

For the Maximum Stress Failure Criterion, the strength ratios SR are the reciprocals of the corresponding
failure indices FI.
For example, the sixth strength ratio for the Maximum Stress Failure Criterion is . Note that if the
maximum allowable stresses are not defined or if the actual stresses are 0, then the strength ratio is set to 100.
2. Maximum Strain Failure Criterion
At each integration point, Marc calculates six failure indices and six strength ratios. The six failure indices are
given by:
(7-22)
(7-23)
(7-24)
(7-25)
maximum allowable in-plane shear stress.
maximum allowable 23 shear stress.
maximum allowable 31 shear stress.
S
12
S
23
S
31
SR
1.0
FI
--------- =
S
31

31
( )
1
e
1t
-------
\ .
| |
1
e
1c
-------
\ .
| |
1
0 >
1
0 <
if
if
1.
2
e
2t
-------
\ .
| |
2
e
2c
-------
\ .
| |
2
0 >
2
0 <
if
if
2.
3
e
3t
-------
\ .
| |
3
e
3c
-------
\ .
| |
3
0 >
3
0 <
if
if
3.
12
g
12
--------
\ .
| |
4.
361
CHAPTER 7
Material Library
(7-26)
(7-27)
where
For the Maximum Strain Failure Criterion, the strength ratios at the integration points are the reciprocals of the
corresponding failure indices.
For example, the sixth strength ratio for the Maximum Strain Failure Criterion is . Note that if the
maximum allowable strains are not defined or if the actual strains are 0, then the strength ratio is set to 100.
3. Hill Failure Criterion
Assumptions:
a. Orthotropic materials only
b. Incompressibility during plastic deformation
c. Tensile and compressive behavior are identical
At each integration point, Marc calculates the failure index FI as follows:
(7-28)
are the maximum allowable strains in the 1 direction in tension and compression.
is the maximum allowable shear strain in the 12 plane.
23
g
23
--------
\ .
| |
5.
31
g
31
--------
\ .
| |
6.
e
1t
e
1c
,
e
2t
e
2c
,
e
3t
e
3c
,
g
12
g
23
g
31
SR
1.0
FI
--------- =
g
31

31
( )
1
2
X
2
-------

2
2
Y
2
------ -

3
2
Z
2
------
1
X
2
------ -
1
Y
2
------ -
1
Z
2
------ +
\ .
| |
1
X
2
-------
1
Z
2
------
1
Y
2
------ - +
\ .
| |
3
+ +
1
Y
2
-------
1
Z
2
------
1
X
2
------ - +
\ .
| |

12
2
S
12
2
---------

13
2
S
13
2
---------

23
2
S
23
2
--------- + + + F

For plane stress condition, it becomes
(7-29)
where
For the Hill Criterion, the strength ratio at each integration point is given by:
Note that if the failure index is 0 (either because no allowable stresses are prescribed or because actual stresses
are 0), the strength ratio is set to 100.
4. Hoffman Failure Criterion
At each integration point, Marc the failure index FI calculates:
(7-30)
is the maximum allowable stress in the 1 direction
are as before
failure index scale factor
Note:
Hoffman criterion is essentially Hill criterion modified to allow unequal maximum
allowable stresses in tension and compression.
1
2
X
2
-------

1
2
X
2
-------------

2
2
Y
2
-------

12
2
S
12
2
--------- + +
\ .
|
| |
X
Y
Z
S
12
S
23
S
31
, ,
F
SR
1.0
FI
---------- =
C
1

2

3
( )
2
C
2

3

1
( )
2
C
3

1

2
( )
2
C
4
1
C
5
2
+ + + + [
C
6
3
C
7
23
2
C
8
13
2
C
9
12
2
] F + + + +
363
CHAPTER 7
Material Library
with
(7-31)
(7-32)
where: are as before.
For the Hoffman Failure Criterion, the strength ratio SR is obtained by taking the smaller of the absolute values
of the two roots obtained by solving the quadratic equation:
Note: For small ratios of, for example, , the Hoffman criterion can become negative due to the
presence of the linear terms.
C
1
1
2
---
1
Z
t
Z
c
-----------
1
Y
t
Y
c
-------------
1
X
t
X
c
------------- +
\ .
| |
=
C
2
1
2
---
1
X
t
X
c
-------------
1
Z
t
Z
c
-----------
1
Y
t
Y
c
------------- +
\ .
| |
=
C
3
1
2
---
1
X
t
X
c
-------------
1
Y
t
Y
c
-------------
1
Z
t
Z
c
----------- +
\ .
| |
=
C
4
1
X
t
-----
1
X
c
------ =
C
5
1
Y
t
-----
1
Y
c
------ =
C
6
1
Z
t
-----
1
Z
c
------ =
C
7
1
S
23
2
-------- =
C
8
1
S
13
2
-------- =
C
9
1
S
12
2
-------- =
1
X
t
-----
1
X
c
------
\ .
| |
1
1
Y
t
-----
1
Y
c
------
\ .
| |
1
2
X
t
X
c
-------------

2
2
Y
t
Y
c
-------------

12
2
S
12
2
---------

1
2
X
t
X
c
------------- + + + +
)
`

F
X
t
X
c
Y
t
Y
c
Z
t
Z
c
S
12
S
23
S
31
F , , , , , , , , ,
1
X
t
------
A SR ( )
2
B SR ( ) C + + 0 =

where for the general 3-D case:
,
, and .
Note that if the allowable stresses are not defined or if the actual stresses are 0, then the strength ratio is set to
100.
5. Tsai-Wu Failure Criterion
Tsai-Wu is a tensor polynomial failure criterion. At each integration point, Marc calculates the failure index FI
as follows:
(7-33)
where are as before.
(7-34)
Interactive strength constant for the 12 plane
Note:
In order for the Tsai-Wu failure surface to be closed,
See Wu, R.Y. and Stachurski, 2, Evaluation of the Normal Stress Interaction Parameter in the
Tensor Polynomial Strength Theory for Anisotropic Materials, Journal of Composite Materials,
Vol. 18, Sept. 1984, pp. 456-463.
A
C
1

2

3
( )
2
C
2

3

1
( )
2
C
3

1

2
( )
2
C
7
23
2
C
8
13
2
C
9
12
2
+ + + + +
F
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- =
B
C
4
1
C
5
2
C
6
3
+ +
F
------------------------------------------------------ - = C 1.0 =
1
X
t
-----
1
X
c
------
\ .
| |
1
1
Y
t
-----
1
Y
c
------
\ .
| |
2
1
Z
t
-----
1
Z
c
------
\ .
| |
1
2
X
t
X
c
-------------

2
2
Y
t
Y
c
-------------

3
2
Z
t
Z
c
----------- + + + + +
12
2
S
12
2
--------

23
2
S
23
2
--------

13
2
S
13
2
-------- 2F
12
2
2F
23
3
2F
13
3
] + + + + + + F
X
t
X
c
Y
t
Y
c
Z
t
Z
c
S
12
S
23
S
31
F , , , , , , , , ,
F
12
F
23
F
13
1
X
t
-----
1
X
c
------
\ .
| |
1
1
Y
1
------ -
1
Y
c
------
\ .
| |
1
2
X
t
X
c
-------------

2
2
Y
t
Y
c
-------------

12
2
S
12
2
--------- 2F
12
2
+ + + + +
)
`

F
F
12
2
1
X
t
X
c
-------------
1
Y
t
Y
c
------------- < F
23
2
1
Y
t
Y
c
-------------
1
Z
t
Z
c
----------- < F
31
2
1
X
t
X
c
-------------
1
Z
t
Z
c
----------- <
365
CHAPTER 7
Material Library
For the Tsai-Wu failure criterion, the strength ratio SR is obtained by taking the smaller of the absolute values
of the two roots obtained by solving the quadratic equation:
where for the general 3-D case:
, and .
100.
6. User-defined Failure Criteria
Using the UFAIL user subroutine, you can evaluate your own failure criterion as a function of stresses and
strains at each integration point. Two quantities can be returned by the routine:
A user-defined failure index
A user-defined strength ratio.
7. Hashin Failure Criterion
The Hashin failure criterion distinguishes between fiber failure and matrix failure. At each integration point,
Marc calculates the failure index FI for each mode as follows:
Tension fiber mode,
or (7-35)
Compressive fiber mode,
(7-36)
Tensile matrix mode,
(7-37)
A SR ( )
2
B SR ( ) C + + 0 =
A
1
2
X
t
X
c
-------------

2
2
Y
t
Y
c
-------------

3
2
Z
t
Z
c
-----------

12
2
S
12
2
--------

23
2
S
23
2
--------

13
2
S
13
2
-------- 2F
12
2
2F
23
3
2F
13
3
+ + + + + + + +
F
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- =
B
1
X
t
-----
1
X
c
------
\ .
| |
1
1
Y
t
-----
1
Y
c
------
\ .
| |
2
1
Z
t
-----
1
Z
c
------
\ .
| |
3
+ +
F
--------------------------------------------------------------------------------------------------------------- =
C 1.0 =
11
X
t
--------
\ .
| |
2
1
S
2
------
12
2
13
2
+ ( ) + 1 =
11
X
T
=
1
0 <
11
X
c
------------ 1 =
2

3
+ 0 >
1
Y
t
-----
2

3
+ ( )
2
1
S
23
2
--------
23
2
3
( )
1
S
12
--------
12
2
13
2
+ ( ) + +

Compressive matrix mode,
(7-38)
For the Hashin failure criterion, the strength ratio SR for each mode is calculated as follows:
Tension fiber mode: where FI is the failure index for the tension fiber mode
Compressive fiber mode: where FI is the failure index for the compressive fiber mode
Tensile matrix mode: where FI is the failure index for the tensile matrix mode
Compressive matrix mode: where
, and .
100.
A sixth failure index is available for postprocessing. It is the maximum of the first five failure criteria.
Similarly, the minimum strength ratio is available as the seventh strength ratio.
8. Hashin Fabric Failure Criterion
The Hashin Fabric failure criterion is a variant of the Hashin criterion adapted for fabric type materials. The 1
direction is the first fiber direction, the 2 direction is the second fiber direction, and the 3 direction is through
the thickness direction. At each integration point, Marc calculates the failure indices as follows:
Tensile fiber 1 mode,
(7-39)
2

3
0 < +
1
Y
c
------
Y
c
2S
23
-----------
\ .
| |
2
1
\ .
| |

2

3
+ ( )
1
4S
23
2
------------
2

3
+ ( )
2
1
S
23
2
--------
23
2
3
( )
1
S
12
2
--------
12
2
13
2
+ ( )
+ +
+
SR
1.0
FI
---------- =
SR
1.0
FI
--------- =
SR
1.0
FI
---------- =
A SR ( )
2
B SR ( ) C + + 0 =
A
1
4S
23
2
------------
2

3
+ ( )
2
1
S
23
2
--------
23
2
3
( )
1
S
12
2
--------
12
2
13
2
+ ( ) + + =
B
1
Y
c
------
Y
c
2S
23
-----------
\ .
| |
2
1
\ .
| |

2

3
+ ( ) = C 1.0 =
1
0 >
1
X
t
------
\ .
| |
2

12
S
12
---------
\ .
| |
2

13
S
13
--------
\ .
| |
2
+ +
367
CHAPTER 7
Material Library
Compressive fiber 1 mode,
(7-40)
Tensile fiber 2 mode,
(7-41)
Compressive fiber 2 mode,
(7-42)
(7-43)
(7-44)
where: are as before.
For the Hashin Fabric Failure Criterion, the strength ratio SR for each of the six modes is calculated as:
where FI is the failure index for the corresponding mode. Note that if the maximum allowable stresses are not
defined or if the actual stresses are 0, then the associated strength ratios are set to 100.
9. Hashin Tape Failure Criterion
The Hashin Tape failure criterion is a variant of the Hashin criterion adapted for tape type of materials. The 1
direction is in the tape fiber direction, the 2 direction is perpendicular to the fiber direction in the plane of the
tape, and the 3 direction is through the thickness direction. At each integration point, Marc calculates the failure
indices as follows:
Tensile fiber mode,
(7-45)
1
0 <
1
X
c
------
\ .
| |
2

12
S
12
---------
\ .
| |
2

13
S
13
--------
\ .
| |
2
+ +
2
0 >
2
Y
t
------
\ .
| |
2

12
S
12
---------
\ .
| |
2

13
S
13
--------
\ .
| |
2
+ +
2
0 <
2
Y
c
------
\ .
| |
2

12
S
12
---------
\ .
| |
2

13
S
23
--------
\ .
| |
2
+ +
3
0 >
3
Z
t
------
\ .
| |
2

12
S
12
---------
\ .
| |
2

13
S
13
--------
\ .
| |
2

23
S
23
--------
\ .
| |
2
+ + +
3
0 <
3
Z
c
------
\ .
| |
2

12
S
12
---------
\ .
| |
2

13
S
13
--------
\ .
| |
2

23
S
23
--------
\ .
| |
2
+ + +
X
t
X
c
Y
t
Y
c
Z
t
Z
c
S
12
S
23
S
13
, , , , , , , ,
SR
1.0
FI
---------- =
1
0 >
1
X
t
------
\ .
| |
2

12
S
12
---------
\ .
| |
2

13
S
13
--------
\ .
| |
2
+ +

(7-46)
(7-47)
(7-48)
where: are as before. is equal to zero or one and is used to determine if the
last term should be used. is the maximum fiber stress for matrix compression.
For the Hashin Tape Failure Criterion, the strength ratio SR for the first three modes is calculated as
where FI is the failure index for the corresponding mode.
For the compressive matrix mode, the strength ratio is calculated as follows:
where
,
, and .
Note that if the maximum allowable stresses are not defined or if the actual stresses are 0, then the associated
strength ratios are set to 100.
10. Puck Failure Criterion
The Puck failure criterion distinguishes just as Hashin between fiber failure and matrix failure. The main
difference is in the matrix failure where the concept of a fracture failure angle is used. This angle is the angle
at which the fracture due to matrix failure will occur. See Figure 7-10 for the definition on how it is defined
with respect to the preferred coordinates system.
1
0 <
1
X
c
------
\ .
| |
2

12
S
12
---------
\ .
| |
2

13
S
13
--------
\ .
| |
2
+ +
2

3
+ 0 >
2

3
+
Y
t
-------------------
\ .
| |
2

22
33
S
23
2
-----------------

12
S
12
---------
\ .
| |
2

13
S
13
--------
\ .
| |
2

23
S
23
--------
\ .
| |
2
+ + +
2

3
+ 0 <
Y
c
2S
23
-----------
\ .
| |
2
1
\ .
| |

2

3
+
Y
c
-------------------
\ .
| |

2

3
+
2S
23
-------------------
\ .
| |
2

2
3
S
23
2
-------------

12
S
12
---------
\ .
| |
2

13
S
13
--------
\ .
| |
2

23
S
23
--------
\ .
| |
2
A
1
1
S
1
------
\ .
| |
2
+ + + + +
X
t
X
c
Y
t
Y
c
S
12
S
23
S
13
, , , , , , A
1
S
1
SR
1.0
FI
---------- =
A SR ( )
2
B SR ( ) C + + 0 =
A

2

3
+
2S
23
-------------------
\ .
| |
2

2
3
S
23
2
-------------

12
S
12
--------
\ .
| |
2

13
S
13
--------
\ .
| |
2

23
S
23
--------
\ .
| |
2
A
1
1
S
1
------
\ .
| |
2
+ + + = =
B
1
Y
c
------
Y
c
2S
23
-----------
\ .
| |
2
1
\ .
| |

2

3
+ ( ) = C 1.0 =
369
CHAPTER 7
Material Library
Figure 7-10 Definition of the Fracture Failure Angle
The Puck failure criterion uses the following input parameters as defined above: . In
addition, four parameters are used for describing the slopes of the failure envelope: ,
, , and . The notation etc. is used in Marc Volume C: Program
Input and in the Marc output while the other notation is from Puck [Ref. 25]. The following parameters are
calculated from the input parameters and used in the following
(7-49)
(7-50)
The first two failure indices are given by
Tensile fiber mode,
(7-51)
(7-52)
The fracture failure angle can be calculated analytically for the plane stress case provided that the
following relation between two of the failure envelope slopes is enforced:
(7-53)
It is recommended that either or is specified in the input. The other is by default calculated using
Equations (7-49) and (7-53). This default setting is also used for other cases than plane stress although then this
relationship is not necessary.
2
1
3
n
t
fp
fp
X
t
X
c
Y
t
Y
c
S
12
, , , ,
p
12c
p
||
()
=
p
12t
p
||
(+)
= p
23c
p
()
= p
23t
p
(+)
= p
12c
R
A
Y
c
2 1 p
23c
+ ( )
-----------------------------
S
12
2p
12c
-------------- 1 2p
12c
Y
c
S
12
-------- + 1
\ .
|
| |
= =
21c
S
12
1 2p
23c
+ =
1
0 >
1
X
t
------
1
0 <
1
X
c
---------
fp
p
23c
p
12c
R
A
S
12
-------- =
p
12c
p
23c

For the plane stress case, we have three failure indices, denoted modes A, B, and C:
Mode A, ,
(7-54)
Mode B, and ,
(7-55)
Mode C, and .
(7-56)
For this plane stress case, the fracture failure angle is calculated from the following formula
(7-57)
It is available as the sixth failure index but only for postprocessing.
For other than plane stress, the failure angle is not determined analytically. Instead, a numerical procedure is
used as outlined below.
The stresses in a failure plane defined by 1, n and t directions in Figure 7-10 are
(7-58)
(7-59)
(7-60)
The failure index as a function of is now formulated as follows, :
(7-61)
and for
2
0 >
fp
0 =
12
S
12
--------
\ .
| |
2
1 p
12t
Y
t
S
12
--------
\ .
| |
2

2
Y
t
------
\ .
| |
2
+ p
12t
2
S
12
-------- +
2
0 < 0

2
12
--------
R
A
21c
-----------
fp
0 =
1
S
12
--------
12
2
p
12c
2
( )
2
+ p
12c
2
+ ( )
2
0 < 0

12
2
---------
21c
R
A
-----------
12
2 1 p
23c
S
12
+ ( )
--------------------------------------
\ .
| |
2

2
Y
c
------
\ .
| |
2
+
\ .
| |
Y
c
2
---------
fp
cos
2
1
2 1 p
23c
+ ( )
-----------------------------

12
2
--------
\ .
| |
2
R
A
S
12
--------
\ .
|
| |
2
1 +
\ .
|
| |
=
n

2
cos
2
3
sin
2
2
23
cos sin + + =
nt

3

2
( ) cos sin
23
cos
2
sin
2
( ) + =
n1

31
sin
12
cos + =

n
0
f ( )
1
Y
t
----- p
1
\ .
| |
2
n
2

nt
R
A
--------
\ .
| |
2

n1
S
12
---------
\ .
| |
2
+ + p
1
n
+ =
n
0 <
371
CHAPTER 7
Material Library
(7-62)
where
(7-63)
(7-64)
The critical failure angle and the corresponding failure index is obtained by evaluating equations (7-58)
through (7-64) with a sufficient number of angles between -90 and 90 degrees. For positive normal stress, it is
saved as the 3rd failure index and as the 4th failure index. The 5th failure index is not used in this case. The
critical failure angle is stored as the 6th failure index for post processing just as for plane stress.
For the Puck Failure Criterion, the strength ratios for each mode are the reciprocals of the corresponding failure
indices, . Note that if the maximum allowable stresses are not defined or if the actual stresses are
0, then the associated strength ratios are set to 100.
A seventh failure criterion is available for postprocessing. It is the maximum of the first five failure criteria.
Similarly, the minimum strength ratio is available as the seventh strength ratio.
11. SIFT Failure Criterion
The Strain Invariant Failure Theory (SIFT) criterion is based on evaluating the failure indices using the strain
invariants. It accounts for micro-mechanical effects of the composite without having a micro-mechanical
model. The SIFT failure criterion also takes into effect, the thermo-mechanical influences on the composite
material. The Strain Invariant Failure Theory criterion is based on the work done by Gosse, J. H. et al, Boeing
Company [Ref. 30].
Strain Invariants: The fundamental hypothesis of the SIFT criterion is that failure is based on the first
invariant and the second invariant (or equivalent strain) of the homogenous medium. In other words, if any of
the two quantities exceeds the critical limit, the failure is said to initiate.
The first invariant, which accounts for the dilatation nature of deformation, is given as
(7-65)
where , , and are the diagonal components of strain tensor. And the equivalent strain, which
accounts for the distortional (or deviatoric) nature of deformation, is given as
(7-66)
The critical values of the quantities are to be evaluated using experiments. Please refer to [Ref. 30] for further
details.
f ( )

nt
R
A
--------
\ .
| |
2

n1
S
12
---------
\ .
| |
2
p
2
n
( )
2
+ + p
2
n
+ =
p
1
p
23t
R
A
---------

nt
2
nt
2
n1
2
+
-------------------------
p
12t
S
12
---------

n1
2
nt
2
n1
2
+
------------------------- + =
p
2
p
23c
R
A
----------

nt
2
nt
2
n1
2
+
-------------------------
p
12c
S
12
----------

n1
2
nt
2
n1
2
+
------------------------- + =
fp
SR
1.0
FI
--------- =
J
1

11

22

33
+ + =
11

22

33
eq
2 3
i j
i j
( ) [ ]
1 2
=

Micro-mechanical enhancements: In a typical composite analysis, the deformation of the structure is not
typically linked to the stress and strain states of the heterogeneous phases of the composite and most predictions
of composite structure failure have been based upon stresses and strains within the homogenized system. The
SIFT theory considers the heterogeneous effects without having a homogeneous model. For a given state of
strain in the homogeneous medium, the SIFT criterion has a methodology called influence function formulation
to evaluate the heterogeneous states of strain for the fiber and matrix medium.
Representative volume element: The parameters required to account for the micro-mechanical effects of the
model is deduced using the representative volume element (RVE). The configuration of the RVE can be based
on the volume fraction and the fiber arrangement in the matrix of the composite. Two configurations of fiber
arrangement square and hexagonal are shown below.
Figure 7-11 Square and Hexagonal Array Representative Volume Elements for Finite Element Analysis
[Ref. 29]
Figure 7-12 Representative Volume Elements for Determining Influence Parameters
(a) Square Arrangement (b) Diagonal Arrangement
373
CHAPTER 7
Material Library
Influence function formulation: Considering the representative volume element being subjected to an
arbitrary state of strain, including both thermal and mechanical components with
the effective coefficient of thermal expansion of the homogenized medium. The state of strain at a
prescribed point within the representative volume element can be determined from the following relationship:
(7-67)
where is the influence function matrix at the k
th
point in the RVE and is the thermal vector of strain,
per unit thermal change, induced at the point of interest within the boundaries of the unconstrained
representative volume element.
Essentially, the formulation helps to amplify the strain to account for the heterogeneous effects, which is then
used to calculate the strain invariant quantities. The enhanced strain invariants are then compared against the
respective critical values for predicting failure.
Methodology to obtain the influence matrices: The user is required to input the data required for this
formulation in the form of influence matrices. The influence matrices can be evaluated by performing a
separate finite element analysis of a model called representative volume element (RVE) a heterogeneous
model containing fiber and matrix elements with respective materials properties. This can be achieved using
the Mentat product as shown below.
At a given point of interest, , the components of the influence function matrix can be determined uniquely by
prescribing a canonical (simple) state of deformation of the representative volume element and carrying out
three-dimensional finite-element analyses within the representative volume element to determine the
components of the strain tensor at the specified point, . For example, let , and
and then,
(7-68)
j

j
T ( ) i j , ( ) 1 6 =
i
k
i
k
T M
i j
k
j

j
T ( ) A
i
k
T + = i j , 1 6 =
M
i j
k
A
i
k
k
k
j
0 = j ( 2 6 ) =
1
1 =
T 0 =
1
k
M
11
k
1
or M
11
k
1
k
= =
2
k
M
21
k
1
or M
21
k
2
k
= =
3
k
M
31
k
1
or M
31
k
3
k
= =
4
k
M
41
k
1
or M
41
k
4
k
= =
5
k
M
51
k
1
or M
51
k
5
k
= =
6
k
M
61
k
1
or M
61
k
6
k
= =

It is to be noted that for any point, , a single finite-element analysis with boundary conditions appropriate to
the condition, yields 6 of the 36 coefficients in the influence function matrix. Five more finite element
analyses are required to determine the remaining terms of . However, these same 6 finite element analyses
can produce the terms of at another point, , within the medium as well. Thus, only 6 finite element
models need to be constructed for a given representative volume element geometry to determine any number
of influence function matrices. Though, the user is open to choose specific points, SIFT specifies the points at
which the parameters needs to be determined. These points are called interrogation points (or locations). For
the matrix phase, the points chosen differ for the two representative volume elements. These points represent
both the point the greatest distance from the fiber phase and the two points where the fibers are closest to one
another. For the hexagonal array, the point where the least distance between fibers occurs is chosen along with
two other intermediate points.
Figure 7-13 Interrogation Locations within the Matrix Phase for Hexagonal and Square Array for Fiber
Diameter, D. [Ref. 29]
The displacement boundary conditions for the prescribed extensional strain for the square array RVE in the x
direction are:
The canonical shearing displacements at the RVE faces, for the rectangular array are shown as follows for the
yz plane:
k
1
1 =
M
i j
k
M
i j
k
k
u 0 at x 0 and u 10
6
at x 1 = = = =
v 0 at y 0 1 , = =
w 0 at z 0 1 , = =
y 0 y , 1 and z 0 z , 1 : u 0 v ,
z
2
--- 10
6
w ,
y
2
--- 10
6
= = = = = = =
x 0 x , 1 : u 0 = = =
375
CHAPTER 7
Material Library
Thermal enhancement vectors: In the case of a uniform temperature change for the representative volume
element, boundary conditions were imposed to allow for free translation normal to the surfaces: ,
, and for the square array and for the hexagonal array, , , and for the
square array and for the hexagonal array, , , and , while holding deformations normal
to the planes , , and to zero. This set of nodal boundary conditions is intended to yield
a traction free state of deformation of the RVE. In this case, the unit cell free thermal deformation is equal to
that defined by the coefficients of thermal expansion of the homogenized unidirectional lamina,
Let
The thermal strain enhancement vector can be evaluated as:
Failure Indices and Strength Ratio
The failure index and the strength ratios are calculated as follows:
a. The strain is amplified using equation (7-67) for each of the MATRIX amplification matrices
b. The strain is amplified using equation (7-67) for each of the FIBER amplification matrices
c. The first invariants are calculated for each of amplified strains for the MATRIX material (The fiber
material is not considered for calculation of this invariant ([Ref. 29] and [Ref. 30]))
d. The equivalent strains are calculated for each of amplified strains for the MATRIX material
e. The equivalent strains are calculated for each of amplified strains for the FIBER material
f. The maximum of , , is compared against the respective user-specified critical values.
g. The failure indices are calculated as the ratio of the above max values with the respective critical values. The
order of failure indices are corresponding to , , . In other words, the first failure index is based on
, the second based on point and the third based on point .
h. The strength ratio is the reciprocal of the failure index.
Mentat Procedure for SIFT Influence Matrices
In order to make it easy for the users to compute the strain enhancement coefficients, the Marc installation is equipped
with Python scripts to make the micro-mechanical model and to calculate the parameters. The procedure to obtain the
strain enhancement matrices and thermal enhancement vectors - given the properties of the constituent materials of
the composite and the volume fraction of fiber - is detailed below.
x 1 =
y 1 = z 1 = x 1 = y 1 = z 1 =
x 1 = y 1 = z 3 =
x 0 = y 0 = z 0 =
j
T
i
k
i
k
T M
i j
k
j
x
j
T ( ) A
i
k
T = i j , 1 6 = ( )
j

j
T = j 1 3 = ( )
A
i
k

i
k
i
k
T
T
--------------------------- = i 1 6 = ( )
J1 ( )
e
m
( )
e
f
( )
J1
e
m
e
f
J1
e
m
e
f
J1
e
m
e
f

The Python scripting utility is used to run the scripts by choosing the following options from the Mentat GUI.
UTILS -> PYTHON -> RUN
The scripts are located in \mentat\utilities\python directory under the Mentat installation directory. The scripts listed
below are to be copied to the users working directory.
sift_01_create_parameters.py
sift_02_compute_strain_enhancement.py
sift_03_apply_strain_enhancement.py
The steps are explained as follows:
1. Save any existing model.
2. Open a new model.
3. Run the Python script named sift_01_create_parameters.py by choosing UTILS->PYTHON->RUN. This script
creates the input parameter keywords required for the input of material parameters and volume fraction.
4. Modify the volume fraction of fiber and the material parameters of the constituent materials of the composite
as desired by choosing UTILS->PYTHON->PARAMETERS. The values available are representative in nature.
5. Run the Python script named sift_02_compute_strain_enhancemnt.py. This script builds the representative
volume element (RVE) model using the volume fraction and the material properties and runs seven simulations
(jobs) required for computation of the strain enhancement parameters - the influence matrices and thermal
strain enhancement vector.
6. After ensuring that all the seven simulations are completed, open any existing model (e7x38.mud for example)
or a new model and run the script named sift_03_apply_strain_enhancemnt.py. This script postprocesses the
simulation results of the RVE and computes the strain enhancement parameters. In addition, it creates a new
material in the users Mentat model and applies the computed properties in the DAMAGE EFFECTS -> STRAIN
INVARIANT -> INFLUENCE FUNCTION PROPERTIES section.
Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D
Composite Brick Elements
Another addition made for composite analysis is the calculation of interlaminar shears. These interlaminar shears are
printed in the local coordinate system above and below each layer selected for printing by PRINT CHOICE or PRINT
Notes: 1. The scripts are to be present in the current working directory.
2. It takes several minutes for the RVE model simulations to complete. Before running the third script,
the user needs to ensure that the result files named sift_job1.t16, sift_job2.t16 and so on are available
in the current working directory.
3. The RVE simulations are required to be run only once for a given set of constituent materials and
volume fraction of fiber. The results files and the third script, which are to be placed in the current
working directory, can be used for as many user models as required to create the influence matrices
and thermal enhancement vectors.
377
CHAPTER 7
Material Library
ELEMENT. These values are also available for postprocessing. The TSHEAR parameter must be used for activating the
parabolic shear distribution calculations.
In Marc, the distribution of transverse shear strains through the thickness for thick shell and beam elements was
assumed to be constant. From basic strength of materials and the equilibrium of a beam cross section, it is known that
the actual distribution is more parabolic in nature. As an additional option, the formulations for elements 1, 22, 45, 75,
89, 140, 149, 150, and 185 (TSHEAR will be switched off for elements 185, 149 and 150 when the elements are
stacked) have been modified to include a parabolic distribution of transverse shear strain. The formulation is exact for
beam element 45, but is approximate for the other thick shell elements. Nevertheless, the approximation is expected
to give improved results from the previous constant shear distribution. Furthermore, interlaminar shear stresses for
composite beams and shells can now be easily calculated.
The generalized stiffness matrix for the complete section excluding transverse shear terms is given by:
(7-69)
where X, Y, and Z are 3 x 3 matrices
A unique X, Y direction in the plane of the section is defined by a rotation around the element normal which maximizes
the value of in the above equation.
We then assume that the stresses in the X and Y direction are uncoupled, this gives:
and (7-70)
If we assume only bending and transverse shear in the section, all section forces are zero and inverting the above
equation gives:
F = section forces
M = section moments
= strain at mid plane of section
= curvature
1,2 are in-plane directions
F
11
F
22
F
12
M
11
M
22
M
12
X Y
Y Z
11
22
12
11
22
12
=
X
11
F
xx
M
xx
)

`

X
xx
Y
xx
Y
xx
Z
xx
xx
xx
)

`

=
F
yy
M
yy
)

`

X
yy
Y
yy
Y
yy
Z
yy
yy
yy
)

`

=

and (7-71)
For a point in the section, we can now define the stresses as:
(7-72)
Since we assumed that all stresses in the X and Y direction are uncoupled, the equilibrium conditions through the
thickness are given by:
and (7-73)
where and are the transverse shear stresses.
From beam theory, we have:
and (7-74)
where is the bending moments and is the shear forces.
Combining the Equations (7-72), (7-73), and (7-74) gives:
and (7-75)
We can integrate this through the thickness giving:
and (7-76)
with the boundary conditions that the shear stresses at the top and the bottom of the shell are zero.
We can now define the transverse shear stiffness by matching the shear strain energy over the section obtained with
the transverse shear stresses given in (7-76). This yields the flexibility matrix S:
(7-77)
where is the transverse shear flexibilities of the material through the thickness.
Inversion of the flexibility matrix gives the transverse shear stiffness of the section.
xx
xx
)

`

H
1x
M
2x
M
xx
{ } =

yy
yy
)

`

H
1y
M
2y
M
yy
{ } =
xx
z ( ) E z ( )
xx
z ( ) E z ( )
xx

xx
z + { } E z ( ) H
1x
H
2x
z + { }M
xx
= = = and
yy
z ( ) E z ( )
yy
z ( ) E z ( )
yy

yy
z + { } E z ( ) H
1y
H
2y
z + { }M
yy
= = =
zx
z
-----------

xx
z ( )
x
-------------------- + 0 =

zy
z
-----------

yy
z ( )
y
-------------------- + 0 =
zy

zx
V
x
M
xx
x
-------------- + 0 = V
y
M
yy
y
-------------- + 0 =
M V
zx
z
----------- E z ( ) H
1x
H
2x
z + { } V
x
=

zy
z
----------- E z ( ) H
1y
H
2y
z + { } V
y
=
zx
z ( ) E z ( ) H
1x
H
2x
z + { }V
x
z d
t 2
t 2
}
=
zy
z ( ) E z ( ) H
1y
H
2y
z + { }V
y
z d
t 2
y
}
=
1
2
---
V
x
V
y
)
`

S [ ]
V
x
V
y
)
`

1
2
---

zx
z ( )
zy
z ( )
)
`
G
xx
z ( ) G
xy
z ( )
G
yx
z ( ) G
yy
z ( )
zx
z ( )
zy
z ( )
)
`

z d
t 2
t 2
}
=
G
S
379
CHAPTER 7
Material Library
Interlaminar Stresses for Continuum Composite Elements
In Marc, the interlaminar shear and normal stresses are calculated by averaging the stresses in the stacked two layers.
The stresses are transformed into a component tangent to the interface and a component normal to the interface. The
two components, considered as shear stress and normal stress, respectively, are printed out in the output file.
By using POST code 501 or 511 (see Marc Volume C: Program Input) representing interlaminar normal and shear
stresses respectively, the interlaminar normal or shear stress can be written into a post file in the form of a stress tensor
defined in the global coordinate directions. Mentat can be used to plot the principal directions of the stress tensor which
show the magnitude and the direction of the stress, and the changes based on deformation.
Progressive Composite Failure
Marc supports progressive failure analysis for composites and other elastic materials. The material is assumed to be
linear elastic up to the point of failure. Failure is indicated by the failure criteria described in the previous section.
When failure occurs, the element stiffness is degraded. Marc offers three different methods for the material
degradation as described below. This is flagged through the FAIL DATA model definition option. While Marc allows
up to three failure criteria to be used for failure index calculation, only the first criterion is used for progressive failure.
The material will not heal; the damaged elements keep the degraded properties after unloading.
Model 1 Selective Gradual Degradation
This model uses a selective degradation of the moduli depending on failure mode. The moduli are decreased gradually
when failure occurs. Within an increment, it attempts to keep the highest failure index less than or equal to one.
Whenever a failure index F larger than one occurs, stiffness reduction factors r
i
are calculated based upon the value of
the failure indices. The incremental contribution to the total reduction factor is calculated as
(7-78)
This is done differently for different failure criteria as described below. Six such reduction factors are stored and
updated. They are then used for scaling the respective material modulus according to
The Poissons ratios are scaled in the same way as the corresponding shear modulus.
r
i
1 e
1 F
( ) =
E
11
new
r
1
E
11
or i g
=
E
22
new
r
2
E
22
or i g
=
E
33
new
r
3
E
33
or i g
=
G
12
new
r
4
G
12
or i g
=
G
23
new
r
5
G
23
or i g
=
G
31
new
r
6
G
31
or i g
=

For the maximum stress and maximum strain criteria the reduction factors are calculated separately from each separate
failure index: r
1
is calculated from the first failure index as given by Equation 7-16 above, r
2
is calculated from the
second failure index from Equation 7-17 etc. Thus, there is no coupling of the different failure modes for these criteria.
For the failure criteria which only have one failure index: Tsai-Wu, Hoffman and Hill, all six reduction factors are
decreased in the same way, using the smallest of the r
i
:s.
For the criteria which distinguish between fiber and matrix failure (Hashin, Hashin-tape and Puck) there is a more
complex coupling between the failure modes. There is a default behavior which can be influenced by a number of input
parameters. The default is as follows.
r
1
depends on fiber failure (first and second failure index)
r
2
depends on matrix failure (third, fourth and fifth failure index)
r
3
behaves the same as r
1
( )
r
4
behaves the same as r
2
( )
r
5
and r
6
behave the same as r4 ( )
In the FAIL DATA option, there are five parameters available for controlling the way the moduli are reduced. With the
exception of a
1
, they are only used for the Hashin variants and Puck.
a
1
Residual stiffness factor: The stiffness is never reduced to less than this factor. The default is 0.01.
a
2
Matrix compression factor. With this factor, r
2
can reduce less due to failure in matrix compression. Experiments
show that certain materials show less degradation of the matrix properties in compression than in tension. See for
example [Ref. 27]. For the case that F
mc
indicates failure in matrix compression, Equation 7-78 is modified into
(7-79)
a
2
defaults to 0; so the default is full reduction
a
3
Shear stiffness factor. This factor is used for taking into account the effect that the shear stiffness G
12
can reduce
less than the matrix stiffness E
2
. See [Ref. 28] for experimental data on this subject. With F
m
indicating a matrix failure
we have
(7-80)
The combined effect of a
2
and a
3
on the shear stiffness reduction for the case of matrix compression failure is then
a
4
E
33
reduction from fiber failure: This factor controls the reduction of E
33
due to fiber and matrix failure. The
default is as mentioned above that E
33
reduces due to fiber failure. With this factor this can be changed to vary linearly
with fiber and matrix failure. With F
f
indicating a fiber failure and F
m
a matrix failure we have
(7-81)
r
3
r
1
=
r
4
r
2
=
r
5
r
6
r
4
= =
r
2
1 a
2
( ) 1 e
1 F
mc
( ) =
r
4
1 a
3
( ) 1 e
1 F
m
( ) =
r
4
1 a
2
( ) 1 a
3
( ) 1 e
1 F
mc
( ) =
r
3
1 a
4
( ) 1 e
1 F
f
( ) a
4
1 e
1 F
m
( ) =
381
CHAPTER 7
Material Library
a
5
Shear reduction from fiber failure: With this factor it is possible to control the reduction of the shear stiffness
due to fiber failure. By default it only reduces due to matrix failure. With F
f
and F
m
as in the previous we have
(7-82)
The Hashin-fabric failure criterion reduces the first three reduction factors from the respective failure index. The three
shear reduction factors are taken from the worst of the first three factors. The factors a
2
through a
5
are not used for this
criterion.
In addition, it is also possible to use the UPROGFAIL user subroutine to explicitly define the reduction factors r
1

through r
6
.
Model 2 Selective Immediate Degradation
This model uses selective degradation just as Model 1, but the stiffness is abruptly decreased. As soon as failure is
indicated, the stiffnesses are set to a
1
the residual stiffness factor. The same rules as in Model 1 for how the different
factors are defined depending on the type of failure is applied here.
Model 3 Original Marc Method
This is the original method in Marc.
1. Upon failure, the material moduli for orthotropic materials at the integration points are set to the smallest of the
original moduli, and the smallest is set to 10% of the original.
2. Upon failure, for isotropic materials, the failed moduli are taken as 10% of the original moduli.
3. If there is only one modulus, such as in a beam or truss problem, the failed modulus is taken as 10% of the
original one.
The different options are flagged through the FAIL DATA model definition option. The reduction factor is available for
postprocessing as the element post code number 80, Damage. This single value contains the smallest of the six
reduction factors.
Mixture Model
The mixture models allow the user to define a composite material that consists of multiple components for all element
types. Effective material properties are formed based upon the volume fraction of each material component. There are
three variations of the mixture model which have consequences on the type of material that may be used in each
component. The component material is defined using the conventional ISOTROPIC, ORTHOTROPIC,
ANISOTROPIC, MOONEY, etc. options. The mixture model acts as a continuum in the sense that debonding between
the components is not considered in any of the models.
Mixture Model 1
This model is applicable to linear elastic materials and heat transfer. An effective material property is formed as
r
4
1 a
5
( ) 1 e
1 F
m
( ) a
5
1 e
1 F
f
( ) =
E
ef f
V
f
i
E
i
i 1 =
N
=

where is the value of the ith component, is the volume fraction of the ith component, and is the number of
components. The same summation is done for all properties. If an orthotropic component is used, it is done for all nine
material properties.
Note that for orthotropic or anisotropic properties, is assumed that all components have the same preferred directions.
If the component has temperature dependent properties, then the temperature dependence is evaluated for each
component which can be expressed as
The stresses obtained are an effective stress associated with the mixture material and not the stress in each component.
Mixture Model 2
This model is applicable to linear elastic materials and heat transfer. An effective material stress-strain relationship is
formed as
where
For heat transfer, analogous equations are used:
Again, temperature dependent properties are dealt with on a component basis, and orthotropic or anisotropic mixtures
must have the same preferred orientations.
Mixture Model 3
This model is applicable to nonlinear material behavior. This requires additional allocation of memory to store the
associated state variables (plastic strain, shift tensors, etc.) for each component. It is similar to model 2 in that an
effective material stress-strain relationship is formed as
E
i
V
f
i
N
E
ef f
T ( ) V
f
i
E
i
T ( )
:
N
=
D
eff
V
f
i
D
i
i 1 =
N
i
D
i
=
ef f
D
ef f
=
q
i
D
i
T =
q
ef f
D
ef f
T =
383
CHAPTER 7
Material Library
but now, includes material nonlinear effects. for the case of elastic-plastic materials, we now have two types of
quantities which are not completely consistent with one another. On a component level:
where again the indicates the component number which is the correct expression. We also calculate effective
quantities. These quantities are accumulated in the usual manner.
There are also effective quantities calculated as:
which are also accumulated.
where the is the equivalent plastic strain. Note that the effective plastic strain is not used in a subsequent
calculation and is only provided for output.
Given a mixture of two identical nonlinear materials at a 50%-50% mix, the output of the effective equivalent plastic
strain will not be the same as for a solitary material.
D
eff
V
f
i
D
i
i 1 =
N
=
D
i
i
D
el
i

t h
i

pl
i
( ) =
i
n 1 +
i
n
i
i
+ =
n 1 +
pl i ( )
n
pl i ( )
pl i ( )
+ =
ef f
V
f
i
i
i 1 =
N
pl ef f ( )
V
f
i
pl i ( )
i 1 =
N
pl ef f ( )
V
f
i
pl i ( )
i 1 =
N
n 1 +
pl ef f ( )
n
pl ef f ( )
pl eff ( )
+ =
n 1 +
pl ef f ( )
n
pl ef f ( )
pl eff ( )
+ =
pl

Mixture model 3 can be used with most nonlinear material behavior, but there are a few restrictions in this release
which include:
All damage models introduced via the DAMAGE option
Gasket material
Soils
User-defined generalized stress-strain law.
ORNL
Rigid Plastic
Viscoelasticity
Cohesive
Grain size effects
Thermo-Pore
Mass Density
The effective mass density for all of the mixture models is calculated as
Specific Heat
The effective specific heat for all of the mixture models is calculated as
Coefficient of Thermal Expansion
There are two ways to calculate the coefficient of thermal expansion; the first directly uses the rule of mixtures.
The second formulation:
where, again, if temperature dependent material properties are used, these are applied at the component level.
ef f
V
f
i
i
i 1 =
N
=
C
p
ef f
V
f
i
C
p
i
i 1 =
N
ef f
i
i 1 =
N
ef f
E
i
V
f
i
E
i
V
f
i
------------------------------------ =
ef f
( )
E
i
T ( ) V
f
i
i
T ( )
E
i
T ( ) V
f
i
------------------------------------------------------ =
385
CHAPTER 7
Material Library
Mixtures and Composites
It is possible to have a shell element or a composite brick element where the layer of a composite material may be a
mixture model that consists of multiple components.
Restrictions
If the VOID CHANGE or POROSITY CHANGE is used to define a state variable, and the material properties are a
function of the void ratio, then all components will use the same void ratio.
A mixture material cannot have as a component another mixture material.
Mixture materials should not be used with the PSHELL option.
Mixture materials cannot be used with design sensitivity or design optimization where the design variable is an elastic
property in the mixture material or one of the components.
Gasket
Engine gaskets are used to seal the metal parts of the engine to prevent steam or gas from escaping. They are complex
(often multi-layer) components, usually rather thin and typically made of several different materials of varying
thickness. The gaskets are carefully designed to have a specific behavior in the thickness direction. This is to ensure
that the joints remain sealed when the metal parts are loaded by thermal or mechanical loads. The through-thickness
behavior, usually expressed as a relation between the pressure on the gasket and the closure distance of the gasket, is
highly nonlinear, often involves large plastic deformations, and is difficult to capture with a standard material model.
The alternative of modeling the gasket in detail by taking every individual material into account in the finite element
model of the engine is not feasible. It requires a lot of elements which makes the model unacceptably large. Also,
determining the material properties of the individual materials might be cumbersome.
The gasket material model addresses these problems by allowing gaskets to be modeled with only one element through
the thickness, while the experimentally or analytically determined complex pressure-closure relationship in that
direction can be used directly as input for the material model. The material must be used together with the first-order
solid composite element types 149 (three-dimensional solid element), 151 (two-dimensional plane strain element) or
152 (two-dimensional axisymmetric element). In that case, these elements consists of one layer and have only one
integration point in the thickness direction of the element.
Gaskets can be used in mechanical, thermal or thermo-mechanically coupled analyses. The usual staggered scheme of
a heat transfer pass followed by a structural pass is used for coupled analyses. For the heat transfer part, the elements
used to model the gaskets are type 175 (three-dimensional first-order solid element), type 177 (two-dimensional first-
order planar element), or type 178 (two-dimensional first-order axisymmetric element).
Constitutive Model
The behavior in the thickness direction, the transverse shear behavior, and the membrane behavior are fully uncoupled
in the gasket material model. In subsequent sections, these three deformation modes are discussed.
Local Coordinate System
The material model is most conveniently described in terms of a local coordinate system in the integration points of
the element (see Figure 7-14). For three-dimensional elements, the first and second directions of the coordinate system

are tangential to the midsurface of the element at the integration point. The third direction is the thickness direction of
the gasket and is perpendicular to the midsurface. For two-dimensional elements, the first direction of the coordinate
system is the direction of the midsurface at the integration point, the second direction is the thickness direction of the
gasket and is perpendicular to the midsurface, and the third direction coincides with the global three direction.
Figure 7-14 The Location of the Integration Points and the Local Coordinate Systems in Two- and Three-dimensional
Gasket Elements
In a total Lagrange formulation, the orientation of the local coordinate system is determined in the undeformed
configuration and is fixed. In an updated Lagrange formulation, the orientation is determined in the current
configuration and is updated during the analysis.
Thickness Direction - Compression
In the thickness direction, the material exhibits the typical gasket behavior in compression, as depicted in Figure 7-15.
After an initial nonlinear elastic response (section AB), the gasket starts to yield if the pressure p on the gasket exceeds
the initial yield pressure p
y0
. Upon further loading, plastic deformation increases, accompanied by (possibly
nonlinear) hardening, until the gasket is fully compressed (section BD). Unloading occurs in this stage along nonlinear
elastic paths (section FG, for example). When the gasket is fully compressed, loading and unloading occurs along a
new nonlinear elastic path (section CDE), while retaining the permanent deformation built up during compression. No
additional plastic deformation is developed once the gasket is fully compressed.
The loading and unloading paths of the gasket are usually established experimentally by compressing the gasket,
unloading it again, and repeating this cycle a number of times for increasing pressures. The resulting pressure-closure
data can be used as input for the material model. If the pressure also varies with temperature, then pressure-closure
data at different temperatures can be provided. The user must supply the loading path and may specify up to ten
unloading paths. In addition, the initial yield pressure p
y0
must be given. The initial yield pressure can also be varied
with temperature and spatial coordinates. The loading path should consist of both the elastic part of the loading path
and the hardening part, if present. If no unloading paths are supplied or if the yield pressure is not reached by the
loading path, the gasket is assumed to be elastic. In that case, loading and unloading occurs along the loading path.
1
2
Midsurface
Integration Point
1
2
3
Midsurface
Integration Point
387
CHAPTER 7
Material Library
Figure 7-15 Pressure-closure Relation of a Gasket
The loading and unloading paths must be defined using the TABLE model definition option (see Marc Volume C:
Program Input) and must relate the pressure on the gasket to the gasket closure. Optionally, at each closure value, the
loading and unloading paths can also be defined as functions of temperature and spatial coordinates using multi-variate
TABLEs. The unloading paths specify the elastic unloading of the gasket at different amounts of plastic deformation;
the closure at zero pressure is taken as the plastic closure on the unloading path. If unloading occurs at an amount of
plastic deformation for which no path has been specified, the unloading path is constructed automatically by linear
interpolation between the two nearest user supplied paths. The unloading path, supplied by the user, with the largest
amount of plastic deformation is taken as the elastic path at full compression of the gasket.
For example, in Figure 7-15, the loading path is given by the sections AB (elastic part) and BD (hardening part) and
the initial yield pressure is the pressure at point B. The (single) unloading path is curve CDE. The latter is also the
elastic path at full compression of the gasket. The amount of plastic closure on the unloading path is c
p1
. The dashed
curve FG is the unloading path at a certain plastic closure c
p
that is constructed by interpolation from the elastic part
of the loading path (section AB) and the unloading path CD.
The compressive behavior in the thickness direction is implemented by decomposing the incremental gasket closure
into an elastic and a plastic part:
(7-83)
Of these two parts, only the elastic part contributes to the pressure. The constitutive equation is given by the following
equation:
.
Here, is the consistent tangent to the pressure-closure curve.
Plastic defomation develops when the pressure equals the current yield pressure . The latter is a function of the
amount of plastic deformation developed so far and is given by the hardening part of the loading path (section BD in
Figure 7-15).
p
y0
p
y1
p
y
c
p0
c
p1
c
p
c
y0
c
y1
c
y
A
B
C
D
E
F
G
Gasket Closure Distance c
G
a
s
k
e
t

P
r
e
s
s
u
r
e
p
loading path
unloading path
c c
e
c
p
+ =
p D
c
c
e
D
c
c c
p
( ) = =
D
c
p p
y

Initial Gap
The thickness of a gasket can vary considerably throughout the sealing region. Since the gasket is modeled with only
one element through the thickness, this can lead to meshing difficulties at the boundaries between thick regions and
thin regions. The initial gap parameter can be used to solve this. The parameter basically shifts the loading and
unloading curves in the positive closure direction. As long as the closure distance of the gasket elements is smaller
than the initial gap, no pressure is built up in the gasket. The sealing region can thus be modeled as a flat sheet of
uniform thickness and the initial gap parameter can be set for those regions where the gasket is actually thinner than
the elements of the finite element mesh used to model it. The initial gap can optionally be varied as a function of spatial
coordinates by using a table.
Thickness Direction - Tension
The tensile behavior of the gasket in the thickness direction is linear elastic and is governed by a tensile modulus .
The latter is defined as a pressure per unit closure distance (that is, length). can optionally be varied as a function
of temperature and spatial coordinates by using a multi-variate table.
Transverse Shear and Membrane Behavior
The transverse shear is defined in the 2-3 and 3-1 planes of the local coordinate system (for three-dimensional
elements) or the 1-2 plane (for two-dimensional elements). It is linear elastic and characterized by a transverse shear
modulus .

can optionally be varied as a function of temperature and spatial coordinates by using a multi-variate
table.
The membrane behavior is defined in the local 1-2 plane (for three-dimensional elements) or the local 3-1 plane (for
two-dimensional elements) and is linear elastic and isotropic. Youngs modulus and Poissons ratio that
govern the membrane behavior are taken from an existing material that must be defined using the ISOTROPIC model
definition option. Multiple gasket material can refer to the same isotropic material for their membrane properties (see
also the GASKET model definition option in Marc Volume C: Program Input).
Thermal Expansion
The thermal expansion of the gasket material is isotropic and the thermal expansion coefficient is taken from the
isotropic material that also describes the membrane behavior.
Constitutive Equations
As mentioned above, the behavior in the thickness direction of the gasket is formulated as a relation between the
pressure p on the gasket and the gasket closure distance c. In order to formulate the constitutive equations of the gasket
material, this relation (Equation (7-84)) must first be written in terms of stresses and strains. This depends heavily on
the stress and strain tensor employed in the analysis. For small strain analyses, for example, the engineering stress and
strain are used. In that case, the incremental gasket closure and the incremental pressure are related to the incremental
strain and the incremental stress by
and (7-84)
in which h is the thickness of the gasket.
D
t
D
t
G
t
G
t
E
m

m
c h = p =
389
CHAPTER 7
Material Library
7
Material Library
The resulting constitutive equation for three-dimensional elements, expressed in the local coordinate system of the
integration points, now reads
(7-85)
in which . For two-dimensional elements, the equation is given by
(7-86)
For large deformations in a total Lagrange formulation, in which the Green-Lagrange strains and the second Piola-
Kirchhoff stresses are employed, as well as in an updated Lagrange environment, in which the logarithmic strains and
Cauchy stresses are being used, similar but more complex relations can be derived.
Nonlinear Hypoelastic Material
The hypoelastic model is able to represent a nonlinear elastic (reversible) material behavior. For this constitutive
theory, Marc assumes that
(7-87)
where is a function of the mechanical strain and is a function of the temperature.
The stress and strains are true stresses and logarithmic strains, respectively, when used in conjunction with the LARGE
STRAIN.
11
22
33
12
23
31
E
m
1
m
2
-----------------
m
E
m
1
m
2
----------------- 0 0 0 0
m
E
m
1
m
2
-----------------
E
m
1
m
2
----------------- 0 0 0 0
0 0 C 0 0 0
0 0 0
E
m
2 1
m
+ ( )
------------------------- - 0 0
0 0 0 0 G
t
0
0 0 0 0 0 G
t
11
22
33

33
p
12
23
31
=
C hD
c
=
11
22
33
12
E
m
1
m
2
----------------- 0
m
E
m
1
m
2
----------------- 0
0 C 0 0
m
E
m
1
m
2
----------------- 0
E
m
1
m
2
----------------- 0
0 0 0 G
t
11
22

22
p
33
12
=
i j
L
i j kl
kl
g
i j
+ =
L g

When used in conjunction with the LARGE DISP option only, Equation (7-87) is expressed as
(7-88)
where are the Green-Lagrangian strain and second Piola-Kirchhoff stress, respectively.
A HYPOELASTIC or NLELAST model definition option is necessary to invoke this model. These models can be used
with any stress element, including Herrmann formulation elements.
The tensors and are defined by you in the HYPELA2 user subroutine or via the NLELAST model definition option.
HYPELA2 User Subroutine
In order to provide an accurate solution, should be a tangent stiffness evaluated at the beginning of the iteration. In
addition, the total stress should be defined as its exact value at the end of the increment. This allows the residual load
correction to work effectively.
In HYPELA2, additional information is available regarding the kinematics of deformation. In particular, the
deformation gradient ( ), rotation tensor ( ), and the eigenvalues ( ) and eigenvectors ( ) to form the stretch tensor
( ) are also provided. This information is available only for the continuum elements namely: plane strain, generalized
plane strain, plane stress, axisymmetric, axisymmetric with twist, and three-dimensional cases.
For more information on the use of user subroutines, see Marc Volume D: User Subroutines and Special Routines.
NLELAST Model Definition
The NLELAST model definition in Marc supports the Nastran NLELAST capability and simplified nonlinear elasticity
models. These models all have the form that a strain energy function does not exist but have been found to be useful
in solving engineering simulations, where only a limited amount of test data is available. The theory and algorithms
are adequate to trace the stress-strain curve accurately for the uniaxial loading cases. However, the theory is not based
on the classical theory of finite elasticity. Consequently, some of the constitutive relations may be violated in the
multiaxial stress state and, as a result, limit the use of this model to small strain.
Six models are implemented in Marc to support nonlinear elasticity based upon the input of a uniaxial stress strain law.
These models are summarized as:
1. Nastran like NLELAST model based upon MATS1 input option.
2. Hypoelastic strain invariant model.
3. Model based upon working in principal strain space which results in induced anisotropic behavior.
4. Linear elasticity with tension and/or compression cut-off.
5. Bi-modulus elasticity with tension and/or compression cut-off.
6. Nonlinear orthotropic elasticity based upon non-rotating preferred directions.
The HYPELA2 user subroutine can be used if special material behavior is required.
S
i j L
i j kl
E
kl
g
i j
+ =
E S ,
L g
L
F R N
U
391
CHAPTER 7
Material Library
Basic assumptions and definitions
A material model is called elastic if the stress is a function of the strain. This implies two important properties: the
material is reversible and the stress at a point in the material depends only on some measure of strain at that point.
Elastic materials implemented with this option are in the hypoelastic framework.
This group of nonlinear elastic materials is therefore grouped in Mentat under Materials -> HYPOELASTIC.
The nonlinearity in the material models introduced here results from the definition of the stress as a function of the
strain path. The models are anticipated to behave correctly in the small strain region only (less than 5%).
Input of uniaxial stress-strain data
In the next sections the theory and input is described for each of the different material models. As a basis it is assumed
that the user has a set of engineering stresses and engineering strains available. This means that in the uniaxial test:
The measured elongations are compared to the original length of the test specimen, resulting in:
The measured forces are compared to the original cross-section of the test specimen, resulting in:
For analysis without large incremental or total rotation or large incremental or total strains, the engineering values can
be used as is. In some models, certain characteristics still have to be derived. This is described in the sections per
material type.
For analysis where the LARGE DISP, UPDATE, and FINITE, or PLASTICITY,3, or LARGE STRAIN parameters are
used, the uniaxial test data cannot be used as is. Engineering strain and stress are measures that are invalid once the
strain in the analysis model is no longer small (approximately 5%). For these large strain simulations, the true stress
and strain definition have been adopted in Marc. This nonlinear strain measure is dependent upon the final length of
the model. Since the NLELAST material models are only valid in this small range it is not necessarily required to
convert the engineering data.
Converting engineering strain/stress to true strain/stress
True strain is defined as:
By integration, it can be shown that this directly defines the true strain as:
(7-89)
Similarly, the uniaxial engineering stress data can be converted to true stress (also known as Cauchy stress) for
incompressible materials by means of:
(7-90)
and, in general, by
engineering
L L
0
=
engineering
F A
0
=
true
L L
}
=
true
1
engineering
+ ( ) ln =
true

engineering
1
engineering
+ ( ) =
t rue

engineering
1 v
engineering
( )
2
=

By using Equations (7-89) and (7-90), the uniaxial test data can quickly be transformed into true stress/strain data.
Figure 7-16 shows the deviation of engineering stress/strain compared to the engineering definition. In the remainder
of this section on nonlinear elasticity the engineering stresses and strains can be mutually exchanged with their true
counterpart. The tables can be either engineering or true data depending on the options activated in the analysis.
Figure 7-16 Comparison Between Engineering and True Stress/Strain
Type 1: Nastran like nonlinear elasticity model
The stress curve is supplied in a table with independent variable mechanical strain (id=69). An additional independent
variable in the table may be temperature (id=12). Note that a table with independent variable mechanical strain is a
requirement for the analysis.
It is optional to define Poissons ratio as a function of temperature. The initial Youngs modulus is the slope of the
equivalent stress-strain curve is derived by Marc from the table at = 0. This imposes the continuous definition of the
slope at = 0.
Theoretical Basis
The nonlinear elastic capability in Nastran was designed to satisfy the equivalence of the deformation work per unit
volume in the simple tension to the strain energy per unit volume (conservation of energy), while the work done for
deformation may be defined by a stress-strain curve in simple tension, i.e.,
It is further assumed that the effective strain may be defined by:
Comparison between engineering and true stress/strain
-200
-150
-100
-50
0
50
100
150
200
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
true or engineering strain
t
r
u
e

o
r

e
n
g
i
n
e
e
r
i
n
g

s
t
r
e
s
s
True Stress
Engineering Stress
d
}
<> d { }
}
=
1
2
--- E
2
1
2
--- <> D
e
[ ] d { } =
393
CHAPTER 7
Material Library
From the total differential of the above equation, we obtain
Substituting the latter in the first equation above, where stresses may be expressed in terms of total strains, i.e.
The tangential matrix for such material may be obtained by differentiating the latter equation, i.e.
In the original Nastran, implementation is set to 1. In the Marc implementation, testing proved that setting this value
to 0.0 (default) improves convergence significantly.
Solution Algorithm
The user specifies the nonlinear stress-strain curve, in a TABLE parameter along with a NLELAST model
definition. The NLELAST entry does require a table for Youngs modulus; whereas, strain dependency for Poissons
ratio is optional.
The NLELAST material interfaces with the element routines with the following data:
Input: , , , ,
Output: , ,
The computational procedure is described below:
Step 1: Upon entry the new strain state is computed by:
Step 2: The effective strain is computed based on by
where
is simply formed using the Hookean law.
d
1
E
------ <> D
e
[ ] { } =
{ }

E
------ D
e
[ ] { } =
D
ne
[ ]
{ }
{ }
-------------

E
------ D
e
[ ]

E
------

------ -

---
\ .
| |
D
e
[ ] { }<> D
e
[ ]
T
+ = =
( )
{ }
old
{ }
ol d
{ } E
{ }
new
{ }
new
D
ne
[ ]
{ }
new
{ }
old
{ } + =
{ }
new
2
1
E
--- <>
e
{ } =
e
{ } D
e
[ ] { } =
D
e

Step 3: The effective stress is determined by looking-up the user-specified stress-strain curve for . .
Step 4: The new stress state is determined by scaling the stress-strain law:
Step 5: The tangential matrix is determined by
Furthermore is the slope at the current strain.
The tables for NLELAST must be supplied in the new table format.
Combining Uniaxial Tension and Compression Stress-strain Curves
Some materials exhibit appreciably different behavior in compression from that in tension even in the small strain
range. For uniaxial loading the magnitude of the strain in that direction becomes the effective strain, i.e.,
for uniaxial tension in x
for uniaxial compression in x
In order to remain compatible with the Nastran implementation of this material model, the user has the option to ignore
the uniaxial compression data, even if it is supplied on input.
In case of this unsymmetric material behavior, we need to be able to distinguish between a state of compression and a
state of tension. There are two known data points for one effective strain , namely the effective stress for uniaxial
tension and the effective stress for uniaxial compression . Some method of interpolation or extrapolation is
required to predict the effective stress for the general stress state using two known data points.
The first stress invariant has been adopted for interpolation/extrapolation as follows,
Considering that the pure shear is in the midway between simple tension and simple compression, it seems appropriate
to use the first stress invariant. Hydrostatic tension and compression cases will impose lower- and upper-bounds for
extrapolation, i.e.,
for uniaxial tension/compression
for pure shear
for hydrostatic pressure
The instantaneous modulus should be interpolated or extrapolated in the same manner.
( )
{ }
new
E
------
e
{ } =
D
ne
[ ]

E
------ D
e
[ ]

E
2
---------

------ -

---
\ .
| |

e
{ }
e
{ }
T
+ =
------ -

x
=

x
=
t
( )
c
( )
I
1
( )
I
1

x

y

z
+ + =
I
1

x
=
I
1
0 =
I
1
3p =
-------
\ .
| |
395
CHAPTER 7
Material Library
Solution Algorithm for Bilateral Stress-strain Relations
The new stress state is proportional in magnitude to the effective stress , which should be determined as follows:
1. Compute the effective stress based on ; i.e.,
for 3-D
for plane stress
for plane strain
2. Compute the first invariant of :
where for plane stress
3. Determine the ratio by normalizing

by , i.e.,
where signifies the relative distance from the midpoint of and at . It would be implausible to process
a large value of r (such is the case with a hydrostatic load). Therefore, , will be confined to a plausible range,
. The value will be reset to the limit if lies outside the range.
4. Look up the user specified stress-strain curve in the TABLE entry and determine and ; i.e.,
5. Determine based on , . and ; i.e.,
For the tangent matrix the instantaneous modulus should be determined using the same ratio (r) as follows:
1. Compute the instantaneous slope at for tension, i.e.,
for
where is the i-th data point in the TABLE entry
e
( )
e
{ }
e
{ } D
e
[ ] { } =
1
2
---
x

y
( )
2
y

z
( )
2
z

x
( )
2
+ + [ ] 3
xy
2
yz
2
zx
2
+ + ( ) + =

x
2
y

y
2
3
xy
2
+ + =
1
2
---
x

y
( )
2
y

z
( )
2
z

x
( )
2
+ + [ ] 3
xy
2
+ =
I
x

y

z
+ + =
z
0 =
r ( ) I
I

c
r
I
1
e
----- - =
r
c

t

r
1 r 1 r 1 = ( ) r
t

c
t
( ) =
c
( ) =
t

c
r

t

c
+
2
------------------ r
t

c
2
------------------ + =
-------
\ .
| |
-------
\ .
| |
t
y
i 1 +
y
i
x
i 1 +
x
i
------------------------ = x
i
x
i 1 +

x
i
y
i
, ( )

2. Compute the instantaneous slope at for compression; i.e.,
for
3. Determine based on , and , i.e.,
Subsequently the values can now be used as in the tension-only algorithm.
Type 2: Strain Invariant Model
This model, like others in Marc is based upon the incremental theory where
And

is a Hookes law based upon the current values of the Youngs modulus and Poissons ratio. In this model
these quantities are a function of the strain invariants , , .
For this model, Marc computes the three strain invariants in each integration point of the element. The three strain
invariants are independent variables in the Youngs modulus vs strain-invariant table that is supplied on input. In
principle a table with all three strain invariants can be supplied if necessary, however, when converting the uniaxial
data into Young vs strain-invariant tables it is usually sufficient to choose one. The choice may depend on the element
type used and the need to model non-symmetric behavior in compression and tension. This is explained further below.
Marc looks up Youngs modulus in the user-supplied table and forms a Hookean stress-strain law.
In order to arrive at proper input, additional data conversion of the uni-axial data needs to be done. Starting from
uniaxial (engineering or true) stress strain data, the elasticity modulus (Young) needs to be derived for a number of
engineering or true strain data points. See Figure 7-17.
Figure 7-17 Uniaxial Stress and Young vs Strain

-------
\ .
| |
c
y
j 1 +
y
j
x
j 1 +
x
j
------------------------ = x
i
x
i 1 +

------ -

-------
\ .
| |
t
-------
\ .
| |
c
r
------ -
1
2
---

------ -
\ .
| |
t
------ -
\ .
| |
c
+
r
2
---

------ -
\ .
| |
t
-------
\ .
| |
c
+ =
d C
el
d
el
=
C
el
I
1
I
2
I
3
Uniaxial stress & Young vs Strain
0
20
40
60
80
100
120
140
160
0 0.01 0.02 0.03 0.04 0.05
Strain
S
t
r
e
s
s
0
2000
4000
6000
8000
10000
Y
o
u
n
g
'
s

m
o
d
u
l
u
s
engineering strain
Young's modulus
397
CHAPTER 7
Material Library
The definition of these invariants in Marc is as follows:
Because only two elastic moduli (Youngs modulus and Poisson ratio) which are functions of the invarients are
defined, the material will behave isotropically in the simulation. In the input table, the three strain invariants are
indicated by means of IDs 70, 71, and 72.
For uniaxial and plane stress elements, a special treatment is followed in Marc. The out-of-plane strains are not
computed internally. Therefore incompressibility (poisson = 0.5, see Figure 7-18) is assumed, such that : the
first invariant is very closely related to the hydrostatic strain . The other invariants automatically follow from
this assumption.
Figure 7-18 Influence of Incompressibility on First Strain Invariant
For a uniaxial element:
and
I
1
trace [ ]
xx

yy

zz
+ + = =
I
2
1 2 trace [ ] [ ] I
1
2
( ) { }
xy
2
yz
2
zx
2
xx
yy

yy
zz

zz
xx
+ + ( ) + + = =
I
3
det [ ]
xx
yy
zz
2
xy
yz
zx

yy
yz
2

yy
zx
2
zz
xy
2
+ + + = =
I
1
0 =
I
1
3 ( )
Influence of incompressibility on first strain invariant
-0.02
-0.015
-0.01
-0.005
0
0.005
0.01
0.015
0.02
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
Engineering strain
F
i
r
s
t

s
t
r
a
i
n

i
n
v
a
r
i
a
n
t
First strain invariant
(poisson = 0.3)
First strain invariant
(poisson = 0.5)
xy

yz
0 = =
yy

zz
=
I
1
0 >
yy

zz
1
2
---
xx
= = = =
I
2
3
4
---
xx
2
=
I
3
1
4
---
xx
3
=

For a plane stress element:
It is important to realize that the incompressibility assumption for different elements has a direct impact on the material
data supplied in the input deck. For different element types, the input data derived from the uniaxial test data will be
different. We compare the behavior of a plain stress element (e.g., element 3), a hexahedral element (e.g., element 7),
a plane strain element (e.g., element 11) and a beam element (e.g., element 52).
For the materials, we define a Poissons ratio of 0.3 and we prescribe a fixed displacement which brings us into the
nonlinear part of the stress-strain curve. Note that Poissons ratio may also vary with the uniaxially measured strain,
in this case this has to be accounted for in the 2-D and 3-D elements.
3-D Element Three Direct Components of Strain
For the hexahedral element, we may choose any of the invariants as independent variable. In order to get a
corresponding behavior with the uniaxial test, we have to supply the first invariant in the following form:
2-D Elements Two Direct Components of Strain (= Plane Stress)
Although the plane stress elements in Marc have a third strain component unequal to zero, Marc does not compute it.
It has been chosen that the user input for the stress-strain data has to follow the incompressibility assumption
. In this case, we may use the definition of the second invariant in order to define a dependency on strain,
supposing a uniaxial loading:
(uniaxiality)
from , we deduce (see above):
yz

zx
0 = =
I
1
0 >
zz

xx

yy
+ ( ) = = =
I
2

xy
2
xx
2
yy
2
xx
yy
+ + + =
I
3

xx

yy
+ ( )
xy
2
xx
yy
( ) =
xx

engineering
=
yy

zz

engineering
= =
I
1

xx

yy

zz
+ + 1 2 ( )
engineering
or: ( I
1
1 2 ( )
true
) = = =
I
1
0 = ( )
xx

engineering
=
xy

yz

zx
0 = = =
yy

engineering
=
I
1
0 =
I
2

xx
2
yy
2
xx
yy
+ + =
I
2

2
1 + ( )
engineering
2
=
399
CHAPTER 7
Material Library
2-D Elements Three Direct Components of Strain (= Plane Strain)
For the plane strain elements, the third direct strain term is zero. In this case, incompressibility is not assumed and the
strain invariants can be easily derived from the conditions: and under uniaxial loading:
(uniaxiality and plain strain)
1-D Elements One Direct Components of Strain
For the beam and truss elements, we cannot use the first invariant as independent variable either, since . We
may quickly derive that:
(or )
Loading Under Tension/Compression
It must be noted that for both beam and plane stress elements a nonsymmetric behavior in compression cannot be
described by means of the second invariant because (see Figure 7-19).
The third invariant can be used for this purpose; e.g., for the 1-D elements:
(or )
And finally for the 2-D plane stress elements:
(uniaxiality)
yy
0 =
xx
0 =
xx

engineering
=
xy

yz

zx

zz
0 = = = =
yy
1 ( )
engineering
=
I
1

xx

yy

zz
+ + 1 1 ( ) + { }
engineering
= =
I
2

xx
yy
=
I
2
1 ( )
engineering
2
=
I
1
0 =
I
2
3 4
engineering
2
= I
2
3 4
true
2
=
I
2
0
I
3
1 4
engineering
3
= I
3
1 4
true
3
=
xx

engineering
=
xy

yz

zx
0 = = =
yy

engineering
=

Figure 7-19 Second Strain Invariant (>0)
from , we deduce
Data Approximation
Finally it is important to realize that the second and third invariants are nonlinear functions of the strain. As a result,
even a linear stress-strain curve will form a polynomial of the second or third order for the second and third strain
invariant respectively. It is, therefore, wise to apply the load in a sufficient number of increments, even though the
actual behavior is only linear. This is shown in Figure 7-20.
Type 3: Principal Strain Space Model
In the principal strain space theory, the strain tensors principal values are now computed. The compliance matrix is
formed using a total strain approach; i.e., the secant modulus is computed for each principal strain by looking up the
stress in the user defined table. This results in three values for the modulus. These values are then used to form an
orthotropic stress-strain law together with poissons ratio and a shear modulus, which are also user defined.
The definition of the orthotropic stress-strain law is described in Volume C (chapter 3). It must be noted that the
constants have to fulfill certain requirements in order to guarantee stability. These criteria are checked by Marc.
However, if stability cannot be guaranteed, Marc will continue to solve the increment at hand, and moreover that it
will continue if the increment converges. From this point on, the solution may not be trustworthy anymore and care
should be taken.
Second strain invariant (> 0)
0.0000E+00
2.0000E-04
4.0000E-04
6.0000E-04
8.0000E-04
1.0000E-03
1.2000E-03
1.4000E-03
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
Engineering strain
S
e
c
o
n
d

s
t
r
a
i
n

i
n
v
a
r
i
a
n
t
I
1
0 =
I
3

xx

yy
+ ( )
xy
2
xx
yy
( ) =
I
3
1 ( )
engineering
3
401
CHAPTER 7
Material Library
Figure 7-20 Influence of Number of Data Points on the Third Strain Invariant
Theoretical Background
The Principal strain space theory can be defined as the incremental stress strain law which is based upon an orthotropic
Hookes law expressed with respect to a set of axis aligned with the principal axes based upon the strain tensor. The
moduli ( , , ) are evaluated at the current values of the principal strains.
Consider to be the reference material system of the element. For a continuum element, this is the global system,
while for a shell element this is the , , system. Now, consider the total strain at the beginning of the increment
. One can obtain the principal strains and the eigenvectors of this system.
The principal strains are ,

,

and the eigenvectors are , , , The matrix that transforms the system to
the principal system is .
The state is then rotated into the principal, this includes the stresses, strains, thermal strains and creep strains.
, etc.
Based upon the principal strains, the material properties for the current value of the principal strains are determined.
The evaluated material constants are tested to make sure stability is not violated, also positive definiteness of the
compliance matrix is checked.
Then the constitutive law is evaluated based upon the current Youngs moduli, Poisson ratios, and shear modulus.
Based upon either the estimated or calculated incremental strain, the incremental stress in principal space is
determined. Subsequently the total stress in principal space is calculated and the quantities are rotated back to the
system.
Influence of # of datapoints on third strain invariant
-2.0000E-05
-1.5000E-05
-1.0000E-05
-5.0000E-06
0.0000E+00
5.0000E-06
1.0000E-05
1.5000E-05
2.0000E-05
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
Engineering strain
T
h
i
r
d

s
t
r
a
i
n

i
n
v
a
r
i
a
n
t
E G
E
1

2

3
1

2

3
n
1
n
2
n
3
E
R
n 1 +
R
T
R =
E
n 1 +
R
n 1 +
p
R
T
R
p
p
+ ( )R
T
= =

The constitutive law is also rotated back into the system.
This is then used in the tangent stiffness matrix calculation.
Type 4: (Bimodulus) Elasticity Model
This is conventional Hookes law, where the Youngs modulus and Poisson ratio may be function of temperature or
position, but not strain or stress. Further below, a description is given of the treatment regarding tension and
compression limits.
Type 5: (Bimodulus) Elasticity Model with Either No Tension, Limited Tension, No Compression,
or Limited Compression Enhancements
In this model, the hydrostatic strain will be calculated and if the value is positive, the tensile values of Youngs modulus
and Poissons ratio will be used. If the hydrostatic strain is 0, as in the very beginning of the analysis, the tensile values
will be used.
For Model 4
Given total strains, the total stresses are calculated based upon tensile properties. This results in and ,
For Model 5
To distinguish a tension or compression stress state the hydrostatic strain is computed.
If the hydrostatic strain is positive the total stresses are calculated based upon tensile properties. Resulting in and
If the hydrostatic strain is negative, the total stresses are calculated based upon compressive properties. This results in
and ,
E
C RC
p
R
T
=
Linear Elasticity Bimodulus Elasticity
C
n 1 +
C
ten
n 1 +
C
com

n 1 +
403
CHAPTER 7
Material Library
For Either Model 4 or 5
The treatment of the cut-offs works as follows, if they are defined in the simulation.
The principal stresses ,

,

and the eigenvectors are , , are obtained.
If Tensile stress limit, then principal direction will be reduced.
If Compressive stress limit, then principal direction will be reduced.
The reduction implies that a new secant modulus is computed for the current reduced direction. The maximum secant
modulus in one of the reduced directions is used to form a new stress-strain law.
The transformation matrix: is obtained between Element system and Principal system .
A stiffness matrix is recreated using the secant modulus of the current limit stress vs strain state.This is also done to
improve stability.
No Tension Material No Compression Material
Bimodulus with both Tension and Compression Cut-off
1

2

3
n
1
n
2
n
3
I
> i
I
< i
R
ep
E P
C
z
R
ep
C
p
R
epT
=

Rotate stresses to principal system and maximize the stress to the limit in ith component, and then rotate back into
element system:
Note that there is no need to form

because is from isotropic elasticity, and is a proper orthogonal matrix.
Therefore the stiffness matrix can be formed in the global system with the limited secant modulus.
Type 6: Nonlinear Orthotropic Elasticity Model
This model is based upon orthotropic Hookes law, written with respect to a fixed preferred material orientation. The
difference is that the nine constants, , , , , , , , , may be a function of the spatial
position, temperature, and the strain in the associated direction. This can be expressed as:
, etc.
, etc.
, etc.
The material properties are evaluated based upon the current value of the independent variables, and the Hookes law
is used to create the current tangent modulus. The incremental stress follows from these moduli and the incremental
strains.
The evaluated material constants are evaluated to make sure stability is not violated, also it is determined if the
compliance matrix is positive definite.
As the material orientations do not change, the usual procedure for defining them is used.
Thermo-Mechanical Shape Memory Model
NiTi alloys with near-equiatomic composition exhibit a reversible, thermoelastic transformation between a high-
temperature, ordered cubic (B2) austenitic phase and a low-temperature, monoclinic (B19) martensitic phase. The
density change and thus the volumetric are small and on the order of 0.003. The transformation strains are, thus mainly
deviatoric, of the order of 0.07-0.085. However, these small dilational strains do not necessarily lead to a lack of
pressure sensitivity in the phenomenology. The behavior of nitinol is different depending on whether the materials are
subjected to hydrostatic tension or compression.
Typical phenomenology is shown in Figure 7-21 taken from Miyazaki et al. (1981). The curves indicate that upon
cooling, the material transformation from austenite to martensite begins once the temperature is reached. Upon
further cooling, the volume fraction on martensite is a given function of temperature; the volume fraction becomes
100% martensite when the temperature is reached. Upon heating, transformation from martensite to austenite
n 1 +
p
R
epT
n 1 +
R
ep
=
n 1 +
z
R
ep
n 1 +
p z
R
epT
=
C
p
C R
C
E
11
E
22
E
33

12

23

31
G
12
G
23
G
31
E
11
E
11
X T
11
, , ( ) =
12

12
x T
12
, , ( ) =
G
12
G
12
x T
12
, , ( ) =
M
s
M
f
405
CHAPTER 7
Material Library
begins only after temperature is reached. This re-transformation is complete when the temperature is reached.
Finally, note that the four transformation temperatures are stress dependent. The experimental data indicate the ,
, , and may be approximated from their stress-free values, , , , and by
,
; and
,
.
where is the von Mises equivalent stress. At a sufficiently high temperature, often called the temperature,
transformation to martensite does not occur at any level of stress.
The transformation characteristics such as the transformation temperatures depend sensitively on alloy composition
and heat treatment.
Figure 7-21 Austenite to Martensite and Martensite to Austenite Decomposition
A
s
A
f
M
s
M
f
A
s
A
f
M
s
0
M
f
0
A
s
0
A
f
0
M
s
M
s
0

eq
C
m
-------- + =
M
f
M
f
0

eq
C
m
-------- + =
A
s
A
s
0

eq
C
a
-------- + =
A
f
A
f
0

eq
C
a
-------- + =
eq
M
d
0
77 150 200 250 300
Temperature (K)
M
s A
f
200
400
600
T
e
n
s
i
l
e

S
t
r
e
s
s

(
M
P
a
)
Stress = 0
Austenite to martensite &
martensite to austentie
decomposition
A
s
M
f
Note:
After partial
transformation,
decomposition
begins at A
s
.
M
S
A
f
50 60 70 80 90 100 0 10 20 30 40
Temperature
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
M
a
r
t
e
n
s
i
t
e

V
o
l
u
m
e

F
r
a
c
t
i
o
n

Transformation Induced Deformation
For the discussion of the thermo-mechanical response of NiTi, the data of Miyazaki et al. (1981) is shown in
Figure 7-22. Following this thermal history, it is observed that, when unstrained specimens with fully austenitic
microstructures are cooled, the transformation to martensite begins at a temperature of 190K; the transformation is
complete at 128K. This established the so-called martensite start ( ) and martensite finish ( ) temperatures at
190K and 128K, respectively. With the imposition of an applied uniaxial tensile stress, the low temperature martensite
is favored and the and temperatures increase. Upon heating a specimen with fully martensitic microstructure,
the reverse transformation is observed to begin at a temperature of 188K and to be complete at 221K. These define the
austenite start ( ) and austenite finish ( ) temperatures, respectively. Uniaxial tension tests are carried out in
temperature ranges where , , and where is defined as the temperature above
which the yield strength of the austenitic phase is lower than the stress required to induce the austenite-to-martensite
transformation.
Figure 7-22 Thermal History
In the temperature range where , the microstructures are all martensitic. The stress versus strain curves display
a smooth parabolic type of behavior which is consistent with deformation caused by the movement of defects such as
twin boundaries and the boundaries between variants. Note that unloading occurs nearly elastically and that the
accumulated deformation, caused by the reorientation of the existing martensite and the transformation of any pre-
M
s
0
M
f
0
M
s
0
M
f
0
A
s
0
A
f
0
T M
s
< M
s
T A
f
< < A
f
T T
c
< < T
c
0 0 0
100
200
300
(a) 77K (b) 153K (c) 164K
0 0 0
100
200
300
400
(d) 224K (e) 232K (f) 241K
200
400
600
(g) 263K (h) 273K (i) 276K
M
s
= 190K
A
F
= 221K
0 2 4 0 2 4 0 2 4
Strain (%)
T
e
n
s
i
l
e

S
t
r
e
s
s

(
M
P
a
)
T M
f
<
407
CHAPTER 7
Material Library
existing austenite, remains after the specimen is completely unloaded. Note also that the accumulated deformation is
entirely due to oriented martensite and this would be recoverable upon heating to temperatures above the ( )
range. This would, then, display the shape memory effect.
Pseudoelastic behavior is displayed in the temperature range . In this range, the initial microstructures
are essentially all austenitic, and stress induced martensite is formed, along with the associated deformation; upon
unloading, however, the martensite is unstable and reverts to austenite thereby undoing the accumulated deformation.
Note that, as expected, the stress levels rise with increasing temperature. In this range, the transformation induced
deformation is nearly all reversible upon unloading.
At temperatures where , plastic deformation appears to precede the formation of stress induced martensite. The
unloading part of the stress versus strain behavior displays nonlinearity and the unloading is now associated with
permanent (plastic) deformation. Permanent deformation, due to plastic deformation of the austenite, is
nonrecoverable and, if such deformation is large, shape memory behavior is lost.
Constitutive Theory
The model formulated below is based on the kinematics of small strains, although the extension to large strain is
straightforward. Accordingly, the incremental strain, , is simply the sum of the following contributions:
(7-91)
In Equation (7-91), is the incremental elastic, or lattice, strain rate; is the incremental thermal strain,
is the incremental visco-plastic strain, and is the incremental strain associated with thermoelastic phase
transformations. The incremental elastic strain is taken to be simply related to a set of elastic moduli, , and the
incremental Cauchy stress rate, , as
(7-92)
To calculate the coefficient of thermal expansion of the composite, the rule of mixtures is used as
.
In the above equations, the superscripts and refer to the austenite and martensite values, and is the volume
fraction of martensite.
Phase Transformation Strains
As noted earlier, the phase transformation induced strains are a result of the formation of oriented, stress induced,
martensite and the reorientation of randomly oriented thermally induced martensite. To account for this, is
expressed as
A
s
A
f
A
f
T T
c
< <
T T
c
>

el
t h
pl
ph
+ + + =
el
t h
pl
ph
L
L
el
=
1 f ( )
A
f
M
+ =
A M f
Ph
Ph
TRI P
TWI N
+ =

where
. (7-93)
and
. (7-94)
where and represents McCauleys bracket where , .
In Equation (7-93), represents the rate at which martensite is formed, is the magnitude of the deviatoric
part of the transformation, and is the volumetric part of the transformation strain. The function is
schematically depicted in Figure 7-23, and is a measure of the extent to which the martensite transformation strains
are aligned with the deviatoric stress. is the equivalent stress defined as:
where is the deviatoric stress.
Figure 7-23 Function
The first two terms in Equation (7-93) describe the development of transformation induced strains due to the formation
of stress induced (partially oriented) martensite. is the change of formation of austenite; for example, the rate at
which the volume fraction of martensite decreases. The last term in Equation (7-93), therefore, represents the recovery
of the accumulated phase transformation strain.
TRI P
f
+ ( )
g
eq
( )
eq
T
3
2
---
'
eq
-------- f
+ ( )
T
I f
- ( )
Ph
+ + =
TWI N
fg
eq
( )
eq
T
3'
2
eq
------------
eq
{ }
eq

ef f
g
{ } =
f f
+ ( )
f
- ( )
+ = { } x { }
1
2
---
x x +
x
----------------
\ .
| |
= x o
f
+ ( )
eq
T
T
g
eq
( )
eq
eq
3
2
---
d
:
d
=
d
2.0 1.5 1.0 0.5 0.0
-1.0
0.1
0.3
0.5
0.7
0.9
1.1
g
stress/g
0
g
eq
( )
f
- ( )
409
CHAPTER 7
Material Library
Note that there is no dilatational contribution to since is fixed. Note that the twinning strain rate is zero
when is less than , or when the magnitude of the stress change is negative ( ). Hence, can be
considered as a stress below which no twinning is possible.
The function represents the extent to which the transformation strains are coaxial with the applied deviatoric stress.
This function can be calibrated with the experimental data. Note for uniaxial stress-strain curves performed below the
martensite finish temperature, the material starts as 100% martensite, and that other than elastic strains, the
deformation is dominated by the twinning of the randomly oriented martensite.
A functional form that leads to sufficient fit to most experimental data has been implemented in Marc.
In most cases, the first term is sufficient, and a value of and yields the best results. is a stress level
used to non-dimensionalize the constants, and can be chosen such that when the . In some cases, it
is necessary to include the higher powers of equivalent stress for better experimental fits. In these, cases suggested
values for and . However, depending on the values of and , this could lead to
maxima or minima values of in the range of interest. Note that and it should be a monotonically
increasing function (an increase in the stress level should lead to an increase in the increment of the phase strains).
Thus, cut off values of are provided in Marc, such that when reaches its maximum value at the stress
level , it is held constant at the value . For the proper selection of , see the following section.
Experimental Data Fitting for Thermo-mechanical Shape Memory Alloy
The properties and transformation/retransformation behavior of Nititol depend upon alloy chemistry, microstructure
and the thermal processing applied to the specimens and eventually to the components built from them. Every time
any of the above change, it might be necessary to redo the calibration. Calibration of Nitinol experimental data works
best if, in fact, all specimens can be initially rendered as 100% austenite. The list of properties that require calibration
is given as follows:
The unstressed transformation temperatures, .
The coefficients , and that provide the stress dependence of the transformation temperatures.
The elastic constants ( and ).
Coefficients of thermal expansion, .
The calibration of the detwinning function, that provides the description of the degree to which
martensite is co-axial with the deviatoric stress state.
TWI N
f
eq

ef f
g
eq
0 <
ef f
g
g
g
eq
( ) 1 g
a
eq
g
0
--------
\ .
| |
g
b
g
c
eq
g
0
--------
\ .
| |
g
d
g
e
eq
g
0
--------
\ .
| |
g
f
+ + exp =
g
a
0 < g
b
2 = g
0
g 1
eq
g
0
g
d
2.55 or 2.75 = g
e
3 = g
c
g
e
g 0 g 1
g g g g
max
=
eq

max
g
g
0
= g
max
g
0
M
s
0
M
f
0
A
s
0
A
f
0
, , ,
C
M
C
A
E
M
E
A
v
M
, , v
A
A
,
g
eq
( )

The yield stress of the pure martensite and austenite phase ( and ), and their strain hardening
properties.
The calibration of the transformation strains, and .
Transformation Temperatures and Their Stress Dependence ( , and )
For almost any use of shape memory alloy, it is highly desirable that one knows the Transformation Temperatures
(TTRs) of the alloy. The TTRs are those temperatures at which the alloy changes from the higher temperature austenite
to the lower temperature martensite or vice versa. There are in common use with NiTi alloys to provide helpful data
to product designers Constant Load, DSC and Active . The detail procedures to obtain TTRs using the above
methods are shown in website (www.sma-inc.com). It is recommended that combined dilatometry and DSC tests be
performed on unstressed specimens of thermally processed material to establish both the unstressed transformation
temperatures and the thermal expansion coefficients. Those tests would provide a baseline set of values for
, and . Note that the TTRs are stress dependent parameters, but it is difficult, in practice, to prepare
totally unstressed samples. In order to determine the TTRs at zero stress, experimental data must be obtained at two
or more stress levels. The particular transformation point of interest can then be extrapolated to zero stress. The
estimations of TTRs,
,
and are shown in Figure 7-24. The typical range of TTRs is 200 to 100C. So, it is
difficult to recommend the default values. As references, there are examples for two different SMA materials below.
SMA 1)
, , , , ,
SMA 2)
, , , , ,
Figure 7-24 Typical Stress vs. Temperature Curve Showing the Stress Dependence of Martensite and Austenite Start
and Finish Temperature
Y
M
Y
A
eq
T
v
T
M
s
0
M
f
0
A
s
0
A
f
0
C
M
, , , , C
A
A
f
M
s
0
M
f
0
A
s
0
, , A
f
0
C
M
C
A
M
s
0
: 50C M
f
0
: 100C A
s
0
: 5C A
f
0
: 20C C
M
: 6.0Mpa/C C
A
: 8.0Mpa/C
M
s
0
: 190K M
f
0
: 128K A
s
0
: 188K A
f
0
: 221K C
M
: 5.33Mpa/K C
A
: 6.25Mpa/K
S
t
r
e
s
s
Temperature
C
M
C
A
M
f
0
M
s
0
A
s
0
A
f
0
411
CHAPTER 7
Material Library
Elastic Constants ( , and )
Literature estimates for the elastic moduli of martensite and austenite are typically in the range of
, , . However, most experimental data
appears to be significantly different than these. It is, therefore, suggested that estimates of these moduli should be made
using actual experimental data for the materials being calibrated. Initial loading from a state corresponding to 100%
austenite produces a linear elastic response from which can be readily estimated as in Figure 7-25. The modulus
of martensite ( ) is also estimated for the unloading line, again as illustrated in Figure 7-25. In this figure, the
loading should be performed to produce 100% martensite and thus the unloading occurs with the elastic response of
martensite.
Figure 7-25 Typical Stress-strain Curves in the Pseudo-elastic Regime, Depicting the Elastic Moduli
Thermal expansion coefficients ( )
A recommended method for measuring thermal expansion coefficients is through the use of dilatometry whereby
carefully controlled cycles of temperature can be made. An alternative to this type of precise calibration, is to use
literature values that have been shown to be consistent with values measured on actual specimens.
These values are given as follows (See for example, TiNi Smart Sheet)
; and
Detwinning and the calibration of the g function
The phenomenology of the NiTi phase transformation is such that the alignment of the martensite varies with the
prevailing deviatoric stress. This intensity is measured via the von Mises equivalent stress, . As shown in
Equations 7-93 and 7-94, the scaling function that provides the description of the degree to which the martensite is
aligned is . The most direct path to calibrating this function is to fit it to the uniaxial stress vs. strain curve
for pure, randomly oriented martensite conducted at a temperature below the temperature. Such a curve is shown
E
M
E
A
v
M
, , v
A
E
M
28000-41000 Mpa = E
A
60000-83000Mpa = v
M
v
A
0.33 = =
E
A
E
M
Strain
S
t
r
e
s
s
Typical pseudoelastic
T = T
1
C
response
E
A
E
M
A
,
M
6.6 10
6
C 3.67 10
6
F = =
A
11.0 10
6
C 6.11 10
6
F = =
g
eq
( )
eq
g
eq
( ) g
M
f
0

as Figure 7-26. The solid curve shown in Figure 7-26 is the actual measured record of uniaxial stress vs. total strain
for a specimen of 100% martensite tested at a temperature sufficiently low to ensure it remains 100% martensite. The
dot curve is simply a convenient fit to it. The parameter is the equivalent deviatoric transformation strain. Note
that the function is defined as it relates to the development of deviatoric strain due to the alignment of martensite
variants. As mentioned in the previous section, in general, the variables , , ,
, and yield a good match to many experimental results. It is often observed
that there exists a threshold equivalent stress level below which detwinning does not occur; this stress is referred to as
. The value of at this stress is . Note that from Equation (7-94), twinning strain is zero when
. In addition, it is also found, in practice, that the function tends to approach unity at a finite equivalent
stress level, called . By definition, . Also, should be chosen to match the general shape of the
function. Since the ratio is less than one, the higher powers take effect later, and thus can be added to
lower the middle slope of the curve and to fix the final slope of the curve. However, depending on the relative
values of ,

and , this curve might reach a maximum in the range of interest, and therefore, it should be cut-
off at its maximum value . The value of which is reach at a stress value are also
supplied as an input to Marc. Usually is a good approximation. But, the selection of
depends on the experimental measurement.
Figure 7-26 Typical Stress-Strain Curve of 100% Martensite Tested Below Temperature
Others
Yield stresses of the pure martensite and austenite phases for NiTi: and .
eq
T
g
g
a
0 < g
b
2f 2.0 = = g
c
0
g
d
2.25 and 2.75 = g
e
0 g
f
3.0 =
ef f
g
g g
ef f
g
f f
g
( ) =
eq

ef f
g
< g
0
g
g
0
g
( ) 1 = g
a
eq
g
0
g
c
g
e
g
a
g
b
g
c
g
max
g
max

eq
g
0

max
g
=
g
0
2
ef f
g
10
eff
g
= g
0
Strain
S
t
r
e
s
s
TWI N
g
eq
( )
eq
T
=
g
0
g
0
Experimental
Model
M
f
0
Y
M
Y
A
Y
M
70-140 Mpa =
Y
A
195 690 Mpa =
413
CHAPTER 7
Material Library
The calibration of the transformation strains for NiTi:
Mechanical Shape Memory Model
Shape-memory alloys can undergo reversible changes in the crystallographic symmetry-point-group. Such changes
can be interpreted as martensitic phase transformations, that is, as solid-solid diffusion-less phase transformations
between a crystallographically more-ordered phase (the austenite or parent phase) and a crystallographically less-
ordered phase (the martensite). Typically, the austenite is stable at high temperatures and high values of the stress. For
a stress-free state, we indicate with the temperature above which only the austenite is stable and with the temperature
below which only the martensite is stable.
The phase transformations between austenite and martensite are the key to explain the superelasticity effect. For the
simple case of uniaxial tensile stress, a brief explanation follows (Figure 7-27). Consider a specimen in the austenitic
state and at a temperature greater than A; accordingly, only the austenite is stable at zero stress. If the specimen is
loaded, while keeping the temperature constant, the material presents a nonlinear behavior (ABC) due to stress-
induced conversion of austenite into martensite. Upon unloading, while again keeping the temperature constant, a
reverse transformation from martensite to austenite occurs (CDA) as a result of the instability of the martensite at zero
stress. At the end of the loading/unloading process, no permanent strains are present and the stress-strain path is a
closed hysteresis loop.
Figure 7-27 Superelasticity
At the crystallographic level, if there is no preferred direction for the occurrence of the transformation, the martensite
takes advantage of the existence of different possible habit plates (the contact plane between the austenite and the
martensite during a single-crystal transformation), forming a series of crystallographically equivalent variants. The
product phase is then termed multiple-variant martensite and it is characterized by a twinned structure. However, if
(deviatoric transformation strain): 0.05-0.085
(volumetric transformation strain): 0 0.003.
(detwinning stress): 100-150 Mpa
Note:
The current model uses a nonsymmetric Jacobian matrix. It is recommended that the nonsymmetric solver
be used to improve convergence.
eq
T
v
T
ef f
g
g
0
2
ef f
g
10
ef f
g
A
B
C

there is a preferred direction for the occurrence of the transformation (often associated with a state of stress), all the
martensite crystals trend to be formed on the most favorable habit plane. The product is then termed single-variant
martensite and is characterized by a detwinned structure. According to the existence of different types of single-variant
martensite species, the conversion of each single-variant martensite into different single variants is possible. Such a
process, known as a reorientation process, can be interpreted as a family of martensite phase transformations and is
associated with changes in the parameters governing the single-variant martensite production (hence, it is often
associated to nonproportional change of stress).
In addition to the thermo-mechanical shape memory model, a superelastic shape memory alloy model is also
implemented in Marc based on the work of Auricchio [Ref. 1] and [Ref. 2]. This work has been enhanced to allow
different elastic properties for the Austenite and Martensite phases.
The superelastic shape memory model has been implemented in Marc in the framework of multiplicative
decomposition. We assume the deformation gradient, as the control variable, and the martensite fraction, as the
only scalar internal variable. We also introduce a multiplicative decomposition of in the form:
where is the elastic part and is the phase transition part.
Assuming an isotropic elastic response, the Kirchhoff stress and the elastic left Cauchy-Green tensor , defined
as: , share the same principal directions. Accordingly, the following spectral decompositions can be
introduced:
with the elastic principal stretches and the deviatoric part, according to the relation:
(7-95)
where is the second-order identity tensor, is the pressure, defined as , and is the trace
operator. We can write Equation (7-95) with the following component form:
(7-96)
F
S
F
F F
e
F
t r
=
F
e
F
t r
b
e
b
e
F
e
F
eT
=

A
n
A
n
B
A 1 =
3
=
d
A
d
n
A
n
B
A 1 =
3
=
b
e
A
e
( )
2
n
A
n
B
A 1 =
3
A
e
d
pI
d
+ =
I p p tr ( ) 3 = t r
.
( )
A
p
A
d
+ =
415
CHAPTER 7
Material Library
with
, .
Phase Transformations and Activation Conditions
We consider two phase transformations: the conversion of austenite into martensite (A S) and the conversion of
martensite into austenite (S A). To model the possible phase-transformation pressure-dependence, we introduce a
Drucker-Prager-type loading function:
(7-97)
where is a material parameter and indicates the Euclidean norm, such that:
.
Indicating variants in time with a superposed dot, we assume the following linear forms for the evolution of :
for (A S) (7-98)
for (S A) (7-99)
where
,
with , , , and material constants. The scalar quantities and embed the plastic-
transformation activation condition hence, allowing a choice between Equations (7-98) and (7-99) and they are
defined by the relations:
, if , or . Otherwise, .
, if , or . Otherwise, .
where
, .
p K
e
=
A
d
2Ge
A
e
=
F ( )
d
3p + =
A
d
( )
2
A 1 =
3
1 2
=
S H
AS
1
S
( )
F
F R
f
AS
--------------------- =
S H
SA
S
F
F R
f
SA
--------------------- =
R
f
AS
f
AS
2
3
--- +
\ .
| |
= R
f
SA
f
SA
2
3
--- +
\ .
| |
=
s
AS
f
AS
s
SA
f
SA
H
AS
H
SA
H
AS
1 = R
s
AS
F R
f
AS
< < F
0 > H
AS
0 =
H
SA
1 = R
f
SA
F R
s
SA
< < F
0 < H
SA
0 =
R
s
AS
s
AS
2
3
--- +
\ .
| |
= R
s
SA
s
SA
2
3
--- +
\ .
| |
=

Time-discrete Model
The time-discrete model is obtained by integrating the time-continuous model over the time interval [ ]. In
particular, we use a backward-Euler integration formula for the rate-equations evaluating all the nonrate equations at
time . Written in residual form and clearing fraction from Equations (7-98) and (7-99), the time-discrete evolutionary
equations specialize to:
(7-100)
(7-101)
where
(7-102)
The quantity in Equation (7-102) can be computed expressing as a function of and requiring the satisfaction
of the discrete equation relative to the corresponding active phase transition.
The detailed solution algorithm for stress update and consistent tangent modulus are given in the work of Auricchio
[Ref. 2].
In the enhanced version of this model, the user enters different elastic constants for the two phases. In this case, the
effective elastic moduli are taken as:
Experimental Data Fitting for Mechanical Shape Memory Alloy
The experiment for mechanical shape memory alloy is quite simple.
1. To determine the transformation stresses ( ):
Initial Stress for Austenite to Martensite
Final Stress for Austenite to Martensite
Initial Stress for Martensite to Austenite
Initial Stress for Martensite to Austenite
t
n
t ,
t
R
AS
F R
f
AS
( )
s
H
AS
1
S
( ) F F
n
( ) 0 = =
R
SA
F R
f
SA
( )
s
H
SA
S
F F
n
( ) 0 = =
S
t d
t
n
t
}

S

S n ,
= =
S
F
S
E E
A
1
S
( ) E
M
S
+ =

A
1
S
( )
M
S
+ =
S
AS
f
AS
S
SA
f
SA
, , ,
S
AS
f
AS
S
SA
f
SA
417
CHAPTER 7
Material Library
Uniaxial tension test is performed at the same temperature at which the simulation is desired. Here is one
example set for SMA materials.
, , ,
2. :It is measured from the difference between the response in tension and compression.
3. : epsL is a scalar parameter representing the maximum deformation obtainable only by detwinning of the
multiple-variant martensite (or maximum strain obtainable by variant orientation). Classical values for
epsL are in the range 0.005 and 0.10. Marc sets the default value as 0.07.
Conversion from Thermo-Mechanical to Mechanical SMA
Case 1) if the behavior in tension and compression are the same, the value is set to 0.
Case 2) if the behavior in tension and compression have a difference as in the classical case of SMA, the
value is usually set to 0.1 if there is no compression data for the phase transformation. One value
for the phase transformation in compression, say (sigAS_s_compression) is available, is
calculated as follows:
Note:
The mechanical shape memory model only supports ndi = 3 case (3-D, plane-strain and
axisymmetric elements). It does not support either ndi = 1 or ndi = 2 cases (1-D and plane-stress elements).
Table 7-2 Conversion Table
Thermo-Mechanical SMA Mechanical SMA Enhanced Mechanical SMA
S
AS
500Mpa =
f
AS
600Mpa =
S
AS
300Mpa =
f
SA
200Mpa =
s
AS
sqrt 2 3 ( )
S
AS ,
S
AS
( )
S
AS ,
S
AS
+ ( ) =
L
E
A
E 0.5 E
A
E
M
+ ( ) =
E
A
A
0.5
A
M
+ ( ) =

A
E
M
E 0.5 E
A
E
M
+ ( ) =
E
M
M
0.5
A
M
+ ( ) =

M
eq
T
L
sqrt
3
2
---
\ .
| |
eq
T
=
L
sqrt
3
2
---
\ .
| |
eq
T
=
C
M
C
M
C
M
C
A
C
A
C
A
T
o
T
o
T
o

Elastomer
An elastomer is a polymer which shows nonlinear elastic stress-strain behavior. The term elastomer is often used to
refer to materials which show a rubber-like behavior, even though no rubbers exist which show a purely elastic
behavior. Depending upon the type of rubber, elastomers show a more or less strongly pronounced viscoelastic
behavior. Marc considers both the viscous effects and the elastic aspects of the materials behavior. These materials are
characterized by their elastic strain energy function.
Elastomeric materials are elastic in the classical sense. Upon unloading, the stress-strain curve is retraced and there is
no permanent deformation. Elastomeric materials are initially isotropic. Figure 7-28 shows a typical stress-strain curve
for an elastomeric material.
Table 7-3 Prediction from Linear Algebra
The Relationship between Mechanical Model
and Thermo-Mechanical Model
s
AS
T
o
M
s
0
( )C
m
=
f
AS
T
o
M
f
0
( )C
m
=
s
SA
T
o
A
s
0
( )C
a
=
f
SA
T
o
A
f
0
( )C
a
=
T
T
o
A
f
M
s
A
s
M
f
C
M
C
A
f
SA
s
SA
s
AS
f
AS
419
CHAPTER 7
Material Library
Figure 7-28 A Typical Stress-Strain Curve for an Elastomeric Material
Calculations of stresses in an elastomeric material requires an existence of a strain energy function which is usually
defined in terms of invariants or stretch ratios. Significance and calculation of these kinematic quantities is discussed
next.
In the rectangular block in Figure 7-29, , , and are the principal stretch ratios along the edges of the block
defined by
(7-103)
Figure 7-29 Rectangular Rubber Block
In practice, the material behavior is (approximately) incompressible, leading to the constraint equation
(7-104)
,

S
t
r
e
s
s
e, Strain
100%
1

2

3
i
L
i
u
i
+ ( ) L
i
=
L
3
2
L
2
L
2
L
1
1
L
1
Undeformed
Deformed
3
L
3
3
1 =

the strain invariants are defined as
(7-105)
Depending on the choice of configurations, for example, reference (at ) or current ( ), you obtain
total or updated Lagrange formulations for elasticity. The kinematic measures for the two formulations are discussed
next.
A. Total Lagrangian Formulation
The strain measure is the Green-Lagrange strain defined as:
(7-106)
where is the right Cauchy-Green deformation tensor defined as:
(7-107)
in which is the deformation gradient (a two-point tensor) written as:
(7-108)
The Jacobian is defined as:
(7-109)
Thus, the invariants can be written as:
(7-110)
in which is the permutation tensor. Also, using spectral decomposition theorem,
(7-111)
I
1

1
2
2
2
3
2
+ + =
I
2

1
2
2
2
2
2
3
2
3
2
1
2
+ + =
I
3

1
2
2
2
3
2
=
t 0 = t n 1 + =
E
i j
1
2
--- C
i j

i j
( ) =
C
i j
C
i j
F
ki
F
kj
=
F
kj
F
kj
x
k
X
j
--------- =
J
J
1
3
det C
i j
( )
1
2
---
= =
I
1
C
i i
=
I
2
C
i j
C
i j
C
i i
( )
2
( )
2
------------------------------------------ =
I
3
1
6
--- e
i j k
e
pqr
C
i p
C
j q
C
kr
det C
i j
( ) = =
(implied sum on i)
e
i j k
C
i j

A
2
N
i
A
N
j
A
=
421
CHAPTER 7
Material Library
7
Material Library
in which the stretches are the eigenvalues of the right Cauchy-Green deformation tensor, and the
eigenvectors are .
B. Updated Lagrange Formulation
The strain measure is the true or logarithmic measure defined as:
(7-112)
where the left Cauchy-Green or finger tensor is defined as:
(7-113)
Thus, using the spectral decomposition theorem, the true strains are written as:
(7-114)
where is the eigenvectors in the current configuration. It is noted that the true strains can also be
approximated using first Pad approximation, which is a rational expansion of the tensor, as:
(7-115)
where a polar decomposition of the deformation gradient is done into the left stretch tensor and
rotation tensor as:
(7-116)
The Jacobian is defined as:
(7-117)
and the invariants are now defined as:
(7-118)
A
2
C
i j
N
i
A
i j
1
2
--- ln b
i j
=
b
i j
b
i j
F
i k
F
j k
=
i j

A
( ) ln
A 1 =
3
n
i
A
n
j
A
=
n
i
A
i j
2 V
i j

i j
( ) V
i j

i j
+ ( )
1
=
F
i j
V
i j
R
i j
F
i j
V
i k
R
kj
=
J
J
1
3
det b
i j
( )
1
2
---
= =
I
1
b
i i
=
I
2
1
2
--- b
i j
b
i j
b
i i
( )
2
( ) =
I
3
1
6
--- e
i j k
e
pqr
b
i p
b
j q
b
kr
det b
i j
( ) = = and

It is noted that either Equation (7-110) or Equation (7-118) gives the same strain energy since it is scalar and
invariant. Also, to account for the incompressibility condition, in both formulations, the strain energy is split
into deviatoric and volumetric parts as:
(7-119)
Thus, the generalized Mooney-Rivlin (gmr) and the Ogden models for nearly-incompressible elastomeric
materials are written as:
(7-120)
where and are the first and second deviatoric invariants.
A particular form of the generalized Mooney-Rivlin model, namely the third order deformation (tod) model, is
implemented in Marc.
(7-121)
where
Simpler and popular forms of the above strain energy function are obtained as:
(7-122)
Use the MOONEY model definition option to activate the elastomeric material option in Marc and enter the material
constants .
The TEMPERATURE EFFECTS model definition option can be used to input the temperature dependency of the
constants and . The UMOONY user subroutine can be used to modify all five constants , , , ,
and . Using the table driven input, all material parameters can reference tables to define temperature dependent
behavior. For viscoelastic, the additional VISCELMOON model definition option must be included.
is the deviatoric third order deformation form strain energy function,
are material constants obtained from experimental data
W W
deviatoric
W
volumetric
+ =
W
devi at or i c
gmr
C
mn
I
1
3 ( )
m
I
2
3 ( )
n
n 1 =
N
m 1 =
N
=
I
1
I
2
W
devratoric
tod
C
10
I
1
3 ( ) C
01
I
2
3 ( ) C
11
I
1
3 ( ) I
2
3 ( ) C
20
I
1
3 ( )
2
+ + + + =
C
30
I
1
3 ( )
3
W
deviatoric
tod
C
10
C
01
C
11
C
20
C
30
, , , ,
W
deviatoric
mr
C
10
I
1
3 ( ) C
01
I
2
3 ( ) + = Mooney-Rivlin
W
deviatoric
nh
C
10
I
1
3 ( ) = Neo-Hookean
C
10
C
01
C
11
C
20
C
30
, , , ,
C
10
C
01
C
01
C
10
C
11
C
20
C
30
423
CHAPTER 7
Material Library
The form of strain energy for the Ogden model in Marc is,
(7-123)
where are the deviatoric stretch ratios while , , and are the material constants obtained
from the curve fitting of experimental data. This capability is available in Mentat.
If no bulk modulus is given, it is taken to be virtually incompressible. This model is different from the Mooney model
in several respects. The Mooney material model is with respect to the invariants of the right or left Cauchy-Green strain
tensor and implicitly assumes that the material is incompressible. The Ogden formulation is with respect to the
eigenvalues of the right or left Cauchy-Green strain, and the presence of the bulk modulus implies some
compressibility. Using a two-term series results in identical behavior as the Mooney mode if:
and and and
The material data is given through the OGDEN model definition option or the UOGDEN user subroutine. For
viscoelastic behavior, the additional VISCELOGDEN model definition option must be included.
In the Arruda-Boyce strain energy model, the underlying molecular structure of elastomer is represented by an eight-
chain model to simulate the non-Gaussian behavior of individual chains in the network. The two parameters,
and ( is the chain density, is the Boltzmann constant, is the temperature, and is the number of statistical
links of length l in the chain between chemical crosslinks) representing initial modules and limiting chain extensibility
and are related to the molecular chain orientation thus representing the physics of network deformation.
As evident in most models describing rubber deformation, the strain energy function constructed by fitting experiment
data obtained from one state of deformation to another fails to accurately describe that deformation mode. The Arruda-
Boyce model ameliorates this defect and is unique since the standard tensile test data provides sufficient accuracy for
multiple modes of deformation.
The model is constructed using the eight chain network as follows [Ref. 3]:
Consider a cube of dimension with an unstretched network including eight chains of length , where
the fully extended chain has an approximate length of Nl. A chain vector from the center of the cube to a corner can
be expressed as:
(7-124)
Using geometrical considerations, the chain vector length can be written as:
(7-125)
W
devi at or i c
ogden

k
------
1
k
3 + +
\ .
| |
k 1 =
N
k
J

k
3
-------
k
= C
mn

k

k
1
2C
10
=
1
2 =
2
2C
01
=
2
2 =
nk
N n k N
0
r
0
Nl =
C
1
0
2
------
1
i

0
2
------
2
j

0
2
------
3
k + + =
r
chain
1
3
------- Nl
1
2
2
2
3
2
+ + ( )
1 2
=

and
(7-126)
Figure 7-30 Eight Chain Network in Stretched Configuration
Using statistical mechanics considerations, the work of deformation is proportional to the entropy change on stretching
the chains from the unstretched state and may be written in terms of the chain length as:
(7-127)
where is the chain density and is a constant. is an inverse Langevin function correctly accounts for the limiting
chain extensibility and is defined as:
(7-128)
where Langevin is defined as:
(7-129)
With Equations (7-126) through (7-129), the Arruda-Boyce model can be written
(7-130)
chain
r
chain
r
0
------------
1
3
------- I
1
( )
1 2
= =
0
C
1
k
j
i
W nkN
r
chain
Nl
------------
\
|
ln

sinh
--------------
.
|
C
+ =
n C
L
1
r
chain
Nl
------------
\ .
| |
=
( )
1
--- coth =
W
dev
Arruda-Boyce
nk
1
2
--- I
1
3 ( )
1
20N
---------- I
1
2
9 ( )
11
1050N
2
------------------- I
1
3
27 ( ) + + =
19
7000N
3
------------------- I
1
4
81 ( )
519
673750N
4
------------------------- I
1
5
243 ( ) ] + +
425
CHAPTER 7
Material Library
Also, using the notion of limiting chain extensibility, Gent [Ref. 5] proposed the following constitutive relation:
(7-131)
The constant is independent of molecular length and, hence, of degree of cross linking. The model is attractive
due to its simplicity, but yet captures the main behavior of a network of extensible molecules over the entire range of
possible strains.
The Arruda-Boyce and Gent model can be invoked by using the ARRUDBOYCE and GENT model definition options,
respectively.
The volumetric part of the strain energy is for all the rubber models in Marc:
(7-132)
when is the bulk modulus.
It can be noted that the particular form of volumetric strain energy is chosen such that:
1. The constraint condition is satisfied for incompressible deformations only; for example:
(7-133)
2. The constraint condition does not contribute to the dilatational stiffness.
The pressure is obtained by taking the derivative of the volumetric strain energy function with respect to the
invariant yielding.
(7-134)
The equation has a physical significance in that for small deformations, the pressure is linearly related to the
volumetric strains by the bulk modulus .
The discontinuous or continuous damage models discussed in the models section on damage can be included with the
generalized Mooney-Rivlin, Ogden, Arruda-Boyce, and Gent models to simulate Mullins effect or fatigue of
elastomers when using the updated Lagrangian approach. In the total Lagrangian framework however, this is available
for the Ogden model only.
The rubber foam model which is based on Ogden formulation has a strain energy form as follows:
(7-135)
W
dev
Gent
E
6
--- I
m
3 ( ) 1
I
1
3
I
m
3
---------------
\ .
|
| |
log =
EI
m
U
9
2
--- K J
1 3
1 ( )
2
=
K
f I
3
( )
> 0 if I
3
0 >
0 = if I
3
1 =
< 0 if I
3
0 <
p 3K J 1 ( ) =
K
W

n
n
------
1
n
3 + + ( )

n
n
------ 1 J
\ .
| |
n 1 =
N
+
n 1 =
N
=

where , , are material constants. When all terms are equal to zero, the second part of in Equation
(7-135) is omitted.
You can define any other invariant based models through the use of the UENERG user subroutine when using the
MOONEY option in model definition for elasticity in total or Updated Lagrangian framework. A more general and
easy-to-use the UELASTOMER user subroutine (uelastomer.f) can be used to define a general strain energy
function in the Updated Lagrangian framework.
Once the strain energy function is defined, the stresses and material tangent can be evaluated for the total and Updated
Lagrangian formulations as:
A. Total Lagrangian Formulation:
The stress measure in the total Lagrangian formulation is the symmetric second Piola-Kirchhoff stress ,
calculated as:
(7-136)
The material elasticity tangent is:
(7-137)
B. Updated Lagrangian Formulation:
The stress measure in the Updated Lagrangian formulation is the Cauchy or true stress calculated as:
(7-138)
The spatial elasticity tangent is:
(7-139)
The material constants for the Mooney-Rivlin form can be obtained from experimental data. The Mooney-Rivlin form
of the strain energy density function is
(7-140)
For the Mooney-Rivlin model, the force and deformation for a uniaxial test specimen can be related as
(7-141)
which can be written in the form:
(7-142)
n

n

n

n
W
S
i j
S
i j
W
E
i j
---------- 2
W
C
i j
---------- = =
D
i j kl
2
W
E
i j
E
kl
---------------------- 4

2
W
C
i j
C
kl
----------------------- = =
i j
2
J
---
W
b
i k
---------- b
kj
=
L
i j kl
4
J
--- b
i m

2
W
b
mj
b
kn
------------------------ b
nl
=
W
deviatoric
mooney-rivlin
C
10
I
1
3 ( ) C
01
I
2
3 ( ) + =
P 2A
0
1
1
1
3
------
\ .
| |

1
C
10
C
01
+ ( ) =
P
2A
0

1
1
1
2
------
\ .
| |
------------------------------------ C
10
C
01
1
--------- + =
427
CHAPTER 7
Material Library
where is the force of the specimen, is the original area of the specimen, and is the uniaxial stretch ratio. This
equation provides a simple way to determine the Mooney-Rivlin constants. The Mooney-Rivlin constitutive equation
is applicable if the plot of versus should yield a straight line of slope and intercept
( ) on the vertical axis as shown in Figure 7-31.
Figure 7-31 Plots for Various Rubbers in Simple Extension for Mooney-Rivlin Model
If only the Youngs modulus is supplied, and full uniaxial data are not available then
(7-143)
is a reasonable assumption. The constants then follow from the relation:
(7-144)
The material coefficients for the models can be obtained from Mentat. This allows you to select which model is most
appropriate for your data.
Updated Lagrange Formulation for Nonlinear Elasticity
The total Lagrange nonlinear elasticity models in Marc have been augmented with a formulation in an Updated
Lagrange framework. Hence, Rezoning can be used for elastomeric materials based upon the current configuration.
This is specially useful in large deformation analysis since typically excessive element distortion in elastomeric
materials can lead to premature termination of analysis. The new formulation accommodates the generalized Mooney-
P A
0

1
P 2A
0

1
1
1
2
------
\ .
| |
1
----- C
01
C
01
C
10
+
1
1
----- 1 =
0.5
0.1
0.6 0.7 0.8 0.9 1.0
1/
B
0.2
0.3
0.4
/
2
(
-
1
/
2
)

(
N
/
m
n
2
)
A
C
D
E
F
G
E
C
01
0.25C
10
6 C
10
C
01
+ ( ) E

Rivlin and Ogden material models preserving the same format and strain energy functions as the total Lagrange
formulation. In addition, the Arruda-Boyce and the Gent models are only available in the updated Lagrange
framework.
The updated Lagrangian rubber elasticity capability can be used in conjunction with both continuous as well as
discontinuous damage models. Thermal, as well as viscoelastic, effects can be modeled with the current formulation.
Using the table driven input, all of the material properties can reference tables, which facilitates the use of temperature
dependent material properties. The singularity ratio of the system is inversely proportional to the order of bulk
modulus of the material due to the condensation procedure.
A consistent linearization has been carried out to obtain the tangent modulus. The singularity for the case of two- or
three-equal stretch ratios is analytically removed by application of LHospitals rule. The current framework with an
exact implementation of the finite strain kinematics along with the split of strain energy to handle compressible and
nearly incompressible response is eminently suitable for implementation of any nonlinear elastic as well as inelastic
material models. In fact, the finite deformation plasticity model based on the multiplicative decomposition,
is implemented in the same framework.
To simulate elastomeric materials, incompressible element(s) are used for plane strain, axisymmetric, and three-
dimensional problems for elasticity in total Lagrangian framework. These elements can be used with each other or in
combination with other elements in the library. For plane stress, beam, plate or shell analysis, conventional elements
can be used. For updated Lagrangian elasticity, both conventional elements (as well as Herrmann elements) can be
used for plane strain, axisymmetric, and three-dimensional problems.
Time-independent Inelastic Behavior
In uniaxial tension tests of most metals (and many other materials), the following phenomena can be observed. If the
stress in the specimen is below the yield stress of the material, the material behaves elastically and the stress in the
specimen is proportional to the strain. If the stress in the specimen is greater than the yield stress, the material no longer
exhibits elastic behavior, and the stress-strain relationship becomes nonlinear. Figure 7-32 shows a typical uniaxial
stress-strain curve. Both the elastic and inelastic regions are indicated.
Figure 7-32 Typical Uniaxial Stress-Strain Curve (Uniaxial Test)
F F
e
F
F
p
=
Note: Stress and strain are total quantities.
Stress Inelastic
Region
Elastic Region
Yield
Stress
Strain
429
CHAPTER 7
Material Library
Within the elastic region, the stress-strain relationship is unique. As illustrated in Figure 7-33, if the stress in the
specimen is increased (loading) from zero (point 0) to (point 1), and then decreased (unloading) to zero, the strain
in the specimen is also increased from zero to , and then returned to zero. The elastic strain is completely recovered
upon the release of stress in the specimen.
The loading-unloading situation in the inelastic region is different from the elastic behavior. If the specimen is loaded
beyond yield to point 2, where the stress in the specimen is and the total strain is , upon release of the stress in
the specimen the elastic strain, , is completely recovered. However, the inelastic (plastic) strain, , remains in the
specimen. Figure 7-33 illustrates this relationship.

Figure 7-33 Schematic of Simple Loading - Unloading (Uniaxial Test)
Similarly, if the specimen is loaded to point 3 and then unloaded to zero stress state, the plastic strain remains in
the specimen. It is obvious that is not equal to . We can conclude that in the inelastic region:
Plastic strain permanently remains in the specimen upon removal of stress.
The amount of plastic strain remaining in the specimen is dependent upon the stress level at which the
unloading starts (path-dependent behavior).
The uniaxial stress-strain curve is usually plotted for total quantities (total stress versus total strain). The total stress-
strain curve shown in Figure 7-32 can be replotted as a total stress versus plastic strain curve, as shown in Figure 7-34.
The slope of the total stress versus plastic strain curve is defined as the workhardening slope (H) of the material. The
workhardening slope is a function of plastic strain.
2

2
2
e
2
p
Stress
Total Strain = Strain and Elastic Strain
Yield Stress
Strain
3
p
2

2
p
2
e
+ =
3

3
p
3
e
+ =
3
p
3
e
2
e
2
p
0
1
2
3
3
p
2
p
3
p

Figure 7-34 Definition of Workhardening Slope (Uniaxial Test)
The stress-strain curve shown in Figure 7-32 is directly plotted from experimental data. It can be simplified for the
purpose of numerical modeling. A few simplifications are shown in Figure 7-35 and are listed below:
1. Bilinear representation constant workhardening slope
2. Elastic perfectly-plastic material no workhardening
3. Perfectly-plastic material no workhardening and no elastic response
4. Piecewise linear representation multiple constant workhardening slopes
5. Strain-softening material negative workhardening slope
In addition to elastic material constants (Youngs modulus and Poissons ratio), it is essential to include yield stress
and workhardening slopes when dealing with inelastic (plastic) material behavior. These quantities can vary with
parameters such as temperature and strain rate. Since the yield stress is generally measured from uniaxial tests, and
the stresses in real structures are usually multiaxial, the yield condition of a multiaxial stress state must be considered.
The conditions of subsequent yield (workhardening rules) must also be studied.
Figure 7-35 Simplified Stress-Strain Curves (Uniaxial Test)
Plastic Strain
H = tan (Workhardening Slope)
= d/d
p
Total Stress
(1) Bilinear Representation
(3) Perfectly Plastic
(5) Strain Softening
(2) Elastic-Perfectly Plastic
(4) Piecewise Linear Representation

431
CHAPTER 7
Material Library
Yield Conditions
The yield stress of a material is a measured stress level that separates the elastic and inelastic behavior of the material.
The magnitude of the yield stress is generally obtained from a uniaxial test. However, the stresses in a structure are
usually multiaxial. A measurement of yielding for the multiaxial state of stress is called the yield condition. Depending
on how the multiaxial state of stress is represented, there can be many forms of yield conditions. For example, the yield
condition can be dependent on all stress components, on shear components only, or on hydrostatic stress. A number
of yield conditions are available in Marc, and are discussed in this section.
von Mises Yield Condition
Although many forms of yield conditions are available, the von Mises criterion is the most widely used. The success
of the von Mises criterion is due to the continuous nature of the function that defines this criterion and its agreement
with observed behavior for the commonly encountered ductile materials. The von Mises criterion states that yield
occurs when the effective (or equivalent) stress () equals the yield stress (
y
) as measured in a uniaxial test.
Figure 7-36 shows the von Mises yield surface in two-dimensional and three-dimensional stress space.
For an isotropic material:
(7-145)
where
1
,
2
, and
3
are the principal Cauchy stresses.
Figure 7-36 von Mises Yield Surface
can also be expressed in terms of nonprincipal Cauchy stresses.
(7-146)
The yield condition can also be expressed in terms of the deviatoric stresses as:
(7-147)

1

2
( )
2
2

3
( )
2
3

1
( )
2
+ + [ ]
1 2
2 =
Yield
Surface
Elastic
Region
(b) p-Plane
Yield
Surface
Elastic
Region
(a) Two-dimensional Stress Space
3
d
2
d
1
d
2

x

y
( )
2
y

z
( )
2
z

x
( )
2
6
xy
2
yz
2
zx
2
+ + ( ) + + + [ ]
1 2
= ( ) 2
3
2
---
i j
d
i j
d
=

where is the deviatoric Cauchy stress expressed as
(7-148)
For isotropic material, the von Mises yield condition is the default condition in Marc. The initial yield stress is
defined in the ISOTROPIC and ORTHOTROPIC options. A user-defined finite strain, isotropic plasticity material
model can be implemented through the UFINITE user subroutine. In this case, the finite strain kinematics is taken care
of in Marc. You have to do the small strain return mapping only. See Marc Volume D: User Subroutines and Special
Routines for more details.
Hills [1948] Yield Function
The anisotropic yield function (of Hill) and stress potential are assumed as
(7-149)
where is the equivalent tensile yield stress for isotropic behavior.
Ratios of actual to isotropic yield (in the preferred orientation) are defined in the array YRDIR for direct tension
yielding, and in YRSHR for yield in a shear (the ratio of actual shear yield to isotropic shear yield). Then the
above are defined by:
(7-150)
(7-151)
(7-152)
(7-153)
(7-154)
(7-155)
i j
d
i j
d
i j
1
3
---
kk
i j
=
y
a [
1

y

z
( )
2
a
2

z

x
( )
2
a
3

x

y
( )
2
3a
4
zx
2
+ + + + =
3a
5
yz
2
3a
6
xy
2
]
1 2
2 + +
3
a
1
a
1
1
YRDIR 2 ( )**2
-------------------------------------- -
1
YRDIR 3 ( )**2
-------------------------------------- -
1
YRDIR 1 ( )**2
-------------------------------------- - + =
a
2
1
YRDIR 3 ( )**2
-------------------------------------- -
1
YRDIR 1 ( )**2
-------------------------------------- -
1
YRDIR 2 ( )**2
-------------------------------------- - + =
a
3
1
YRDIR 1 ( )**2
-------------------------------------- -
1
YRDIR 2 ( )**2
--------------------------------------
1
YRDIR 3 ( )**2
-------------------------------------- + =
a
4
2
YRSHR 3 ( )**2
---------------------------------------- =
a
5
2
YRSHR 2 ( )**2
--------------------------------------- =
a
6
2
YRSHR 1 ( )**2
---------------------------------------- =
433
CHAPTER 7
Material Library
For anisotropic material, use the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition options to
indicate the anisotropy. Use the ORTHOTROPIC option or the ANPLAS user subroutine for the specification of
anisotropic yield condition (constants a
1
through a
6
, as defined above), and the ORIENTATION model definition option
or the ORIENT user subroutine, if necessary, to specify preferred orientations.
Hills (1948) Yield Criterion has been extensively used in sheet metal forming, especially for steel. The experimental
data can be related to the Marc input for the Hills Yield Criterion for the shells or plane stress case as given below.
Figure 7-37 Axes of Anisotropy
The sample of tensile coupon cut from a sheet in the three directions, (rolling), 45 and 90 (transverse) is
tested to obtain , respectively. Similarly, the anisotropy parameter defined as:
(7-156)
is obtained for the 0, 45, and 90 directions. The yield stress in the third (thickness) direction can be written as:
(7-157)
The direct stress coefficients are now:
(7-158)
(7-159)
(7-160)
N (Thickness Direction)
R (Rolling Direction)
T (Transverse Direction)
0 =

0
,
45
, and
90
=
r

width
thickness
-------------------- =
N

0
r
90
1 r
0
+ ( )
r
0
r
90
+
---------------------------
90
r
0
1 r
90
+ ( )
r
0
r
90
+
--------------------------- = =
YRDIR 1 ( )

0
av
-------- =
YRDIR 2 ( )

90
av
-------- =
YRDIR 3 ( )

N
av
-------- =

where is the initial yield stress on the stress-strain curve used. If the stress-strain curve is averaged from all
directions, is defined as in orthotropic plasticity. Similarly, the shear coefficients can
be evaluated at:
(7-161)
(7-162)
It is noted that the transverse direction through the thickness is usually considered isotropic which is reasonable on
physical grounds. Also, notice that for complete isotropy, and , which
yields the von Mises yield criterion.
Note that the use of the Hill Yield Function in Marc with all coefficients set to unity will give results varying from the
von Mises case due to differences in return mapping in both cases. By default, the von Mises case uses the mean
normal return scheme whereas the Hill case uses the multistage radial return scheme. One can refer to Chapter 5:
Structural Procedure Library for more details on these return mapping methods. The Hill results with unit ratio that
can be reproduced by using the von Mises yield criterion on conjunction with the Additive Decomposition with
Radial Return option (PLASTICITY,4). In this case, the von Mises yield criterion also uses the multi-stage radial return
mapping procedure instead of the default mean-normal scheme.
Barlats (1991) Yield Function
Barlat et al. [Ref. 6] proposed a general criterion for planar anisotropy that is particularly suitable for aluminum alloy
sheets. This criterion has been shown to be consistent with polycrystal-based yield surfaces which often exhibit small
radii of curvature near uniaxial and balanced biaxial tension stress states. An advantage of this criterion is that its
formulation is relatively simple as compared with the formulation for polycrystalline modeling and, therefore, it can
be easily incorporated into finite element codes for the analysis of metal forming problems.
For three dimensional deformation, the yield function is defined as (Barlat et al. [Ref. 6])
(7-163)
where are principal values of a symmetric matrix defined with respect to the components of the Cauchy
stress as
(7-164)
av
av

av
0
2
45

90
+ +
4
----------------------------------------- =
YRSHR 1 ( ) YRDIR 3 ( )
3
2r
45
1 +
-------------------- =
YRSHR 2 ( ) YRSHR 3 ( ) 1.0 = =
0

45

90
= = r
0
r
45
r
90
1 = = =
f
f S
1
S
2
m
S
2
S
3
m
S
3
S
1
m
+ + 2
m
= =
S
i 1 2 3 , , =
S
i j
S
C
3

xx

yy
( ) C
2

zz

xx
( )
3
-------------------------------------------------------------------------------- C
6
xy
C
5
zx
C
6
xy
C
1

yy

zz
( ) C
3

xx

yy
( )
3
-------------------------------------------------------------------------------- C
4
zy
C
5
zx
C
4
zy
C
2

zz

xx
( ) C
2

yy

zz
( )
3
-------------------------------------------------------------------------------
=
435
CHAPTER 7
Material Library
In Equation (7-164), the symmetry axes (x, y, z), which represent the mutually orthogonal axes of anisotropy, are
aligned with the initial rolling, transverse, and normal directions of the sheet. During deformation, the anisotropic yield
surface of each material element rotates so that the symmetry axes are all in different directions during deformation.
Therefore, it is necessary to trace the rotation of the yield surface during deformation in order to calculate the plastic
strain increment properly. The rotation of the anisotropy axes is carried out based on the polar decomposition method.
The material coefficients, in Equation (7-164) represent anisotropic properties. When , the
material is isotropic and Equation (7-163) reduces to the Tresca yield condition for or , and the von Mises
yield criterion for or 4. The exponent is mainly associated with the crystal structure of the material. A
higher value has the effect of decreasing the radius of curvature of rounded vertices near the uniaxial and
balanced biaxial tension ranges of the yield surface, in agreement with polycrystal models. Values of for FCC
materials (like aluminum) and for BCC materials (like steel) are recommended. The yield surface has been
proven to be convex for . Figure 7-38 shows the yield surfaces obtained from von-Mises, Hill and Barlat yield
functions for Aluminum 2008-T4 alloy.
Figure 7-38 Comparison of Yield Surfaces Obtained from von Mises, Hill and Barlat Yield Functions
Utilizing the normality rule, the associated plastic strain increment is obtained from the yield function as
(7-165)
where is a scalar function. The calculation of in Equation (7-165) is lengthy but straightforward. The stress
integration to obtain along with the calculation of are shown in the works by Chung and Shah, [Ref. 7],
and Yoon et al,.[Ref. 8].
C
i 1 6 =
C
i 1 6 =
1 =
m 1 =
m 2 = m
m
m 8 =
m 6 =
m 1
1.5
1.0
0.5
-0.5
-1.0
0.0
1.5 1.0 0.5 -0.5 -1.0 0.0
Mises
Hill
(1948)
Barlats 6D
yy
xx
i j
p
f
i j
p
i j
---------- =
i j
----------
i j
p f
-------------

In order to obtain four unknown independent coefficients with the assumption of
(isotropic properties for transverse directions), it is necessary to use four stresses from the experimental data
, where are the tensile yield stresses at 0
o
, 45
o,
and 90
o
from the rolling direction, and
is the balanced biaxial yield stress measured from bulge test. A detailed procedure to calculate the coefficients of
Barlats yield function are summarized in the work of Yoon at el., [Ref. 9]. In Marc, Barlats coefficients are
automatically calculated from user inputs for in Mentat. If biaxial data is not available, generally
Marc assumes . Barlats yield function can be accessed from ISOTROPIC, ORTHOTROPIC,
or ANISOTROPIC model definition options and can be used in conjunction with the ORIENTATION option.
Mohr-Coulomb Material (Hydrostatic Stress Dependence)
Marc includes options for elastic-plastic behavior based on a yield surface that exhibits hydrostatic stress dependence.
Such behavior is observed in a wide class of soil and rock-like materials. These materials are generally classified as
Mohr-Coulomb materials (generalized von Mises materials). Ice is also thought to be a Mohr-Coulomb material. The
generalized Mohr-Coulomb model developed by Drucker and Prager is implemented in Marc. There are two types of
Mohr-Coulomb materials: linear and parabolic. Each is discussed on the following pages.
Linear Mohr-Coulomb Material
The deviatoric yield function, as shown in Figure 7-39, is assumed to be a linear function of the hydrostatic stress.
(7-166)
where (7-167)
(7-168)
(7-169)
Figure 7-39 Yield Envelope of Plane Strain (Linear Mohr-Coulomb Material)
C
1
C
2
C
3
C
6
, , , ( ) C
4
C
5
1 = =
0

45

90

b
, , , ( )
0

45

90
, ,
0

45

90

b
, , ,
b

0
=
f J
1
J
2
1 2

3
------- + 0 = =
J
1

i i
=
J
2
1
2
---
i j
d
i j
d
=
Yield Envelope
c
x

y
+
2
-------------------
437
CHAPTER 7
Material Library
Analysis of linear Mohr-Coulomb material based on the constitutive description above is available in Marc through
the ISOTROPIC model definition option. Through the ISOTROPIC option, the values of and are entered. Note
that, throughout the program, the convention that the tensile direct stress is positive is maintained, contrary to its use
in many soil mechanics texts.
The constants and can be related to and by
(7-170)
where is the cohesion and is the angle of friction.
Parabolic Mohr-Coulomb Material
The hydrostatic dependence is generalized to give a yield envelope which is parabolic in the case of plane strain (see
Figure 7-40).
(7-171)
The parabolic yield surface is obtained in Marc through the ISOTROPIC model definition option. Enter the values
and through the ISOTROPIC model definition option.
(7-172)
where is the cohesion.
Figure 7-40 Resultant Yield Condition of Plane Strain (Parabolic Mohr-Coulomb Material)

c
c

3 1 12
2
( ) [ ]
1 2
----------------------------------------------
3
1 3
2
( )
1 2
---------------------------------- ; sin = =
c
f 3J
2
3J
1
+ ( )
1 2
0 = =

3 3c
2
2
( ) ( )
1 2
--------------------------------------------- =
2
3 c
2
2
3
------
\ .
| |
=
c
c
c
2
-----
x

y
+
2
-------------------

Buyukozturk Criterion (Hydrostatic Stress Dependence)
This yield criterion [Ref. 4], which originally has been proposed as a failure criterion, has the general form:
(7-173)
Through the ISOTROPIC model definition option, the user has to define and the factor , where has a fixed value
of . The Buyukozturk criterion reduces to the parabolic Mohr-Coulomb criterion if .
Powder Material
Some materials, during certain stages of manufacturing are granular in nature. In particular, powder metals are often
used in certain forging operations and during hot isostatic pressing (HIP). These material properties are functions of
both the temperature and the densification. It should be noted that the soil model discussed in this chapter also exhibits
some of these characteristics.
There are two models for representing granular materials such as powders. The Shima-Oyane model and the
exponential cap model may be used to represent these types of materials. Both of these models are dependent upon the
hydrostatic stress and the relative density. It is felt that the newer exponential Cap model better represents the physical
phenomena as shown in the figure below.
Figure 7-41 Comparison with Experiments [Ref. 1]
f 3J
1
J
1
2
3J
2

2
+ + =

0.2 0 =
439
CHAPTER 7
Material Library
In the Shima-Oyane model incorporated into Marc, a unified viscoplastic approach is used. The yield function is
(7-174)
where is the uniaxial yield stress, is the deviatoric stress tensor, and is the hydrostatic pressure. , are
material parameters.
can be a function of temperature and relative density, , are functions only of relative density.
Typically, we allow:
(7-175)
where is the relative density.
As the powder becomes more dense, approaches 1 and the classical von Mises model is recovered.
It should be noted that the elastic properties are also functions of relative density. In particular, as the material becomes
fully dense, the Poissons ratio approaches 0.5.
As most processes involving powder materials are both pressure and thermally driven, it can be necessary to perform
a coupled analysis. Marc also allows you to specify density effects for the thermal properties, conductivity and specific
heat. The basic input data is entered through the POWDER option. In addition to the TEMPERATURE EFFECTS
option, there is a DENSITY EFFECTS option. The initial relative density is entered through the RELATIVE DENSITY
option.
Figure 1-42 Yield Function of Shima Model
F
1
---
3
2
---
d
d
p
2
2
------ +
\ .
| |
1 2
y
=
y

d
p
y

q
1
q
2
+
q
3
( )
q
4
=
b
1
b
2
b
3
+ ( )
b
4
=
d
:
d
2
3
---
2
y
2
------------------
p
2
y 2
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-

Exponential Cap Powder Model
The exponential cap powder model implemented into Marc is based upon the Sandia Geomodel [Ref. 2] and [Ref. 3],
a generalized plasticity model, wherein the combined effect of the two underlying mechanisms, porosity and
microcracks, is represented through a unified yield function given as,
where
where are the second and third invariants of the stress tensor respectively and is the strength ratio.
is the stress tensor
is the deviatoric backstress tensor
is a internal state variable determining the position of the cap
is the second invariant of the shifted stress tensor
is the maximum allowable translation of the yield surface when kinematic hardening is enabled
is the Gudehus Lode angle function given as
f , , ( ) J
2
2
F
s
I
1
( ) N [ ]
2
F
c
I
1
, ( ) =
J
2
dev ( ) =
N
J
2
J
3
( )
1
2
--- 1
J
3
2
---- -
3
J
2
-----
\ .
| |
3 2
---- 1
J
3
2
-----
3
J
2
---- -
\ .
| |
3 2
+
\ .
|
| |
+ =
J
2
J
3
,
Direction of
plastic strain
on cap
I
1
X
Strain hardening
cap F
c
Shear Failure
Surface F
s
X
0

0
X
0
Cap Nose
Cap Branch
Point
Point
J
2
F
s
F
c
---------------- =
441
CHAPTER 7
Material Library
Shear Failure Surface
The shear failure surface is represented as,
where, are all non negative material parameters.
Cap Function
The cap function can be written as
where, and is the cap eccentricity parameter.
F
s
I
1
( )
F
s
I
1
( ) A
1
A
3
e
A
2
I
1
A
4
I
1
=
A
1
A
2
A
3
A
4
, , ,
f
s
I
1
A
1
A
3
F
c
I
1
, ( )
F
c
I
1
, ( ) 1 = if I
1
>
1
I
1

X
--------------
\ .
| |
2
=
otherwise
X ( ) RF
s
( ) = R
f
c
1.0
X I
1

The volumetric plastic strain is determined from the expression
where , , , and are all material constants.
Evolution equations
The flow rule,
where is the consistency parameter.
Evolution of is governed by,
where
When kinematic hardening is enabled, the backstress evolves as
where,
Kinematic Hardening:
Kinematic hardening can be introduced by use of a shifted stress tensor
The backstress tensor is initialized to zero. Upon onset of yielding, the backstress evolves in proportion to the
deviatoric part of the plastic strain rate.
In order to turn on kinematic hardening, the user needs to specify , the maximum allowable translation of the yield
surface, and a material parameter , which determines the speed at which the yield surface approaches the limit
surface. To suppress kinematic hardening, set both and to zero.
v
p
v
p
W e
D
1
X X
0
( ) D
2
X X
0
( )
2
1 =
W D
1
D
2
X
0
p
f
-------
=
h
min
3
f
------ -
v
p
X
---------
X
-------
----------------- -
I
1

R
--------------
v
p
X
---------
X
-------
----------------- - , =
=
H
H 1
J
2
N
----------
\ .
|
| |
dev
f
------
\ .
| |
=
dev ( ) =

N
H
H N
443
CHAPTER 7
Material Library
Non-associative flow rule
The exponential cap model supports non-associatively by allowing the user to specify a flow function which has
the same form as the yield function . The flow potential can be made to differ from the yield function by assigning
values to , , , and that differ from their counterpart parameters ( , , , and ) used to
define the yield surface. To set an associative flow rule, set all potential flow coefficients ( , , , and
) to zero.
Simplification to standard models
The exponential cap powder model can be simplified to standard models such as von Mises, Linear Mohr-Coulomb
etc. by setting some of the parameters to zero. The following simplifications are documented here:
1. von Mises
The von Mises yield surface can be obtained by the following simplifications:
Set and to 0
Set
Set (cap at infinity)
Set , , ,and to 0
Set
2. Linear Mohr-Coulomb
The Linear Mohr-Coulomb yield surface can be obtained by the following
simplifications:
Set
Set
Set
Set (cap at infinity)
Set , , ,and to 0
Set
f
A
2
PF
A
4
PF
R
PF
PF
A
2
A
4
R
A
2
PF
A
4
PF
R
PF
PF
f J
2
( ) J
2
y
3
------- =
A
2
A
4
A
1
A
3

y
3
------- =
0
1e30 =
R W D
1
D
2
1 =
f J
2
( ) I
1
J
2
y
3
------- + =
A
2
0 =
A
4
=
A
1
A
3

y
3
------- =
0
1e30 =
R W D
1
D
2
1 =

User Input Parameters
Symbol Description SI Units
Youngs modulus, supports table input Pa
Poissons ratio, supports table input dimensionless
Density, supports table input
Kg/m
3
Coefficient of thermal expansion, to be used for thermal analysis 1/K
First constant term in the function defining the shear failure surface Pa
Second constant term in the function defining the shear failure surface
Set and to 0 for simplification to von Mises yield surface
1/Pa
Third constant term in the function defining the shear failure surface
Set to mimic the von Mises yield surface
Pa
Fourth constant term in the function defining the shear failure surface dimensionless
Potential function parameter for setting non-associative flow rule.
for associative flow. Set to zero in input file for associative flow.
1/Pa
Potential function parameter for setting non-associative flow rule. Set
for non-associative flow. Set to zero in input file for associative flow.
dimensionless
Kinematic hardening parameter which governs how quickly the yield surface evolves
towards the limit surface. Set both and to zero to suppress kinematic hardening.
dimensionless
Kinematic hardening parameter which sets the maximum allowable translation of the
yield surface. Set both H and N to zero to suppress kinematic hardening.
Pa
E
C
t h
A
1
A
1
0 >
A
2
A
2
0
A
2
A
4
A
3
A
3
0
A
1
A
3

y
3
------- =
f J
2
( ) J
2

y
=
A
4
A
4
0
A
2
PF
A
2
PF
A
2
=
A
4
PF
A
4
PF
A
2
H
H N
N
N 0
445
CHAPTER 7
Material Library
Plotable output parameters
Cap eccentricity parameter. It is the ratio of the width divided by the height of the
elliptical cap function. This parameter affects the stress level at which dilatation will
occur in triaxial compression.
If not entered, .
dimensionless
This is the initial value of at which the meridional yield profile branches away from
the shear failure surface and morphs into the cap function. must be negative. Set
(cap at infinity) to simulate a von Mises response.
Pa
Limit value of the absolute plastic volumetric strain. This parameter is approximately
equal to the initial porosity of the material and can be obtained from the hydrostatic
crush data.
dimensionless
One third of the slope of the porosity versus pressure crush curve at the elastic limit. 1/Pa
Extra fitting parameter for the hydrostatic crush curve data. Used only when the crush
curve has an inflection point. Else set to zero.
1/Pa
2
The triaxial extension to compression strength ratio (TXE/TXC) Convexity of the
yield surface requires .
If not entered, .
dimensionless
Flow potential analog of the yield surface parameter . Used only for non-associative
flow.
dimensionless
Flow potential analog of the yield surface parameter . Used only for non-associative
flow.
dimensionless
Cap Softening Flag Set this flag to 0 to disallow contraction of the cap. Set to 1 to allow the
powder cap to soften.
Flag
The value of at which the meridional yield profile branches away from
the shear failure surface and morphs into the cap function. This variable is
referred to as Cap Branch Point in Mentat.
Pa
The value of at which the cap surface intersects the x-axis. This variable
is referred to as Cap Nose Point in Mentat.
Pa
R
R 0 >
R 1 =
0
I
1
0
1e30 =
W
D
1
D
2
7
9
---
9
7
---
1 =
R
PF
R
PF
S
I
1
X I
1

Material Parameters
The fully generalized exponential cap powder model requires the user to specify 16 different material parameters.
However, simplifications of the exponential model can be run using as few as two material parameters. Material
parameters for the exponential cap powder model can be inferred by curve fitting the data obtained through a series of
experiments. The experiments needed and the method of curve fitting is detailed below.
Crush Curve Parameters
Step 1: Obtain hydrostatic pressure vs. total volumetric strain data by conducting experiments run to the point of
total pore collapse.
Step 2: Use the elastic unloading curve to determine a shift distance that must be applied at any pressure to remove
the elastic part of the strain. This will give you a plot of pressure vs. plastic volumetric strain.
Current relative density based on total volumetric strain. dimensionless
Stress free relative density based on the plastic volumetric strain. dimensionless
vol
vol
p

p
total
v
W
447
CHAPTER 7
Material Library
Step 3: Use a curve fitting program to fit the above data to the following curve
where
Shear Failure Parameters
Step 1: Conduct load-to-failure triaxial compression tests on the material and record all peak stress states. Each
individual experiment has precisely one stress state at which the second stress variant achieves a peak
value. Characterizing the shear failure surface requires numerous such experiments.
Step 2: For every available load-to-failure experiment, find the stress state at which is larger than any of the
other stress states in that experiment. Construct a table of data ( ) where is the value of
at the stress state at which is at its peak value.
Step 3: Fit the data to the exponential shear failure function using a curve fitting program
and
are the cap eccentricity parameter
is approximately equal to the initial porosity of the material.

3
X
p =
p
v
v
p
W e
D
1
X X
0
( ) D
2
X X
0
( )
2
1 =
X
0

0
Rf
s

0
( ) = R
W
J
2
J
2
I
1
peak
J
2
, I
1
peak
I
1
J
2
f
s
I
1
( ) A
1
A
3
e
A
2
I
1
A
4
I
1
=

More details on obtaining material parameters for the exponential cap powder model can be found in the following
reference:
A.F. Fossum and R.M. Brannon, The Sandia GEOMODEL: Theory and User's Guide, Technical Report SAND2004-
3226, Sandia Nat'l Laboratories, Albuquerque, N.M., 2004.
Obtaining Crush Curve and Shear Failure Parameters by Curve Fitting
in Marc
Marc now supports curve fitting for the crush curve and shear failure parameters through the existing Mentat interface.
The user can input the experimental data via tables and Marc carries out the curve fitting and obtains both the shear
failure and the crush curve parameters. The following steps are used for curve fitting in Marc.
Crush Curve Parameters: ( )
1. Set the value of to in Mentat to indicate that you will be using curve fitting for the crush curve
parameters. The rest of the parameters can be set to 0.
2. Assign a table ID for and include a table of pressure vs. the plastic volumetric strain (X vs. ). Data should
be sorted by X. Sample set of data and the corresponding plot are shown below:
Data
Point
1 -6.00E+09 -0.079988058
2 -5.00E+09 -0.079948061
3 -4.00E+09 -0.079774105
4 -3.00E+09 -0.079017528
5 -2.00E+09 -0.075726998
6 -1.50E+09 -0.071088731
7 -1.00E+09 -0.061415709
8 -8.00E+08 -0.055063898
9 -7.00E+08 -0.051115167
10 -6.00E+08 -0.04654114
11 -5.00E+08 -0.041242797
12 -4.00E+08 -0.03510544
13 -3.00E+08, -0.027996208
14 -2.00E+08 -0.019761202
15 -7.00E+07 -0.007076085
16 -7.00E+06 0
0
W D
1
D
2
, , ,
W 1
W
v
p
X 3 * p =

v
p
449
CHAPTER 7
Material Library
3. Marc internally uses nonlinear regression analysis to fit the data and to obtain the crush curve parameters
( ).
Shear Failure Parameters: ( )
1. Set the value of to in Mentat to indicate that you will be using curve fitting for the shear failure
parameters. The rest of the shear failure parameters can be set to 0.
2. Assign a table ID for and include a table of ( vs. ). Data should be sorted by . Sample set of data
and the corresponding plot are shown:
3. Marc internally uses nonlinear regression analysis to fit the data and to obtain the shear failure parameters
( )
Limitations:
The following limitations should be noted.
1. The implementation treats the material as initially isotropic. Kinematic hardening is the only way to simulate
deformation induced anisotropy.
2. The elasticity model is hypoelastic rather than hyperelastic.
3. While the exponential shear failure curve is quite general, only an elliptic cap function is available for modeling
shear effects on pore collapse and other plastic volume reducing mechanisms.
Data
Point
1 -600000000 631520775
2 -440000000 574163050
3 -400000000 553409562
4 -320000000 500886644
5 -280000000 467829871
6 -220000000 407501319
7 -190000000 371605878
8 -160000000 331244305
9 -130000000 285856240
10 -100000000 234811012
11 -80000000 197291022
12 -60000000 156711243
13 -40000000 112820812
14 -20000000 65348293
15 0 14000000
16 10000000 -13225871
0
W D
1
D
2
, , ,
A
1
A
2
A
3
A
4
, , ,
A
1
1
A
1
I
1
J
2
I
1
I
1
J
2
A
1
A
2
A
3
A
4
, , ,

4. The model is not extensively tested for tensile applications.
5. The triaxial extension/compression ratio is presumed constant. Neither does it change with pressure, nor does
it evolve with time.
6. Temperature dependent plasticity is not incorporated in the material model.
7. While the elastic parameters (Youngs modulus, Poissons ratio) can be made to vary with temperature and
relative density using tables, the same is not supported for the other material parameters.
8. Curve fitting programs used to fit experimental data to the required material parameters are not provided.
Tips and Tricks:
1. Getting the complete set of 16 material parameters for the exponential cap powder model is a daunting task
which requires extensive experimentation and curve fitting. But do remember that in most cases one can make
do with much lesser material parameters by running simplifications of the exponential cap model. Also the
Sandia report, referenced in the material parameters section above, has a handy set of material data for several
geomaterials.
2. Make sure that the material properties satisfy the following constraints:
On violation of any of these conditions, an error message is displayed in the output file. Check output file to
see which condition has been violated.
3. If the cap radius and the strength ratio are not specified, then they are each set to a default value of 1.
4. The exponential cap model assumes a associative flow rule as default when and are not specified
(or specified to be 0) by internally setting and . To set a non-associative flow rule set
values of and different than values of and .
(Unloaded virgin material must be below yield)
(Non-negative slope at low pressures)
(Positive Lode radius)

(Convexity condition)
(Non-negative slope at low pressures)

(Initial cap position should be negative)
(Positive cap radius)

A
1
A
3
N 0 ( )
A
2
A
3
A
4
0 +
A
2
0
A
3
0
A
4
0
0
0
R 0 >
7 9 9 7
R
A
2
PF
A
4
PF
A
2
PF
A
2
= A
4
PF
A
4
=
A
2
PF
A
4
PF
A
2
A
4
451
CHAPTER 7
Material Library
5. Many times, powders are subjected to hydrostatic loading or close to hydrostatic loading. In such cases, is
close to zero and the yield stress is return mapped back to the nose of the cap. When running problems where
such hydrostatic loading is likely, better convergence can be achieved by using only the contribution of the
deviatoric stress to the initial stress stiffness matrix. This option can be found under LOADCASE > STATIC >
SOLUTION CONTROL.
6. The exponential cap model has not been extensively tested under tension as most powder materials are
subjected to compression in HIP/CIP processes. The shear failure function and its derivatives become very
large as the exponent increases in the positive regime. We have internally set the limiting value of
to be 10 and issue the warning of large hydrostatic tension in element whenever this condition is
violated. Small amounts of tension are acceptable as is typically small. Only when one gets into large
amounts of tension and becomes really large the above condition will be violated which may make the return
mapping numerically unstable. In such cases a warning is printed in the output files and the simulation
continues for as long as possible.
Oak Ridge National Laboratory Options
In Marc, the ORNL options are based on the definitions of ORNL-TM-3602 [Ref. 10] for stainless steels and ORNL
recommendations [Ref. 16] for 2 1/4 Cr-1 Mo steel.
The initial yield stress should be used for the initial inelastic loading calculations for both the stainless steels and 2 1/4
Cr-1 Mo steel. The 10th-cycle yield stress should be used for the hardened material. The 100th-cycle yield stress must
be used in the following circumstances:
1. To accommodate cyclic softening of 2 1/4 Cr-1 Mo steel after many load cycles
2. After a long period of high temperature exposure
3. After the occurrence of creep strain
To enter initial and 10th-cycle yield stresses, use the ISOTROPIC or ORTHOTROPIC model definition option.
Effects of Temperature and Strain Rates on Yield Stress - not using table driven input
Marc allows you to input a temperature-dependent yield stress. To enter the yield stress at a reference temperature, use
the ISOTROPIC or ORTHOTROPIC model definition option. To enter variations of yield stress with temperatures, use
the TEMPERATURE EFFECTS and ORTHO TEMP model definition options. Repeat the TEMPERATURE EFFECTS
and ORTHO TEMP model definition options for each material, as necessary. The effect of temperatures on yielding is
discussed further in Constitutive Relations.
Marc allows you to enter a strain rate dependent yield stress, for use in dynamic and flow (for example, extrusion)
problems. To use the strain rate dependent yield stress in static analysis, enter a fictitious time using the TIME STEP
option. The zero-strain-rate yield stress is given on the ISOTROPIC or ORTHOTROPIC options. Repeat the STRAIN
RATE model definition option for each different material where strain rate data are necessary. Refer to Constitutive
Relations for more information on the strain-rate effect on yielding.
J
2
A
2
I
1
A
2
I
1
A
2
I
1

Work or Strain Hardening
In a uniaxial test, the workhardening slope is defined as the slope of the stress-plastic strain curve. The workhardening
slope relates the incremental stress to incremental plastic strain in the inelastic region and dictates the conditions of
subsequent yielding. The yield stress and the workhardening data must be compatible with the procedure used in the
analysis. For small strain analyses, the engineering stress and engineering strain are appropriate. If only the LARGE
DISP parameter is used, the yield stress should be entered as the second Piola-Kirchhoff stress, and the workhard data
be given with respect to plastic Green-Lagrange strains. If the LARGE STRAIN parameter is used, the yield stress must
be defined as a true or Cauchy stress, and the workhardening data with respect to logarithmic plastic strains.
Work hardening can be specified in Marc in the different ways listed below.
1. Table driven input
2. WORK HARD option
3. Chaboche
4. Power Law
5. Rate Power Law
6. Johnson-Cook
7. Kumar
8. User subroutine (WKSLP)
The first two methods enables the user to define the yield stress as a function of independent variables (like equivalent
plastic strain, temperature, and strain rate) through user-specified data. Apart from the above methods, Marc provides
commonly used hardening laws available in the literature like Chaboche, Power law, Rate Power Law, Johnson-Cook,
and Kumar models. The user can also customize the hardening law through the WKSLP user subroutine. Brief
explanations of all the above mentioned classes of models are given below.
Flow Stress Definition - using table driven input
When using the table driven input format, the yield (flow) stress dependence on the temperature, equivalent plastic
strain, and the strain rate is defined simultaneously through a single table. As this table may be either a piecewise linear
description or a mathematical equation based upon these independent variables, it allows a very general definition.
The value of the yield stress is obtained by evaluating the table, including the reference value given on the ISOTROPIC
or ORTHOTROPIC option. The strain hardening slope is obtained when necessary by numerically differentiating the
values given.
Work Hardening Definition - not using table input
When using the nontable input format, the yield (flow) stress dependence on the equivalent plastic strain, temperature,
and strain rate are defined through the WORK HARD, TEMPERATURE EFFECTS, and STRAIN RATE options,
respectively. The latter two options were discussed in the previous section. The WORK HARD option is defined in more
detail in this section.
There are two methods to enter this information, using the WORK HARD option. In the first method, you must enter
workhardening slopes for uniaxial stress data as a change in stress per unit of plastic strain (see Figure 7-43) and the
plastic strain at which these slopes become effective (breakpoint).
453
CHAPTER 7
Material Library
Figure 7-43 Workhardening Slopes
In the second method, you enter a of yield stress, plastic strain points. This option is flagged by adding the word DATA
to the work hard statement.
Chaboche Model
The Chaboche model is described in detail in the Time-dependent Inelastic Behavior section.
Power Law Model
where

is the yield stress, is the strain corresponding to initial yield stress, is equivalent strain, is rate of
equivalent strain rate and , , , and are material parameters.
Rate Power Law Model
where the parameters are same quantities as that of Power law.
Note:
The slopes of the workhardening curves should be based on a plot of the stress versus plastic strain for a
tensile test. The elastic strain components of the stress-strain curve should not be included. The first
breakpoint of the workhardening curve should be 0.0.
Note:
The data points should be based on a plot of the stress versus plastic strain for a tensile test. The elastic
strain components should not be included. The first plastic strain should equal 0.0 and the first stress should
agree with that given as the yield stress in the ISOTROPIC or ORTHOTROPIC options.
E E E E
Strain
Stress
1
p
2
p
3
p
1
p
----------
2
p
----------
3
p
----------
0.0
1
p
1
p
2
p
+
Slope Breakpoint
and so on.
y
A
0
+ ( )
m
B
n
+ =
y

0

A B m n
y
A
m
n
B + =

7
Material Library
Johnson-Cook Model
where

is the yield stress, is the equivalent plastic strain, is the current equivalent strain rate, is strain rate
of material characterization and , , , and are material parameters. , , are, respectively, the
current, room, and melting temperatures of the material in absolute scale.
The following conditions should be noted for the Johnson-Cook model.
should be between and . If , is set to . If ,
.
and . If either condition is violated, the middle term in the above expression is set to 1.0.
Kumar Model
where is the yield stress, is equivalent strain rate, is the activation energy, is temperature, and , , and
are material parameters.
Yada Law
Used for grain size predictions, is a generic variation of Kumar model. If the constants and in the Kumar model
are related to strain and the material constants , , , , , and as given below, we get Yada Law.
WKSLP User Subroutine
The user-subroutine option can be employed to specific requirement of hardening law. The yield stress can be a
function of temperature, strain, strain-rate, etc. The user is required to return the work hardening slope and, optionally,
the yield stress at each integration point. If the latter is not defined, then the current yield stress is calculated internally.
WKSLP can be called multiple times within each cycle to capture the dependence of the slope and the yield stress on
temperature, strain, strain-rate, etc.
Work Hardening Rules
A number of workhardening rules (isotropic, kinematic, and combined) are available in Marc. A description of these
workhardening rules is given below.
y
A B
p
n
+ ( ) 1 C

0
-----
\ .
|
| |
ln +
\ .
|
| |
1
T T
r oom
T
mel t
T
r oom
------------------------------------
\ .
| |
m
\ .
| |
=
y

p
0
A B C m n T T
r oom
T
mel t
T T
r oom
T
mel t
T T
r oom
< T T
r oom
T T
mel t
>
T T
mel t
0.01 =
0 0 >

y
B
0
* h
1

A
----
\ .
| |
1 n
e
Q nRT ( )
sin =
Q T A B
0
n
A n
B
1
B
2
B
3
B
4
B
5
B
6
455
CHAPTER 7
Material Library
Isotropic Hardening
The isotropic workhardening rule assumes that the center of the yield surface remains stationary in the stress space,
but that the size (radius) of the yield surface expands, due to workhardening. The change of the von Mises yield surface
is plotted in Figure 7-44b.
A review of the load path of a uniaxial test that involves both the loading and unloading of a specimen will assist in
describing the isotropic workhardening rule. The specimen is first loaded from stress free (point 0) to initial yield at
point 1, as shown in Figure 7-44a. It is then continuously loaded to point 2. Then, unloading from 2 to 3 following the
elastic slope E (Youngs modulus) and then elastic reloading from 3 to 2 takes place. Finally, the specimen is plastically
loaded again from 2 to 4 and elastically unloaded from 4 to 5. Reverse plastic loading occurs between 5 and 6.
It is obvious that the stress at 1 is equal to the initial yield stress and stresses at points 2 and 4 are larger than ,
due to workhardening. During unloading, the stress state can remain elastic (for example, point 3), or it can reach a
subsequent (reversed) yield point (for example, point 5). The isotropic workhardening rule states that the reverse yield
occurs at current stress level in the reversed direction. Let be the stress level at point 4. Then, the reverse yield can
only take place at a stress level of (point 5).
The isotropic workhardening model (with a work slope of 0) is the default option in Marc.
For many materials, the isotropic workhardening model is inaccurate if unloading occurs (as in cyclic loading
problems). For these problems, the kinematic hardening model or the combined hardening model represents the
material better.
Figure 7-44 Schematic of Isotropic Hardening Rule (Uniaxial Test)
Kinematic Hardening
Under the kinematic hardening rule, the von Mises yield surface does not change in size or shape, but the center of the
yield surface can move in stress space. Figure 7-45b illustrates this condition. Pragers law is used to define the
translation of the yield surface in the stress space.
The loading path of a uniaxial test is shown in Figure 7-45a. The specimen is loaded in the following order: from stress
free (point 0) to initial yield (point 1), 2 (loading), 3 (unloading), 2 (reloading), 4 (loading), 5 and 6 (unloading). As
y

y
6
5
1
2
6
5
4
3
2
1
0
E
E
E
+
4
4
4
3
0
(a) Loading Path (b) von Mises Yield Surface
1

in isotropic hardening, stress at 1 is equal to the initial yield stress , and stresses at 2 and 4 are higher than , due
to workhardening. Point 3 is elastic, and reverse yield takes place at point 5. Under the kinematic hardening rule, the
reverse yield occurs at the level of , rather than at the stress level of . Similarly, if the specimen
is loaded to a higher stress level (point 7), and then unloaded to the subsequent yield point 8, the stress at point 8
is . If the specimen is unloaded from a (tensile) stress state (such as point 4 and 7), the reverse yield
can occur at a stress state in either the reverse (point 5) or the same (point 8) direction.
To invoke the kinematic hardening in Marc, use the model definition options ISOTROPIC, ORTHOTROPIC, or
ANISOTROPIC. To input workhardening slope data, use the TABLE or WORK HARD option or the WKSLP user
subroutine.
For many materials, the kinematic hardening model gives a better representation of loading/unloading behavior than
the isotropic hardening model. For cyclic loading, however, the kinematic hardening model can represent neither
cyclic hardening nor cyclic softening.
Figure 7-45 Schematic of Kinematic Hardening Rule (Uniaxial Test)
Combined Hardening
Figure 7-46 shows a material with highly nonlinear hardening. Here, the initial hardening is assumed to be almost
entirely isotropic, but after some plastic straining, the elastic range attains an essentially constant value (that is, pure
kinematic hardening). The basic assumption of the combined hardening model is that such behavior is reasonably
approximated by a classical constant kinematic hardening constraint, with the superposition of initial isotropic
hardening. The isotropic hardening rate eventually decays to zero as a function of the equivalent plastic strain
measured by
(7-176)
y

y
5

4
2
y
( ) =
4
8

7
2
y
( ) =
(b) von Mises Yield Surface
1
2
3
4
5
6
7
8
0
y
2
y
(a) Loading Path
2
y
p
dt
}
2
3
---
i j
p
i j
p
\ .
| |
1 2
dt
}
= =
457
CHAPTER 7
Material Library
This implies a constant shift of the center of the elastic domain, with a growth of elastic domain around this center
until pure kinematic hardening is attained. In this model, there is a variable proportion between the isotropic and
kinematic contributions that depends on the extent of plastic deformation (as measured by ).
Use the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition option to activate the combined
workhardening option in Marc. Use the TABLE or WORK HARD option or the WKSLP user subroutine to input
workhardening slope data.
There are two formulations in the Marc program. The older one occurs whenever PLASTICITY,4 is not included as a
parameter. The workhardening data at small strains governs the isotropic behavior, and the data at large strains
( ) governs the kinematic hardening behavior. If the last workhardening slope is zero, the behavior is the
same as the isotropic hardening model.
Figure 7-46 Basic Uniaxial Tension Behavior of the Combined Hardening Model
The newer formulation, which is invoked by including a PLASTICITY,4 parameter allows greater generality by
introducing the fractional contribution ( ) to kinematic hardening as a user input. As shown in Figure 7-47, is the
value between and .
Isotropic hardening
Kinematic hardening
Combined hardening (default )
Combined hardening model utilizing kinematic fraction factor is available for Hill and Barlat models in ISOTROPIC,
ORTHOTROPIC, and ANISOTROPIC options.
p
1000 >
Fully Hardened
Pure Kinematic
Range
Combined
Hardening
Range
Initial
Elastic
Range
Initial
Yield
One-half Current
Elastic Range
Kinematic Slope,
Stress
Strain
3
2
---
d
d
p
--------
f
h
f
0 1
f
h
0 =
f
h
1 =
0 f
h
1 < < f
h
0.5 =

Figure 7-47 Schematic Explanation for Kinematic Hardening Fraction
Flow Rule
Yield stress and workhardening rules are two experimentally related phenomena that characterize plastic material
behavior. The flow rule is also essential in establishing the incremental stress-strain relations for plastic material. The
flow rule describes the differential changes in the plastic strain components as a function of the current stress
state.
The Prandtl-Reuss representation of the flow rule is available in Marc. In conjunction with the von Mises yield
function, this can be represented as:
(7-177)
where and are equivalent plastic strain increment and equivalent stress, respectively.
The significance of this representation is illustrated in Figure 7-48. This figure illustrates the stress-space for the
two-dimensional case. The solid curve gives the yield surface (locus of all stress states causing yield) as defined by
the von Mises criterion.
Equation (7-177) expresses the condition that the direction of inelastic straining is normal to the yield surface. This
condition is called either the normality condition or the associated flow rule.
If the von Mises yield surface is used, then the normal is equal to the deviatoric stress.
Figure 7-48 Yield Surface and Normality Criterion 2-D Stress Space
Kinematic Hardening Isotropic Hardening Work Hardening Curve

y
1 f
h
( )
*

y
( ) + = f
y
( ) =
p

p
d
p
d
i j
p
d
p

i j
---------- =
d
p
Yield Surface
2
d
2
d
p
d
p
1
1
p
459
CHAPTER 7
Material Library
Constitutive Relations
This section presents the constitutive relation that describes the incremental stress-strain relation for an elastic-plastic
material. The material behavior is governed by the incremental theory of plasticity, the von Mises yield criterion, and
the isotropic hardening rule.
Let the workhardening coefficient be expressed as:
(7-178)
and the flow rule be expressed as:
where (7-179)
Consider the differential form of the familiar stress-strain law, with the plastic strains interpreted as initial strains
(7-180)
where is the elasticity matrix defined by Hookes law and : denotes the tensor contraction.
After substitution of Equation (7-179), this becomes
(7-181)
Contracting Equation (7-181) by
(7-182)
and recognizing that
(7-183)
with use of Equation (7-178) in place of the left-hand side,
(7-184)
By rearrangement
(7-185)
Finally, by substitution of this expression into Equation (7-181), we obtain
(7-186)
where is the elasto-plastic, small strain tangent moduli expressed as:
(7-187)
The case of perfect plasticity, where , causes no difficulty.
H
H d d
p
=
d
p
d
p
: =

i j
---------- =
d C : d C : d
p
=
C
d C : d C : d
p
=
: d : C : d : C : d
p
=
d : d =
Hd
p
: C : d : C : d
p
=
d
p
: C : d
H : C : +
------------------------------------------- =
d L
ep
:d =
L
ep
L
ep
C
C : ( ) C : ( )
H : C : +
------------------------------------------------------ =
H 0 =

Temperature Effects
This section discusses the effects of temperature-dependent plasticity on the constitutive relation.
The following constitutive relations for thermo-plasticity were developed by Naghdi. Temperature effects are
discussed using the isotropic hardening model and the von Mises yield condition.
The stress rate can be expressed in the form
(7-188)
For elastic-plastic behavior, the moduli are
(7-189)
and for purely elastic response
(7-190)
The term that relates the stress increment to the increment of temperature for elastic-plastic behavior is
(7-191)
and for purely elastic response
(7-192)
where
(7-193)
and
(7-194)
and are the coefficients of thermal expansion.
Strain Rate Effects
This section discusses the influence of strain rate on the elastic-plastic constitutive relation.
Strain rate effects cause the structural response of a body to change because they influence the material properties of
the body. These material changes lead to an instantaneous change in the strength of the material. Strain rate effects
become more pronounced for temperatures greater than half the melting temperature ( ). The following discussion
explains the effect of strain rate on the size of the yield surface.
i j
L
i j kl
kl
h
i j
T
+ =
L
i j kl
L
i j kl
C
i j kl
C
i j mn

mn
-------------
pq
------------ C
pqkl
\ .
| |
D =
L
i j kl
C
i j kl
=
h
i j
X
i j
C
i j kl
kl
C
i j kl
kl
-----------
pq
X
pq
2
3
---
T
-------
\ .
| |
\ .
| |
D =
H
i j
X
i j
C
i j kl
kl
=
D
4
9
---
2

p
--------

i j
---------- C
i j kl

kl
----------- + =
X
i j
C
i j kl
T
---------------
kl
e
=
kl
T
m
461
CHAPTER 7
Material Library
Using the von Mises yield condition and normality rule, we obtain an expression for the stress rate of the form
(7-195)
For elastic-plastic response
(7-196)
and
(7-197)
where
(7-198)
Time-independent Cyclic Plasticity
The cyclic plasticity model is based on the work of Chaboche [Ref. 23]. The current version of Marc consists only of
the basic model and plastic-strain-range memorization. The associated time-dependent model is described on
Time-dependent Cyclic Plasticity.
The model combines the isotropic hardening rule, to describe the cyclic hardening (Figure 7-49a) or softening, and the
nonlinear kinematic hardening to capture the proper characteristic of cyclic plasticity like Bauschinger (Figure 7-49b),
ratchetting (Figure 7-49c), and mean-stress relaxation (Figure 7-49d) effect. The influence of the plastic strain range
on the stabilized cyclic response is taken into account by introducing the plastic-strain-range memorization variable
(Figure 7-49e).
The von Mises yield function is now defined as follows:
where , , and
is the back stress tensor representing the center of the yield surface in stress space.
i j
L
i j kl
kl
r
i j
p
+ =
L
i j kl
C
i j kl
C
i j mn

mn
-------------
pq
------------ C
pqkl
\ .
| |
D =
r
i j
C
i j mn

mn
-------------
2
3
---
p
-------- D =
D
4
9
---
2

p
--------

i j
---------- C
i j kl

kl
----------- + =
f R k + ( ) =
3S
i j
S
i j
2
-----------------
\ .
| |
1
2
---
= s
i j
'
i j
1
3
---
i j
'
kk
= ' X =
X

Figure 7-49 Typical Behavior of Material that can be Simulated with Cyclic Plasticity Model
Isotropic Hardening/Softening
The isotropic hardening/softening determines the size of the elastic region during the plastic loading. In this model, it
is controlled by parameter and . The initial conditions of cyclic hardening are given as and ,
while a cyclic softening is initially described by and . The evolution equation for the variable
is described as follows:
where and are material constants. represents the limit of the isotropic hardening/softening. In case of
hardening, then .
(b) Bauschinger Effect (c) Ratchetting
(d) Mean Stress Relaxation (e) Cyclic Hardening

(a) Cyclic Hardening under
Multiple Cyclic Loading
R k k
y
= R 0 =
k
y
R
0
= R R
0
=
R
R
b R
R ( )
=
b R
R R
1 e
b e
ps
\ .
| |
=
463
CHAPTER 7
Material Library
Nonlinear Kinematic Hardening
The nonlinear kinematic hardening is defined from the linear-Ziegler rule by adding the recall term as shown in the
evolution of the back stress tensor below:
where and are two material constants. stands for linear-kinematic rule.
Plastic-strain-range Memorization
Several experimental observations show that the asymptotic stress value of cyclic hardening can depend on the prior
history. The influence of plastic-strain range on the stabilized cyclic response is evident from the comparison between
the different histories of loading used to obtained the cyclic curve. Therefore, an introduction of new internal variables
that memorize the prior maximum plastic range is introduced by defining a memory surface in the plastic strain
space as follows:
The evolution of the state variables are as follows
where and are the unit normal to the yield surface and to the memory surface defined as follows:
and
The coefficient is introduced in order to induce a progressive memory. For then, the memorization is
instantaneous and stabilization occurs after one cycle. A progressive memory is given by .
The dependency between cyclic plastic flow and the plastic strain range is introduced by considering an asymptotic
isotropic state as follows:
where , and are material constants.
X
C
R k +
-------------- X ( ) X
=
C 0 =
F
2
3
---
e

p
( ) =
H F ( ) n n
*
( )
3 2 1 ( )H F ( ) n n
*
( ) n
*
=
n n
*
f 0 = F 0 =
n
2
3
---

p
----- = n
* 2
3
---

p

------------ =
0.5 =
0.5 <
R
Q
M
Q
0
Q
M
( )e
2
+ =
Q
M
Q
0


Plastic Evolution Process and Elasto-plastic Classical Modular Matrix
The plastic evolution process must conform to the consistency condition, . From this condition the plastic
rate multiplier can be derived as follows:
(7-199)
where .
Since the process involves nonlinear equation, iteration process using predictor-corrector technique is used. The
predictor is calculated using the trial elastic stresses as follows:
(7-200)
and then calculate based on and hardening parameter on A. Using the Taylor expansion at B, then
(7-201)
Having the global plasticity iteration converged, then the iteration to satisfy the plastic strain memorization is started.
If both iterations are converged, then the total plasticity iteration is considered completed.
Inserting Equation (7-202) into Equation and using Equation (7-199), the elasto-plastic classical tangent modular
matrix can be derived as follows:
Time-dependent Inelastic Behavior
Force-displacement relationships vary in different material models. A perfectly elastic material and a perfectly viscous
material can be represented by a spring and a dashpot, respectively (as shown in Figure 7-50). In a perfectly elastic
material, the deformation is proportional to the applied load. In a perfectly viscous material, the rate of change of the
deformation over time is proportional to the load.
In the class of viscoelastic and creeping materials, the application of a constant load is followed by a deformation,
which can be made up of an instantaneous deformation (elastic effect) followed by a continual deformation with time
(viscous effect). Eventually, it can become pure viscous flow. Continued deformation under constant load is termed
creep (see Figure 7-51).
f
0 =
a
T
L
a
T
La
C
R k +
-------------- a
T
X ( ) a
T
X b R
R ( ) + +
--------------------------------------------------------------------------------------------------------------------
=
a
f

------
3
2
---
s
R k +
-------------- = =
B

A
L + =
f
B

B
f
B
a
T
La
C
R k +
-------------- a
T
X ( ) a
T
X b R
R ( ) + +
--------------------------------------------------------------------------------------------------------------------- =
L
EP
L I
aa
T
L
a
T
La
C
R k +
-------------- a
T
X ( ) a
T
X b R
R ( ) + +
--------------------------------------------------------------------------------------------------------------------
\ .
|
|
| |
=
465
CHAPTER 7
Material Library
Figure 7-50 Perfectly Elastic (Spring) and Viscous (Dashpot) Materials
Figure 7-51 The Creep Curve
A viscoelastic material can be subjected to sudden application of a constant deformation. This results in an
instantaneous proportional load (elastic effect), followed by a gradual reduction of the required load with time, until a
limiting value of the load is attained. The decreasing of load for a constant deformation, is termed relaxation (see
Figure 7-52).
Viscoelastic and creeping materials can be represented by models consisting of both springs and dashpots because the
material displays both elastic effects and viscous effects. This implies that the material either continues to flow for a
given stress, or the stress decreases with time for a given strain. The measured relation between stress and strain is
generally very complex.
DASHPOT SPRING
f = ku
f = force
u = Displacement
k = Spring Stiffness
f = u
f = Force
u = Velocity (Time Rate of Change of Displacement
= Viscosity of the Dashpot
(Strain)
t (Time)
0
A
B
C
OA Instantaneous Elastic Effect
AB Delayed Elastic Effect
BC Viscous Flow

Figure 7-52 The Relaxation Curve
Two models that are commonly used to relate stress and strain are the Maxwell and Kelvin (Voigt or Kelvin-Voigt)
models. A description of these models is given below.
The mathematical relation which holds for the Maxwell solid is
(7-202)
In the one-dimensional case for normal stress
(7-203)
This relation can be depicted as a spring and dashpot in series, as shown in Figure 7-53. The integration of Equation
(7-202) yields
(7-204)
Figure 7-53 Maxwell Solid
The strain and stress responses of the Maxwell Solid model are shown in Figure 7-54 and Figure 7-55, respectively.
Figure 7-54 Strain Response to Applied Constant Stress (Maxwell Solid)
(Stress)
t (Time)
0
= +
1
E
--- =
--- =

E
---

--- dt
}
+ =

t
Constant Stress Applied
t
Response to Constant Stress
0
E
------
467
CHAPTER 7
Material Library
Figure 7-55 Stress Response to Applied Constant Strain (Maxwell Solid)
The mathematical relation which holds for the Kelvin (Voigt or Kelvin-Voigt) solid is
(7-205)
This equation is depicted as a spring and dashpot in parallel. (See Figure 7-56). When (no dashpot), the
system is a linearly elastic system in which , the elastic modulus.
When (no spring), the solid obeys Newtons equation for a viscous fluid and , the viscous
coefficient. Thus, we can rewrite Equation (7-205) in the form
(7-206)
In the above relation, we have considered one-dimensional normal stress and strain. The relation holds equally well
for shear stress and shear strain in which , the shear modulus, and , the viscous coefficient.
Equation (7-205) can be rewritten as
(7-207)
Figure 7-56 Kelvin (Voigt or Kelvin-Voigt) Solid
The strain responses of the Kelvin Solid model are depicted in Figure 7-57. For multiaxial situations, these equations
can be generalized to tensor quantities.
To invoke the Maxwell model, use the CREEP parameter. The creep strain can be specified as either deviatoric creep
strain (conventional creep) or dilatational creep strain (swelling). To invoke the Kelvin model, also use the CREEP
parameter and CRPVIS user subroutine.
t
Constant Strain Applied
t
Response to Constant Strain
+ =
0 =
E =
E 0 = =
E
+ =
G = =
G
+ =
E

Figure 7-57 Strain Response to Applied Stress (Kelvin Solid)
Creep (Maxwell Model)
Creep is an important factor in elevated-temperature stress analysis. In Marc, creep is represented by a Maxwell model.
Creep is a time-dependent, inelastic behavior, and can occur at any stress level (that is, either below or above the yield
stress of a material). In many cases, creep is accompanied by plasticity which occurs above the yield stress of the
material. The creep behavior can be characterized as primary, secondary, and tertiary creep, as shown in Figure 7-58.
Engineering analysis is often limited to the primary and secondary creep regions. Tertiary creep in a uniaxial specimen
is usually associated with geometric instabilities, such as necking. The major difference between the primary and
secondary creep is that the creep strain rate is much larger in the primary creep region than it is in the secondary creep
region. The creep strain rate is the slope of the creep strain-time curve. The creep strain rate is generally dependent on
stress, temperature, and time.
The creep data can be specified in either an exponent form or in a piecewise linear curve. To specify creep data, use
the CREEP model definition option. The CRPLAW user subroutine allows alternative forms of creep behavior to be
programmed directly.
(7-208)
(a) Stress Pulse
0
t
1
t
(b) Strain Response to Stress of Infinite Domain
t
t
(c) Strain Response to Stress Pulse of Finite Length
t
1
t
0
E
------
c d
c
dt
-------- =
469
CHAPTER 7
Material Library
Figure 7-58 Creep Strain Versus Time (Uniaxial Test at Constant Stress and Temperature)
Marc offers two schemes for modeling creep in conjunction with plasticity: (a) treating creep strains and plastic strains
separately; and (b) modeling creep strains and plastic strains in a unified fashion (viscoplasticity). Both schemes can
be treated using two different procedures: explicit and implicit.
Creep (Explicit Formulation)
There are six possible modes of input for creep constitutive data.
1. Express the dependence of equivalent creep strain rate on any independent parameter through a piecewise
linear relationship. The equivalent creep strain rate is then assumed to be a piecewise linear approximation to
(7-209)
where A is a constant; is equivalent creep strain rate; and , , , and are equivalent stress, equivalent
creep strain, temperature and time, respectively. The functions , , , and are piecewise linear and entered
in the form as either slope-break point data or function-variable data. This representation is shown in
Figure 7-59. Enter functions , , , and through the CREEP model definition option. (Any of the functions
, , , or can be set to unity by setting the number of piecewise linear slopes for that relation to zero on
the input data.)
2. The dependence of equivalent creep strain rate on any independent parameter can be given directly in power
law form by the appropriate exponent. The equivalent creep strain rate is
Enter the constants , , , , and directly through the CREEP model definition option. This is often
adequate for engineering metals at constant temperature where Nortons rule is a good approximation.
(7-210)
Note: Primary Creep: Fast decrease in creep strain rate
Secondary Creep: Slow decrease in creep strain rate
Tertiary Creep: Fast increase in creep strain rate
Secondary
Creep
Primary
Creep
Tertiary
Creep
Creep Strain
C
Time (t)
c
A f ( ) g
c
( ) h T ( )
dk t ( )
dt
------------- =
c

c
T t
f g h k
f g h k
f g h k
c
A
m
c
( )
n
T
p
qt
q 1
( ) =
A m n p q
c
A =
n

3. Input procedure, one can define a table such that
(7-211)
where are four variables that could include equivalent stress, temperature, time, strain, position,
or one of 30 variables. In this case, the function is piecewise linear.
4. Using the table driven input procedure, one can define the strain rate such that
(7-212)
For example, one could enter the equation
(7-213)
The equation can take form as long as there is no conditional logic required. If this is not the case, the CRPLAW
user subroutine should be used.
5. Define the equivalent creep strain rate directly with the CRPLAW user subroutine.
6. Use the ISOTROPIC option to activate the ORNL (Oak Ridge National Laboratory rules) capability of the
program.
Isotropic creep behavior is based on a von Mises creep potential described by the equivalent creep law
(7-214)
The material creep behavior is described by
(7-215)
During creep, the creep strain rate usually decreases. This effect is called creep hardening and can be a function of
time or creep strain. The following section discusses the difference between these two types of hardening.
Consider a simple power law that illustrates the difference between time and strain-hardening rules for the calculation
of the creep strain rate.
(7-216)
where

is the creep strain, and are values obtained from experiments and t is time. The creep rate can be
obtained by taking the derivative with respect to time
(7-217)
c
A f v
1
v
2
v
3
v
4
, , , ( ) =
v
1
v
2
v
3
v
4
, , ,
cr
A equation v
1
v
2
v
3
v
4
, , , ( ) =
A
n
e
QR T
f
c
T t , , , ( ) =
i j
c
i j
----------
)
`

=
c
t
n
=
c
n
c
d
c
dt
-------- nt
n 1
= =
471
CHAPTER 7
Material Library
Figure 7-59 Piecewise Linear Representation of Creep Data
However, being greater than 0, we can compute the time as
(7-218)
Substituting Equation (7-213) into Equation (7-212) we have
(7-219)
(1) Slope-Break Point Data
Slope Break Point
S
1
X
1
S
2
X
2
S
3
X
3
(2) Function-Variable Data
Function Variable
F
1
X
1
F
2
X
2
F
3
X
3
F
4
X
4
Function F (X)
[for example, t ,
g , h (T),
k (t)]
( )
c
( )
X
1
F
1
X
2
F
2
X
3
F
3
X
4
F
4
S
1
S
2
S
3
Variable X (Such as ,
C
, T, t)
t t
t

c
-----
\ .
| |
1 n /
=
c
nt
n 1
n
1 n
c
( )
n 1 ( ) n
( ) = =

Equation (7-213) shows that the creep strain rate is a function of time (time hardening). Equation (7-219) indicates
that the creep strain rate is dependent on the creep strain (strain hardening). The creep strain rates calculated from these
two hardening rules generally are different. The selection of a hardening rule in creep analysis must be based on data
obtained from experimental results. Figure 7-60 and Figure 7-61 show time and strain hardening rules in a variable
state of stress. It is assumed that the stress in a structure varies from to to ; depending upon the model
chosen, different creep strain rates are calculated accordingly at points 1, 2, 3, and 4. Obviously, creep strain rates
obtained from the time hardening rule are quite different from those obtained by the strain hardening rule.
Figure 7-60 Time Hardening
Figure 7-61 Strain Hardening
1

2

3
1
0
1
2
3
4
t
1
0
1
2
3
4
t
3
473
CHAPTER 7
Material Library
7
Material Library
Oak Ridge National Laboratory Laws
Oak Ridge National Laboratory (ORNL) has performed a large number of creep tests on stainless and other alloy
steels. It has also set certain rules that characterize creep behavior for application in nuclear structures. A summary of
the ORNL rules on creep is given below. The references listed at the end of this section offer a more detailed discussion
of the ORNL rules.
1. Auxiliary Rules for Applying Strain-Hardening to Situations Involving Stress Reversals
The Blackburn Creep Law is required as the CRPLAW user subroutine. The parameter EQCP (first parameter
in CRPLAW) is defined as
(7-220)
when the ORNL constitutive option is flagged through use of the ISOTROPIC option. In all other cases, the
definition
(7-221)
is retained. The equivalent primary creep strain passes into CRPLAW in EQCPNC, the second parameter. The
second parameter must be redefined in that routine as the equivalent (total) creep strain increment. The first
parameter (EQCP) must be redefined as the equivalent primary creep strain increment when the ORNL
constitutive option is flagged. During analysis with the ORNL option, equivalent creep strain stores the
distance between the two shifted origins in creep strain space ( in ORNL-TM-3602). The sign on this value
indicates which origin is currently active, so that a negative sign indicates use of the negative origin ( ).
2. Plasticity Effect on Creep
The effect of plastic strains on creep must be accommodated for the time-dependent creep behavior of 2 1/4 Cr
-1 Mo Steel. Since plastic strains in one direction reduce the prior creep strain hardening accumulated in the
reverse direction, ORNL recommends that the softening influence due to plastic strains be treated much the
same as when reversed creep strain occurs. The following quantities are defined:
(7-222)
(7-223)
where is instantaneous creep strain components
= positive and negative strain origins (7-224)
and
(7-225)
(7-226)
c
2
3
---
i j
c
i j
c
\ .
| |
1 2 /
=
2
3
---
c
i j

i j
c
\ .
| |
1 2 /
=
i j
N
i j

i j
I
i j
( ) G =
+ + +
N
i j
i j
I
i j
( ) G
i j
I
i j
+
i j
;
G
+
G
i j
I
i j
+
( ) 2 3
i j
I
i j
+
( )
i j
I
i j
+
( ) [ ]
1 2
= =
G
G
i j
I
i j
( ) 2 3
i j
I
i j
( )
i j
I
i j
[ ]
1 2
= =

Swelling
Marc allows pure swelling (dilatational creep) effect in a creep analysis. To use the swelling option, perform a regular
creep analysis as discussed earlier. Use the VSWELL user subroutine to define the increment of volumetric swelling
. The increment of volumetric swelling is generally a function of neutron flux, time, and temperatures.
For example, radiation-induced swelling strain model for 20% C. W. Stainless Steel 316 can be expressed as:
(7-227)
where , , and are functions of temperature, is neutron flux, and is time.
Creep (Implicit Formulation)
This formulation, as opposed to that described in the previous section, is fully implicit. A fully implicit formulation is
unconditionally stable for any choice of time step size; hence, allowing a larger time step than permissible using the
explicit method. Additionally, this method is more accurate than the explicit method. The disadvantage is that each
increment may be more computationally expensive. This model is activated using the CREEP parameter. There are
two methods for defining the inelastic strain rate. The CREEP model definition option can be used to define a Maxwell
creep model. The back stress must be specified through the field normally reserved for the yield stress in the
ISOTROPIC or ORTHOTROPIC options. The yield stress must be specified through the field normally reserved for the
10th cycle yield stress in the ISOTROPIC option. There is no plastic strain when the stress is less than the yield stress
and there is no creep strain when the stress is less than the back stress.
The equivalent creep strain increment is expressed as
(7-228)
Enter the constants , , , , and directly through the CREEP model definition option. A more general
expression for the equivalent creep strain rate is given by:
(7-229)
Enter the terms and and the functions , , and through the UCRPLW user subroutine.
The creep strain components are given by:
where is the deviatoric stress at the end of the increment. is a function of temperature, time, etc. An algorithmic
tangent is used to form the stiffness matrix.
Based on a parameter defined in the CREEP parameter, one of three tangent matrices is formed. The first is using an
elastic tangent, which requires more iterations, but can be computationally efficient because re-assembly might not be
V
V
--------
\ .
| |
V
V
-------- Rt
R
--- - l n
1 exp t ( ) ( ) +
1 exp +
------------------------------------------------- + =
R t t
c
A
m
c
( )
n
T
p
qt
q 1
( ) =
A m n p q
c
A
m
g
c
( ) h T ( )
dk t ( )
dt
------------- =
A m g h k
i j
3
2
---
i j
d
------- =
i j
d
A
475
CHAPTER 7
Material Library
required. The second is an algorithmic tangent that provides the best behavior for small strain power law creep. The
third is a secant (approximate) tangent that gives the best behavior for general viscoplastic models. When creep is
specified in conjunction with plasticity, the elastic tangent option is not available.
Viscoplasticity
Explicit Formulation
The creep (Maxwell) model can be modified to include a plastic element (as shown in Figure 7-62). This plastic
element is inactive when the stress ( ) is less than the yield stress ( ) of the material. The modified model is an
elasto-viscoplasticity model and is capable of producing some observed effects of creep and plasticity. In addition, the
viscoplastic model can be used to generate time-independent plasticity solutions when stationary conditions are
reached. At the other extreme, the viscoplastic model can reproduce standard creep phenomena. The model allows the
treatment of nonassociated flow rules and strain softening which present difficulties in conventional (tangent modulus)
plasticity analyses.
The viscoplasticity option can be used to implement very general constitutive relations with the aid of the following
user subroutines: ZERO, YIEL, NASSOC, and CRPLAW. See Nonlinear Analysis in Chapter 5 for details on how to use
these procedures.
Figure 7-62 Uniaxial Representation of Viscoplastic Material
Implicit Formulation
To allow for the implementation of general unified creep-plasticity or viscoplastic models, the UVSCPL user
subroutine is available. This routine requires you to define only the inelastic strain rate. The program automatically
calculates a tangent stiffness matrix (only elastic tangent or secant tangent can be used). This option is activated
by indicating that the material is VISCO PLAS in the ISOTROPIC or ORTHOTROPIC option.
Time-dependent Cyclic Plasticity
The time-dependent effect of the model described in Time-independent Cyclic Plasticity on page 461 is modeled
using the unified viscoplastic framework. The viscoplastic potential is based on overstress quantity as follows:

y
Plastic Element
Inactive if <
y
vp
=
vp
K
n 1 +
-------------
f
K
---- ( )
n 1 +
=

The viscoplastic strain rate is defined as follows:
(7-230)
where
(7-231)
In this case, the viscoplastic stress is .
The hardening rules are chosen to be identical to the time-independent case.
Viscoplastic Evolution Process and Visco-plastic Classical Modular Matrix
The iteration procedure, starting from the trial elastic stress as the predictor, is based on the implicit integration of
Equation (7-231) that can be expressed as follows:
(7-232)
Using Newton iteration scheme, the iterative value of (that is, ) can be expressed as follows:
where is the i-th iteration of the residual of Equation (7-232).
The tangent modular matrix is based on the assumption that . Therefore the classical visco-plastic modular
matrix can be expressed as follows:
This matrix is in line with the consistent model derived in [Ref. 20].
Anand Solder Model
The Anand Solder model allows the user to model a solder material using the unified Anand model which consists of
a simple set of constitutive equations for large, isotropic, viscoplastic problems. There are two basic features in this
vp

3
2
---
' X'
e
-----------------
3
2
---
f = = =
f
K
---- ( )
n
=
vp
K
1
n
---
=
r

t
------
f
K
---- ( )
n
+ =

t r
n
i
1
n
K
----
f
K
---- ( )
n 1
a
T
La
2
3
--- Ca
T
a a
T
X b R
R ( ) + +
\ .
| |
+
--------------------------------------------------------------------------------------------------------------------------------------- =
r
i
r
0 =
L
vp
L I
aa
T
L
a
T
La
2
3
--- Ca
T
a a
T
X b R
R ( )
1
n
K
----
f
K
---- ( )
n 1
t
-------------------------------- + + +
------------------------------------------------------------------------------------------------------------------------------------
\ .
|
|
|
|
|
| |
=
477
CHAPTER 7
Material Library
Anand model. First, this model needs no explicit yield condition and no loading/unloading criterion. The plastic strain
is assumed to take place at all nonzero stress values, although at low stresses the rate of plastic flow may be
immeasurable. Second, this model employs a single scalar as an internal variable to represent the isotropic resistance
to plastic flow offered by the internal state of the material.
The Anand Solder model uses a single scalar internal variable, , which denotes the averaged isotropic resistance to
macroscopic plastic flow offered by the underlying isotropic strengthening mechanisms such as dislocation density,
solid solution strengthening, subgrain, and grain size effects, etc. The deformation resistance s is consequently
proportional to the equivalent stress. The flow equations and the evolution equations are given below:
Flow Equation:
where
where the evolution equation is expressed as:
Evolution Equations:
where
is the single internal variable representing deformation resistance
is the pre-exponential factor
is the multiplier of stress
is the strain rate sensitivity of stress
is the Activation Energy/Boltzmanns Constant
is the hardening constant
is the deformation resistance saturation coefficient
is the strain rate sensitivity of saturation
is the strain rate sensitivity of hardening
s
d
p
dt
-------- A

s
-------
\ .
| |
sinh
1
m
----
Q
kT
-------
\ .
| |
exp =
s s
0
A
m
Q k
ds
dt
----- h
0
B
a B
B
-------
)
`

d
p
dt
-------- =
B 1
s
s
*
----- =
s
*
s
1
A
----
d
p
dt
--------
Q
kT
-------
\ .
| |
exp
n
=
h
0
s
n
a

represents the saturation value of associated with a set of given temperatures and strain rates. Thus the Anand
Solder model has nine material parameters: , , , , , , , , where is the initial value of the deformation
resistance needed to determine the evolution of the deformation resistance.
Determination of Parameters for Anand Model
There are some standard test procedures for determination of the constants for the Anand Solder model. These involve
a steady state creep test followed by use of a nonlinear fitting method to determine the constants. More details about
the exact nature of tests required can be found in [Ref. 31]. The Anand material constants for some standard solder
alloys are included in the table below.
Material parameters of Anand model for some solders:
Viscoelastic Material
Marc has two models that represent viscoelastic materials. The first can be defined as a Kelvin-Voigt model. The latter
is a general hereditary integral approach.
Kelvin-Voigt Model
The Kelvin model allows the rate of change of the inelastic strain to be a function of the total stress and previous strain.
To activate the Kelvin model in Marc, use the CREEP parameter.
The Kelvin material behavior (viscoelasticity) is modeled by assuming an additional creep strain , governed by
(7-233)
Material
Parameters
Solders
60Sn40Pb 62Sn36Pb2Ag 96.5Sn3.5Ag 97.5Pb2.5Sn
A
1.49e7 2.3e7 2.23e4 3.25e12
Q/R
10830 11262 8900 15583
11 11 6 7
0.303 0.303 0.182 0.143
(MPa)
80.42 80.79 73.81 72.73
0.0231 0.0212 0.018 0.00437
(MPa) 2640.75 4121.31 3321.15 1787.02
1.34 1.38 1.82 3.73
(MPa)
56.33 43.32 39.09 15.09
s
*
s
A Q m h
0
s
n a s
0
s
1
( )
K ( )
m
s
n
h
0
a
s
0
i j
k
d
dt
-----
i j
k
A
i j kl
kl
d
B
i j kl
kl
k
=
479
CHAPTER 7
Material Library
where and are defined in the CRPVIS user subroutine and the total strain is
(7-234)
thermal strain components (7-235)
elastic strain components (instantaneous response) (7-236)
plastic strain components (7-237)
creep strains defined via the CRPLAW and VSWELL user subroutines (7-238)
Kelvin model strain components as defined above (7-239)
The CRPVIS user subroutine is called at each integration point of each element when the Kelvin model is used.
Use the AUTO CREEP option to define the time step and to set the tolerance control for the maximum strain in any
increment.
The CREEP option allows Maxwell models to be included in series with the Kelvin model.
Hereditary Integral Model
The stress-strain equations in viscoelasticity are not only dependent on the current stress and strain state (as
represented in the Kelvin model), but also on the entire history of development of these states. This constitutive
behavior is most readily expressed in terms of hereditary or Duhamel integrals. These integrals are formed by
considering the stress or strain build-up at successive times. Two equivalent integral forms exist: the stress relaxation
form and the creep function form. In Marc, the stress relaxation form is used.
The viscoelasticity option in Marc can be used for both the small strain and large strain Mooney or Ogden material
stress-relaxation problems for total Lagrange formulation. It can also be used with all hyperelastic models; i.e.,
Mooney, Ogden, Gent, Arruda-Boyce, Foam and generalized hyperelastic materials if the Updated Lagragian
formulation is used. A description of these models is as follows:
Small Strain Viscoelasticity
In the stress relaxation form, the constitutive relation can be written as a hereditary integral formulation
(7-240)
A B
i j

i j
e
i j
p
i j
c
i j
k
i j
t h
+ + + + =
i j
t h
=
i j
e
=
i j
p
=
i j
c
=
i j
k
=
i j
t ( ) G
i j kl
t ( )
d
kl
( )
d
------------------ d
0
t
}
G
i j kl
t ( )
kl
0 ( ) + =

The functions are called stress relaxation functions. They represent the response to a unit applied strain and have
characteristic relaxation times associated with them. The relaxation functions for materials with a fading memory can
be expressed in terms of Prony or exponential series.
(7-241)
in which is a tensor of amplitudes and is a positive time constant (relaxation time). In the current
implementation, it is assumed that the time constant is isotropic. In Equation (7-241), represents the long term
modulus of the material.
The short term moduli (describing the instantaneous elastic effect) are then given by
(7-242)
The stress can now be considered as the summation of the stresses in a generalized Maxwell model (Figure 7-63)
(7-243)
where
(7-244)
(7-245)
Figure 7-63 The Generalized Maxwell or Stress Relaxation Form
G
i j kl
G
i j kl
t ( ) G
i j kl
G
i j kl
n
exp t
n
( )
n 1 =
N
+ =
G
i j kl
n
n
G
i j kl
G
i j kl
0
G
i j kl
0 ( ) G
i j kl
G
i j kl
n
n 1 =
N
+ = =
i j
t ( )
i j
t ( )
i j
n
t ( )
n 1 =
N
+ =
i j
G
i j kl
kl
t ( ) =
i j
n
G
i j kl
n
exp t ( )
n
[ ]
d
kl
( )
d
------------------ d
0
t
}
=
E
2
2
E
3
3
E
N
N
E
1
1
E
i
E
i
=
481
CHAPTER 7
Material Library
For integration of the constitutive equation, the total time interval is subdivided into a number of subintervals
( ) with time-step . A recursive relation can now be derived expressing the stress
increment in terms of the values of the internal stresses at the start of the interval. With the assumption that the
strain varies linearly during the time interval , we obtain the increment stress-strain relation as
(7-246)
where
(7-247)
and
(7-248)
In Marc, the incremental equation for the total stress is expressed in terms of the short term moduli (See Equation
(7-242)).
(7-249)
In this way, the instantaneous elastic moduli can be specified through the ISOTROPIC or ORTHOTROPIC options.
Moreover, since the TEMPERATURE EFFECTS or TABLE option acts on the instantaneous elastic moduli, it is more
straightforward to use the short term values instead of the long term ones. Note that the set of equations given by
Equation 7-249 can directly be used for both anisotropic and isotropic materials.
Isotropic Viscoelastic Material
For an isotropic viscoelastic material, Marc assumes that the deviatoric and volumetric behavior are fully decoupled
and that the behavior can be described by a time dependent shear and bulk modules. The bulk moduli is generally
assumed to be time independent; however, this is an unnecessary restriction of the general theory.
Both the shear and bulk moduli can be expressed in a Prony series
(7-250)
(7-251)
t
m 1
t
m
, t t
m
t
m 1
=
i j
n
t
i j
t
m
( ) G
i j kl
n
t ( )G
i j kl
n
n 1 =
N
+
kl

n
t ( )
i j
n
t
m
t ( )
n 1 =
N
n
t ( ) 1 exp t ( )
n
( ) =
n
t ( )
n
t ( )
n
t ( ) =
i j
t
m
( ) G
i j kl
0
1
n
t ( ) { }G
i j kl
n
n 1 =
N

kl
t
m
( )
n
t ( )
i j
n
n 1 =
N
= t
m
t ( )
G t ( ) G
G
n
exp t
d
n
( )
n 1 =
N
+ =
K t ( ) K
K
n
exp t
v
n
( )
n 1 =
N
+ =

with short term values given by
(7-252)
(7-253)
Let the deviatoric and volumetric component matrices and be given by
(7-254)
The increment set of equations is then given by
(7-255)
G
0
G
G
n
n 1 =
N
+ =
K
0
K
K
n
n 1 =
N
+ =
d

v
d
4 3 2 3 2 3 0 0 0
2 3 4 3 2 3 0 0 0
2 3 2 3 4 3 0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
=
v
1 1 1 0 0 0
1 1 1 0 0 0
1 1 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
=
t
m
( ) G
0
1
d
n
t ( ) [ ]G
n
n 1 =
N
d
)

`

d
t
m
( ) =
d
n
t ( )
d
n
t
m
t ( )
n 1 =
N
d

v
n
t ( )
v
n
t
m
t ( )
n 1 =
N
v

K
0
1
v
n
t ( ) [ ]K
n
n 1 =
N
v
)

`

v
t
m
( )
483
CHAPTER 7
Material Library
and
(7-256)
Note that the deviatoric and volumetric response are fully decoupled.
The instantaneous moduli need to be given in the ISOTROPIC option. Time dependent values (shear moduli and
time constants ; bulk moduli and time constants ) need to be entered in the VISCELPROP option.
Time-stepping is performed using the TIME STEP or AUTO STEP option in the history definition block.
Note that the algorithm is exact for linear variations of the strain during the increment. The algorithm is implicit;
hence, for each change in time-step, a new assembly of the stiffness matrix is required.
Anisotropic Viscoelastic Material
Equation 7-246 can be used for the analysis of anisotropic viscoelastic materials. The tensor of amplitudes must
be entered through the ORTHOTROPIC option. However, both the and the must be entered using the
VISCELORTH option.
Alternatively, a complete set of moduli (21 components) can be specified in the HOOKVI user subroutine.
The ORIENTATION option or ORIENT user subroutine can be used to define a preferred orientation both for the short
time moduli and the amplitude functions .
Incompressible Isotropic Viscoelastic Materials
Incompressible elements in Marc allow the analysis of incompressible and nearly incompressible materials in plane
strain, axisymmetric and three-dimensional problems. The incompressibility of the element is simulated through the
use of an perturbed Lagrangian variational principle based on the Herrmann formulation.
The constitutive equation for a material with no time dependence in the volumetric behavior can be expressed as
(7-257)
(7-258)
d
n
t
m
( )
d
n
t ( )G
n
d
t
m
( )
d
n
t ( )
d
n
t
m
t ( ) =
v
n
t
m
( )
v
n
t ( )K
n
v
t
m
( )
v
n
t ( )
v
n
t
m
t ( ) =
G
n
d
n
K
n
v
n
G
i j kl
0
G
i j kl
n
n
G
i j kl
0
G
i j kl
n
i j
t
m
( ) 2 G
i j kl
0
1
n
t ( ) [ ]G
i j kl
n
n 1 =
N
)

`

kl
t
m
( )
1
3
---
pp
t
m
( )
kl
=
n
t ( )
i j
( )
n
t
m
( )
n 1 =
N
1
3
-- -
kk
i j
+
pp
t
m
( ) 3K
0
pp
t
m
( ) =

The hydrostatic pressure term is used as an independent variable in the variational principle. The Herrmann pressure
variable is now defined in the same way as in the formulation for time independent elastic materials.
(7-259)
The constitutive Equations (7-257) and (7-258) can then be rewritten
(7-260)
where
(7-261)
(7-262)
Thermo-Rheologically Simple Behavior
The rate processes in many viscoelastic materials are known to be highly sensitive to temperature changes. Such
temperature-dependent properties cannot be neglected in the presence of any appreciable temperature variation. For
example, there is a large class of polymers which are adequately represented by linear viscoelastic laws at uniform
temperature. These polymers exhibit an approximate translational shift of all the characteristic response functions with
a change of temperature, along a logarithmic time axis. This shift occurs without a change of shape. These
temperature-sensitive viscoelastic materials are characterized as Thermo-Rheologically Simple.
A reduced or pseudo time can be defined for the materials of this type and for a given temperature field. This new
parameter is a function of both time and space variables. The viscoelastic law has the same form as one at constant
temperature in real time. If the shifted time is used, however, the transformed viscoelastic equilibrium and
compatibility equations are not equivalent to the corresponding elastic equations.
In the case where the temperature varies with time, the extended constitutive law implies a nonlinear dependence of
the instantaneous stress state at each material point of the body upon the entire local temperature history. In other
words, the functionals are linear in the strains but nonlinear in the temperature.
The time scale of experimental data is extended for Thermo-Rheologically Simple materials. All characteristic
functions of the material must obey the same property. The shift function is a basic property of the material and must
be determined experimentally. As a consequence of the shifting of the mechanical properties data parallel to the time
axis (see Figure 7-64), the values of the zero and infinite frequency complex moduli do not change due to shifting.
Hence, elastic materials with temperature-dependent characteristics neither belong to nor are consistent with the above
hypothesis for the class of Thermo-Rheologically Simple viscoelastic solids.
H

pp
2G
0
1
0
+ ( )
-------------------------------- =
i j
t
m
( ) 2G
e
i j

H
i j
+ ( )
n
t ( )
i j
d
( )
n
t
m
t ( )
n 1 =
N
=
G
e
G
0
1
n
t ( )G
n
[ ]
n 1 =
N
G
0
1
0
+ ( ) G
e
1 2
0
( )
3G
e
------------------------------------------------------------------- =
485
CHAPTER 7
Material Library
Figure 7-64 Relaxation Modulus vs. Time at Different Temperatures
Let be the relaxation modulus as a function of at uniform temperature, . Then
(7-263)
where is measured relative to some arbitrary temperature . The modulus curve shifts towards shorter times
with an increase of temperature; is a positive increasing function for . If denotes the relaxation
modulus as a function of time at uniform temperature , so that,
(7-264)
then
(7-265)
The relaxation modulus (and the other characteristic functions) at an arbitrary uniform temperature is thus expressed
by the base temperature behavior related to a new time scale that depends on that temperature.
There is some mapping of the time coordinate for nonuniform, nonconstant temperature, , which depends on
the position
For a nonuniform, nonconstant temperature, the shift function is and the rate of change of reduced time
can be written as:
Marc offers two explicit forms for entering the shift function. The first is based on the familiar Williams-Landel-Ferry
(WLF) equation. Rewriting the above expression for reduced time as
(7-266)
then the WLF form state that
(7-267)
G
T
T
0
f(T
1
)
ln t
T
1
T
2
f(T
2
)
E ln t ( ) l n t T
E
T
ln t ( ) E
T
0
l n t+f T ( ) [ ]* T
0
T
0
( ) =
f T ( ) T
f T ( ) T T
0
> G
T
t ( )
T
G
T
t ( ) E
T
ln t ( ) =
G
T
t ( ) G
T
0
( ) =
T x t , ( )
a T x t , ( ) ( )
d a
T
T x t , ( ) [ ]dt =
x t , ( )
t d
T
T x t , ( ) [ ]
-------------------------------
0
t
}
=
a
10 T
T ( ) log
C
1
T T
0
( )
C
2
T T
0
( ) +
----------------------------------- - h T ( ) = =

and
(7-268)
Typically, the glassy transition point is taken as the reference temperature in the above relation. The logarithmic shift
can also be expressed in a polynomial expansion about the arbitrary reference point as
(7-269)
Enter the shift function parameters associated with Thermo-Rheologically Simple behavior through the SHIFT
FUNCTION model definition option. As an alternative to the WLF function, Marc allows use of series expansion or
specification via the TRSFAC user subroutine.
In addition to the Thermo-Rheologically Simple material behavior variations of initial stress-strain moduli , the
temperature of the other mechanical properties (coefficient of thermal expansion, etc.) due to changes in temperature
can be specified via the TEMPERATURE EFFECTS option.
Note, however, that only the instantaneous moduli are effected by the TEMPERATURE EFFECTS option. Hence, the
long term moduli given by
(7-270)
can easily become negative if the temperature effects are not defined properly.
Large Strain Viscoelasticity
For an elastomeric time independent material, the constitutive equation is expressed in terms of an energy function
. For a large strain viscoelastic material, Simo generalized the small strain viscoelasticity material behavior to a
large strain viscoelastic material. The energy functional then becomes
(7-271)
where are the components of the Green-Lagrange strain tensor, internal variables and the elastic strain
energy density for instantaneous deformations. In Marc, it is assumed that , meaning that the energy density
for instantaneous deformations is given by the third order James-Green-Simpson form or the Ogden form. When used
with Updated Lagrange, one can also use the Arruda-Boyce or Gent model.
t ( ) 10
h T t ( ) [ ]
dt
0
t
}
=
log
10
A
T
T ( )
i
T T
0
( )
i
i 0 =
m
=
G
i j kl
0
G
i j kl
G
i j kl
0
t ( ) G
i j kl
n
n 1 =
N
=
W
E
i j
Q
i j
n
( )
0
E
i j
( ) Q
i j
n
E
i j
n 1 =
N

I
n
Q
i j
n
( )
n 1 =
N
+ =
E
i j
Q
i j
n
0
W =
487
CHAPTER 7
Material Library
The components of the second Piola-Kirchhoff stress then follow from
(7-272)
The energy function can also be written in terms of the long term moduli resulting in a different set of internal variables
(7-273)
where is the elastic strain energy for long term deformations. Using this energy definition, the stresses are obtained
from
(7-274)
Let the total strain energy be expressed as a Prony series expansion
(7-275)
If, in the energy function, each term in the series expansion has a similar form, Equation (7-275) can be rewritten as
(7-276)
where is a scalar multiplier for the energy function based on the short term values.
The stress-strain relation is now given by
(7-277)
(7-278)
S
i j
E
i j
----------

0
E
i j
---------- Q
i j
n
n 1 =
N
= =
T
i j
n
E
i j
T
i j
n
, ( )
E
i j
( ) T
i j
n
E
i j
n 1 =
N
+ =
S
i j
E ( )
E
i j
-------------------- T
i j
n
n 1 =
N
+ =

n
exp t
n
( )
n 1 =
N
+ =

0
exp t
n
( )
n 1 =
N
+ =
n
S
i j
S
i j
T
i j
n
n 1 =
N
+ =
S
i j

E
i j
----------- 1
n
n 1 =
N
\ .
|
|
| |

0
E
i j
---------- = =

Observing the similarity with the equations for small strain viscoelasticity, the internal variables can be obtained from a
convolution expression
(7-279)
Analogue to the derivation for small strain viscoelasticity, a recursive relation can be derived expressing the stress
increment in terms of values of the internal stresses at the start of the increment.
In Marc, the instantaneous values of the energy function are always given on the MOONEY, OGDEN, or FOAM option,
the equations are reformulated in terms of the short time values of the energy function
(7-280)
(7-281)
It is assumed that the viscoelastic behavior in Marc acts only on the deviatoric behavior. The incompressible behavior
is taken into account using special Herrmann elements.
Large strain viscoelasticity is invoked by use of the VISCELMOON, VISCELOGDEN, or VISCELFOAM option in the
constitutive option of the model definition block. The time dependent multipliers and associated relaxation times
as defined by Equation 7-268 are given in the VISCELMOON, VISCELOGDEN, or VISCELFOAM option. For the
Ogden model, both deviatoric and dilatational relaxation behavior is allowed.
Viscoelasticity can be modeled with Arruda-Boyce and Gent models using VISCELMOON option.
Time-stepping can be performed using the TIME STEP with AUTO LOAD or AUTO STEP option of the history
definition block.
The free energy function versus time data being used for large strain viscoelasticity can be generated by fitting the
experimental data in Mentat or Patran provided the following two tests are done:
1. Standard quasi-static tests (tensile, planar-shear, simple-shear, equi-biaxial tension, volumertic) to determine
the elastomer free energy constants.
2. Standard relaxation tests to obtain stress versus time.
Narayanaswamy Model
The annealing of flat glass requires that the residual stresses be of an acceptable magnitude, while the specification for
optical glass components usually includes a homogenous refractive index. The design of heat treated processes (for
T
i j
n
n
S
i j
0
( )exp t ( )
n
[ ]d
0
t
}
=
S
i j
t
m
( ) 1 1
n
t ( ) [ ]
n 1 =
N

n
\ .
|
|
| |
S
i j
0
t
m
( ) S
i j
0
t
m
t { } =
n
t ( )T
i j
n
t
m
t ( )
n 1 =
N
T
i j
n
t
m
( )
n
t ( )
n
S
i j
0
t
m
( ) S
i j
0
t
m
t ( ) [ ]
n
t ( )T
i j
n
t
m
t ( ) =
n
W
0
489
CHAPTER 7
Material Library
example, annealing) can be accomplished using the Narayanaswamy model. This allows you to study the time
dependence of physical properties (for example, volumes) of glass subjected to a change in temperature.
The glass transition is a region of temperature in which molecular rearrangements occur on a scale of minutes or hours,
so that the properties of a liquid change at a rate that is easily observed. Below the glass transition temperature ,
the material is extremely viscous and a solidus state exists. Above , the equilibrium structure is arrived at easily
and the material is in liquidus state. Hence, the glass transition is revealed by a change in the temperature dependence
of some property of a liquid during cooling. If a mechanical stress is applied to a liquid in the transition region, a
time-dependent change in dimensions results due to the phenomenon of visco-elasticity.
If a liquid in the transition region is subjected to a sudden change in temperature, a time-dependent change in volume
occurs as shown in Figure 7-65. The latter process is called structural relaxation. Hence, structural relaxation governs
the time-dependent response of a liquid to a change of temperature.
Figure 7-65 Structural Relaxation Phenomenon
Suppose a glass is equilibrated at temperature , and suddenly cooled to at . The instantaneous change in
volume is , followed by relaxation towards the equilibrium value . The total change in volume
due to the temperature change is as shown in Figure 7-65b. The rate of volume change depends on a
characteristic time called the relaxation time.
The slope of changes from the high value characteristic of the fluid to the low characteristic of the glass
as shown in Figure 7-66. The glass transition temperature is a point in the center of the transition region. The
low-temperature slope represents the change in volume caused by vibration of the atoms in their potential wells.
T
g
T
g
T
1
T
2
t
0
T(t)
t
(a) Step Input for Temperature
V(0,T
1
)
V(,T
2
)
V(0,T
2
)
t
0
t
g
(T
2
-T
1
)
l
(T
2
-T
1
)
(b) Volume Change as Function of Temperature
T
1
T
2
t
0
g
T
2
T
1
( ) V T
2
, ( )
1
T
2
T
1
( )
dV dT
1
g
T
g
g
V

In the (glassy) temperature range, the atoms are frozen into a particular configuration. As the temperature increases,
the atoms acquire enough energy to break bonds and rearrange into new structures. That allows the volume to increase
more rapidly, so . The difference represents the structural contribution to the volume.
Figure 7-66 Property (Volume) Temperature Plot
When a liquid is cooled and reheated, a hysteresis is observed as shown in Figure 7-67.
Figure 7-67 Volume Change During Cyclic Temperature History
Unfortunately, the notion of a glass transition temperature is insufficient as real glassy materials generally exhibit a
temperature regime, called a transition range, across which their bulk properties gradually change from being solid-
like to liquid-like in nature.
As discussed earlier, properties have a time dependence in the transition range. An explanation for the strong time
dependence lies in that the material resides at a nonequilibrium temperature which lags behind the applied temperature
during the heating-cooling cycle. The nonequilibrium temperature is called the fictive temperature, , as shown in
Figure 7-66. The fictive temperature at is found by extrapolating a line from with slope to
intersect a line extrapolated from with slope (see Figure 7-66). For (well below the glass
transition), reaches a limiting value that is called . If the material were equilibrated at , then
T
1

g
>
1

g
=
V(T)
V(T
1
)
T
2
T
g
T
f
(T
1
)
l
Solidus
State
Liquid
State
Transition Range
T
f
(T
1
): Fictive Temperature
V(T
0
)
T
0
T
1
T
g
T
V
Nonequilibrium
Equilibrium
T
f
T
1
T
f
T
1
( ) , V T
1
( )
g
V T
0
( )
1
T T
2
T
f
T
g
T
f
T
1
( )
491
CHAPTER 7
Material Library
instantaneously cooled to , it would change along the line with slope because no structural rearrangement could
occur. Therefore, it would have the same volume as the continuously cooled sample.
The response of the volume change can be described by:
(7-282)
where is the current value of the fictive temperature. The response function, , which dictates the value of
the fictive temperature is assumed to be linear in its argument and governs both the value of the fictive temperature as
well as the material property of interest.
(7-283)
By virtue of its linearity, Boltzmanns superposition principle can be invoked to calculate the fictive temperature at
any time:
(7-284)
The concept of reduced time, , is introduced in the spirit of Thermo-Rheologically Simple materials to capture
the disparate nonlinear response curves on a single master curve.
The reduced time used in Marc is given by the following expression:
(7-285)
Here is the reference relaxation time of the material evaluated at a suitable reference temperature, . The
relaxation time at the given time and temperature can be represented as:
(7-286)
The parameter allows you to dictate how much of the fictive temperature participates in the prescription of the
relaxation time, and must, therefore, range between 0 and 1.
is the activation energy for the particular process and is the gas constant. A typical response function is:
(7-287)
T
1

g
V T
2
t , ( ) V T
1
, ( )
g
T
2
T
1
( )
l

g
( ) T
f
t ( ) T
2
( ) + + =
T
f
t ( ) M
v
V t ( ) V ( )
V 0 ( ) V ( )
--------------------------------- M
v
t ( ) ( )
T
f
t ( ) T ( )
T 0 ( ) T ( )
--------------------------------- = =
T
f
t ( ) T t ( ) M
v
t ( ) t ( ) ( )
d
dt
------- T t ( ) ( ) t' d ( )

t
}
=
t ( )
t ( )

ref
T t ( ) ( )
--------------------- t d

t
}
=
ref
T
ref

ref
H
R
----
1
T
r ef
----------
x
T
---
1 x ( )
T
f
----------------- -
\ .
| |
exp =
x
H R
M
v
( )

--
\ .
| |
exp =

Multiple structural relaxation times can exist. The response function has, therefore, been implemented as:
For a complete description of the model, it is necessary to prescribe the following:
1. The weight for each term in the series (usually ).
2. The reference relaxation times .
3. The fraction parameter and the activation energy-gas constant ratio.
4. The solid and liquid coefficients of thermal of expansion, and through the VISCEL EXP option.
A stable algorithm is employed to calculate the convolution integrals. For improved accuracy it is recommended that
the time steps used during the simulation be sufficiently small.
In Figure 7-68, the volume of cube of material, which is allowed to contract freely and is experiencing a 100
o
C quench,
is displayed.
Figure 7-68 Volume-Temperature-Time Plot
Temperature Effects and Coefficient of Thermal Expansion
Experimental results indicate that a large number of material properties vary with temperatures. In Marc, almost all
material parameters may be a function of temperatures when using the table driven input procedure. A subset of which
are shown in Table 7-4 for different types of analysis.
Please note that is temperature in the above expressions. With the exception of heat transfer, Joule heating, or
coupled thermal-mechanical analysis, the temperature is a state variable.
M
v
( ) W
g
( )
i
i
----
\ .
| |
exp
i 1 =
n
=
W
g
( )
i
W
g
( )
i
1
i ref ,
x
g

1
Temperature
V
o
l
u
m
e
T
im
e
T=100
g
T
l
T
T
493
CHAPTER 7
Material Library
Piecewise Linear Representation
In Marc, the temperature variation of a material constant may be entered as a piecewise linear function of
temperature as shown in Figure 7-66 or as an equation. The base value or reference value is entered in the ISOTROPIC,
ORTHOTROPIC, ANISOTROPIC, MOONEY, OGDEN, etc. model definition options. Using the nontable input format,
these should be at the lowest temperatures. The variation with temperature is entered on the TEMPERATURE
EFFECTS and ORTHO TEMP or TABLE model definition options.
Table 7-4 Temperature-Dependent Material Properties
Analysis Type Material Properties
Stress Analysis Modulus of elasticity (Youngs Modulus)
Poissons Ratio
Yield Stress
Workhardening Slope
Mooney Constants
Heat Transfer Analysis Thermal Conductivity
Specific Heat
Emissivity
Couples Thermo-Electrical
(Joule Heating Analysis)
Electric Resistivity
Hydrodynamic Heating Viscosity
E T ( )
T ( )
y
T ( )
h T ( )
T ( )
C
01
T ( ) C
10
T ( ) ,
C
11
T ( ) C
20
T ( ) ,
C
30
T ( )
K T ( )
C T ( )
T ( )
T ( )
T ( )
F T ( )

7
Material Library
Figure 7-69 Piecewise Linear Representation of Temperature-Dependent Material Properties
Temperature-Dependent Creep
In Marc, the temperature dependency of creep strain can be entered in two ways. The creep strain rate may be entered
as a piecewise linear function. If the creep strain can be expressed in the form of a power law
(7-288)
where and are two experimental constants, input the experimental constants through the CREEP model
definition option.
For other temperature dependency, you may use an equation or use the CRPLAW user subroutine for explicit creep and
UCRPLW user subroutine for implicit creep to input the variation of creep strain with temperature.
(1) Slope-Break Point Data
Slope Break Point
S
1
= (F
2
- F
1
)/(T
2
- T
1
) T
1
S
2
= (F
3
- F
2
)/(T
3
- T
2
) T
2
S
3
= (F
4
- F
3
)/(T
4
- T
3
) T
3

(2) Function-Variable Data
Function Variable
F
1
T
1
F
2
T
2
F
3
T
3
F
4
T
4
Temperature (T) T
1
Base
Value
F1
Temperature
Dependent
Property
F(T)
F
1
T
2
F
2
T
3
F
3
T
4
F
4
S
1
S
2
S
3
c
AT
m
=
A m
495
CHAPTER 7
Material Library
Marc always uses an instantaneous thermal expansion coefficient definition
(7-289)
or
(7-290)
In many cases, the thermal expansion data is given with respect to a reference temperature
(7-291)
where is a function of temperature:
(7-292)
Clearly, in this case
(7-293)
so the necessary conversion procedure is:
1. Compute and plot Equation (7-291) in the form Equation (7-293)
(7-294)
as a function of temperature.
2. Model Equation (7-294) in the ANEXP user subroutine, or with piecewise linear slopes and breakpoints in the
TEMPERATURE EFFECTS or TABLE option.
The anisotropic coefficient of thermal expansion can be input through either the ORTHOTROPIC model
definition option or the ANEXP user subroutine.
Time-Temperature-Transformation
Certain materials, such as carbon steel, exhibit a change in mechanical or thermal properties when quenched or air
cooled from a sufficiently high temperature. At any stage during the cooling process, these properties are dependent
on both the current temperature and the previous thermal history. The properties are influenced by the internal
microstructure of the material, which in turn depends on the rate at which the temperature changes. Only in instances
where the temperature is changed very gradually does the material respond in equilibrium, where properties are simply
a function of the current temperature. In addition, during the cooling process certain solid-solid phase transformations
can occur. These transformations represent another form of change in the material microstructure which can influence
the mechanical or thermal properties. These transformations can be accompanied by changes in volume.
d
i j
t h
i j
dT = in general
d
i j
t h
dT
i j
= for the isotropic case
t h
= T T
0
( )
a
T ( ) =
d
t h
d
dT
------- T T
0
( ) + dT =

d
dT
------- T T
0
( ) + =

The occurrence of phase change is also dependent on the rate of cooling of the material. This relationship is shown in
a typical cooling diagram (see Figure 7-70). The curves A, B, and C in Figure 7-70 represent the temperature history
of a structure that has been subjected to a different cooling rate. It is obvious that the structural material experiences
phase changes at different times and temperatures, depending on the rate of cooling. Under cooling rate A, the material
changes from phase 1 to phase 4 directly. The material undergoes three phase changes (phase 1 to phase 2 to phase 3
to phase 4) for both cooling rates B and C. However, the phase changes take place at different times and temperatures.
The Time-Temperature-Transformation (TIME-TEMP) option allows you to account for the time-temperature-
transformation interrelationships of certain materials during quenching or casting analyses.
Use the T-T-T parameter to invoke the time-temperature-transformation. Input all the numerical data required for this
option through the TIME-TEMP model definition option.
In a transient heat transfer analysis, the thermal properties which can be defined as a function of time and temperature
are the thermal conductivity and the specific heat per unit reference mass. Here, the effects of latent heat or phase
transformation can be included through the definition of the specific heat.
In a thermal stress analysis, the mechanical properties which can be defined as a function of time and temperature are
the Youngs modulus, Poissons ratio, yield stress, workhardening slope, and coefficient of thermal expansion. The
effects of volumetric change due to phase transformation can be included through the definition of the coefficient of
thermal expansion.
Figure 7-70 Simplified Cooling Transformation Diagram
Test data must be available in a tabular form for each property of each material group. For a given cooling rate, the
value of a property must be known at discrete points over a range of temperatures. There can be several sets of these
discrete points corresponding to measurements at several different cooling rates. The cooling tests must be of a specific
type known as Newton Cooling; that is, the temperature change in the material is controlled such that
(7-295)
Line of Phase Change
Change in Volume
T
e
m
p
e
r
a
t
u
r
e

(
T
)
A
B
C
T
1
Time (t)
T
3
T
2
T
4
T
5
T
1
T
3
T
2
t
1
t
2
T t ( ) A exp at ( ) B + =
497
CHAPTER 7
Material Library
In addition, a minimum and a maximum temperature that bracket the range over which the TIME-TEMP option is
meant to apply must also be given.
For the simulation of the cooling rate effect in finite element analysis, material properties of a structure can be assumed
as a function of two variables: time and temperature. Two-dimensional interpolation schemes are used for the
interpolation of properties.
Interpolation is based on making the time variable discrete. Stress analysis is carried out incrementally at discrete time
stations and material properties are assumed to vary piecewise linearly with temperature at any given time. These
temperature-dependent material properties are updated at each increment in the analysis. For illustration, at time ,
the material is characterized by the phase 1 and phase 4 behaviors at temperature ranges to , and to ,
respectively (see Figure 7-71). Similarly, at time , the material behavior must be characterized by all four phases,
each in a different temperature range (that is, phase 1, to ; phase 2, to ; phase 3, to ; phase 4,
to ). The selection of an interpolation scheme is generally dependent on the form of the experimental data. A linear
interpolation procedure can be effectively used where the properties are expressed as a tabulated function of time
and temperature.
During time-temperature-transformation, the change in volume in a stress analysis is assumed to take place in a
temperature range . The change in volume is also assumed to be uniform in space, such that the effect of the
volume change can be represented by a modification of the coefficient of thermal expansion. For a triangular
distribution of in the temperature range , the value of the modified coefficient of thermal expansion is
(7-296)
where is the change in volume. A schematic of the modified is shown in Figure 7-71.
Figure 7-71 Modified Coefficient of Thermal Expansion for Short-Time Change in Volume
t
1
T
1
T
3
T
3
T
2
t
2
T
1
T
4
T
4
T
5
T
5
T
3
T
3
T
2
T
T ( ) T
m
2
T
------- 1 +
3
1 [ ] =
T ( )
Temperature (T)
C
o
e
f
f
i
c
i
e
n
t

o
f

t
h
e
r
m
a
l

E
x
p
a
n
s
i
o
n

(
)
T
m

Low Tension Material
Marc can handle concrete and other low tension material. The CRACK DATA option assists in predicting crack
initiation and in simulating tension softening, plastic yielding and crushing. This option can be used for the following:
Elements with a one-dimensional stress-strain relation (beam and truss elements)
Elements with a two-dimensional stress-strain relation (plane stress, plane strain, axisymmetric, and
shell elements)
Three-dimensional elements (bricks)
Analytical procedures that accurately determine stress and deformation states in concrete structures are complicated
by several factors. Two such factors are the following:
The low strength of concrete in tension that results in progressive cracking under increasing loads
The nonlinear load-deformation response of concrete under multiaxial compression
Because concrete is mostly used in conjunction with steel reinforcement, an accurate analysis requires consideration
of the components forming the composite structure. Steel reinforcement bars are introduced as rebar elements.
Each rebar element must be input with a separate element number. The REBAR model definition option or REBAR
user subroutine is used to define the orientation of the reinforcement rods.
Uniaxial Cracking Data
The cracking option is accessed through the ISOTROPIC option. Uniaxial cracking data can be specified using the
CRACK DATA option or the UCRACK user subroutine. When the CRACK DATA option is used to specify uniaxial
cracking data, the following must be specified: the critical cracking stress, the modulus of the linear strain softening
behavior, and the strain at which crushing occurs. Material properties, such as Youngs modulus and Poissons ratio,
are entered using the ISOTROPIC option and the WORK HARD option. This model is for a material which is initially
isotropic; if the model is initially orthotropic, see FAIL DATA for an alternative cracking model. A typical uniaxial
stress-strain diagram is shown in Figure 7-72.
499
CHAPTER 7
Material Library
Figure 7-72 Uniaxial Stress-Strain Diagram
Low Tension Cracking
A crack develops in a material perpendicular to the direction of the maximum principal stress if the maximum principal
stress in the material exceeds a certain value (see Figure 7-73). After an initial crack has formed at a material point, a
second crack can form perpendicular to the first. Likewise, a third crack can form perpendicular to the first two. The
material loses all load-carrying capacity across the crack unless tension softening is included.
Tension Softening
If tension softening is included, the stress in the direction of maximum stress does not go immediately to zero; instead
the material softens until there is no stress across the crack. At this point, no load-carrying capacity exists in tension
(see Figure 7-73). The softening behavior is characterized by a descending branch in the tensile stress-strain diagram,
and it may be dependent upon the element size.
E Youngs Modulus
E
s
Tension-Softening Modulus
y
Yield Stress
cr
Critical Cracking Stress
crush
Crushing Strain
Workhardening
crush
cr
E
s
E

Figure 7-73 Crack Development
Crack Closure
After a crack forms, the loading can be reversed; therefore, the opening distance of a crack must be considered. In this
case, the crack can close again, and partial mending occurs. When mending occurs, it is assumed that the crack has
full compressive stress-carrying capability and that shear stresses are transmitted over the crack surface, but with a
reduced shear modulus.
Crushing
As the compressive stress level increases, the material eventually loses its integrity, and all load-carrying capability is
lost; this is referred to as crushing. Crushing behavior is best described in a multiaxial stress state by a crushing surface
having the same shape as the yield surface. The failure criterion can be used for a two-dimensional stress state with
reasonable accuracy. For many materials, experiments indicate that the crushing surface is roughly three times larger
than the initial yield surface.
Analysis
The evolution of cracks in a structure results in the reduction of the load carrying capacity. The internal stresses need
to be redistributed through regions that have not failed. This is a highly nonlinear problem and can result in the ultimate
failure of the structure. The AUTO INCREMENT or AUTO STEP option should be used to control the applied load on
the structure.
Soil Model
Soil material modeling is considerably more difficult than conventional metals, because of the nonhomogeneous
characteristics of soil materials. Soil material usually consists of a large amount of random particles. Soils show unique
properties when tested. The bulk modulus of soil increases upon pressing. Also, when the preconsolidation stress is
exceeded, the stiffness reduces dramatically while the stiffness increases upon unloading. At failure, there is no
resistance to shear, and stiff clays or dense sands are dilatant. Over the years, many formulations have been used,
including linear elastic, nonlinear elastic, Drucker-Prager or Mohr Coulomb, and Cam-Clay and variations thereof. In
Marc, the material models for available soil modeling are linear elasticity, nonlinear elasticity, and the
Cam-Clay model.
y
x
2
501
CHAPTER 7
Material Library
Elastic Models
Linear elasticity is defined in the conventional manner, defining the Youngs moduli and the Poissons ratio in the SOIL
option. The nonlinear elasticity model is implemented through the NLELAST option or the HYPELA2 user subroutine.
Some of the simplest models include the bilinear elasticity, where a different moduli is used during the loading and
unloading path, or to represent total failure when a critical stress is obtained.
A more sophisticated elastic law is the hyperbolic model, where six constants are used. In this model, the tangent
moduli are
(7-297)
Two of the elastic models are the E- and K-G variable elastic models. In the E- model, Poissons ratio is considered
constant and
(7-298)
while, in the K-G model
(7-299)
(7-300)
The key difficulty with the elastic models is that dilatancy cannot be represented.
Cam-Clay Model
The Cam-Clay model was originally developed by Roscoe, and then evolved into the modified Cam-Clay model of
Roscoe and Burland. This model, which is also called the critical state model, is implemented in Marc.
The yield surface is an ellipse in the p, plane as shown in Figure 7-74, and is defined by
(7-301)
where is the preconsolidation pressure, and is the slope of the critical state line.
The Cam-Clay model has the following no properties. At the intersection of the critical state line and the ellipse, the
normal to the ellipse is vertical. Because an associated flow rule is used, all plastic strain at failure is distortional; the
soil deforms at constant volume (Figure 7-74). The strain hardening and softening behavior are shown in Figure 7-75.
Also, if the preconsolidation pressure is large, the soil remains elastic for large stresses. The evolution of the
preconsolidation pressure is
(7-302)
where
(7-303)
E 1
R
f
1 sin ( )
1

3
( )
2c cos 2
3
sin +
---------------------------------------------------------
2

3
pa
------
\ .
| |
n
=
E E
0

E
p
E
+ + =
K K
0

K
p + =
G G
0

G
p
G
+ + =
F

2
M
cs
2
---------- 2pp
c
p
2
+ 0 = =
p
c
M
cs
p
c
p
c
tr
pl
( ) =
1 e +

------------- =

where
Figure 7-74 Modified Cam-Clay Yield Surface
Figure 7-75 Strain Hardening and Softening Behavior
Figure 7-76 Response of Idealized Soil to Hydrostatic Pressure
is the void ratio
is the virgin compression index (see Figure 7-76)
is the recompression index (see Figure 7-76)
e
Critical State Line

Strain Hardening
Ceases
Strain Softening
Ceases
Region
originally
elastic
1
2
3
4
(
q
)
P
CO
v
( )p
q
1
2
q
2
q
q
3
4
q
4
q
1
1
e
p ( ) ln
503
CHAPTER 7
Material Library
The void ratio and the porosity are related by the expression
(7-304)
In the modified Cam-Clay model, it is also assumed that the behavior is nonlinear elastic with a constant Poissons
ratio, the bulk modulus behave as:
(7-305)
Note that this implies that at zero hydrostatic stress, the bulk modulus is also zero. To avoid computational difficulties,
a cutoff pressure of one percent of the preconsolidation pressure is used.
This constitutive law is implemented in Marc using a radial return procedure. It is available for either small or large
strain analysis.When large displacements are anticipated, you should use the LARGE STRAIN parameter.
The necessary parameters for the Cam-Clay soil model can be obtained by the following experiments:
1. Hydrostatic test: determines volumetric elastic bulk modulus, yield stress, and the virgin and recompression
ratio of soils.
2. Shear-box test: determines the slope of critical state line and shear modulus of the soil. However, the tests have
to be calibrated with numerical simulations to get the necessary constants.
3. Triaxial shear test: the most comprehensive experimental information and obviates the need for the first two
tests and obtains all the necessary constants listed above.
Evaluation of Soil Parameters for the Critical State Soil Model
To illustrate how to extract soil parameters, namely , , and , a hypothetical data set for a normally consolidated
clay in presented in Figures 7-77 to 7-83. Now we shall use these data to show the procedure of determination of
parameters for the critical state model. The data presented here pertain to conventional triaxial conditions.
Figures 7-77 and 7-78 show the stress-strain relations for constant pressure tests under different initial conditions
p
0
=10 (69), 20 (139), and 30 (207) psi (kPa), respectively. Here the mean effective pressure of the soil sample is kept
constant throughout the test. The last data point shown on the deviatoric stress (q) versus axial strain (
1
) plot for any
test is considered as the ultimate condition for that test. The void ratio values at the beginning and the end of each test
are given in Table 7-5.
Figures 7-80 to 7-82 show stress-strain relation plots for three fully drained tests performed at initial pressures 10, 20,
and 30 psi. Here, there is no pore pressure development, and the effective mean pressure increases during the test. The
void ratio values at the beginning and end of each test are given in Table 7-5.
Figures 7-83 to 7-85 show stress-strain behavior under undrained conditions. Here there is no change in volume of the
sample. However, fluid pore pressure are developed during the test, thereby reducing the effective stresses until the
ultimate (failure) state is reached.
Figure 7-86 shows the variation of void ratio with mean pressure which is obtained from a hydrostatic compression
(HC) test with three load reversals (that is, unloading-reloading cycles). In this figure, void ratio is plotted to a linear
scale while the hydrostatic stress is plotted to a logarithmic (base 10) scale. In fact, this is the usual way of presenting
one-dimensional consolidation or hydrostatic test results in geotechnical engineering practices.
e
1 e +
------------ =
K
1 e +
------------ p =
M

Determination of parameter, M: The parameter M is the slope of the critical state line on a p-q plot. To determine its
value, the values of p and q at ultimate conditions for each test are plotted as shown in Figure 7-87; the ultimate
condition for each test is assumed to be the last point plotted in Figures 7-77 to 7-85. At the ultimate conditions, the
sample undergoes excessive deformations under constant deviatoric stress, and hence it could be taken as the
asymptotic stress to the curve. The slope of the critical state line (Figure 7-87) is calculated as 1.0 for the soil above.
That is, .
Table 7-5 Test Values for Cam Clay Model
Test
Initial
Pressure, p
0
(effective)
(psi)
Initial
void Ratio,
e
0
Final
Pressure, p
f
(effective)
(psi)
Final
Void Ratio,
e
1
p-constant 10 1.080 10 0.980
p-constant 20 0.959 20 0.860
p-constant 30 0.889 30 0.787
Drained 10 1.080 15 0.908
Drained 20 0.959 30 0.787
Drained 30 0.889 45 0.716
Undrained 10 1.080 05.55 1.080
Undrained 20 0.959 11.09 0.959
Undrained 30 0.889 16.64 0.889
M 1.0 =
505
CHAPTER 7
Material Library
Figure 7-77 Constant Pressure Test: p
0
= 10 psi
0
= 20 psi
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,
q
=
(
p
s
i
)
10
8
6
4
2
0 0.10 0.20
Axial strain,
1
0.05
0.04
0.03
0.02
0.01
0.10 0.20
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n
,
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,
q
=
(
p
s
i
)
25
20
15
10
5
0 0.10 0.20
Axial strain,
1
0.05
0.04
0.03
0.02
0.01
0.10 0.20
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n
,

v

0
= 30 psi
Figure 7-80 Drained Test: p
0
= 10 psi
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

q
=
(
p
s
i
)
30
20
15
10
5
0.10 0.20
Axial strain,
1
0.05
0.04
0.03
0.02
0.01
0.10 0.20
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n
,
25
=
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

(
p
s
i
)
16
8
6
4
2
0.10 0.20
Axial strain,
1
0.10
0.08
0.06
0.04
0.02
0.10 0.20
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n
,
10
12
14
507
CHAPTER 7
Material Library
0
= 20 psi
0
= 30 psi
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

(
p
s
i
)
20
15
10
5
0.10 0.20
Axial strain,
1
0.10
0.08
0.06
0.04
0.02
0.10 0.20
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n
,
25
30
=
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

q
=
(
p
s
i
)
40
30
20
10
0.10 0.20
Axial strain,
1
0.10
0.08
0.06
0.04
0.02
0.10 0.20
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n
,
50

Figure 7-83 Undrained Test: p
0
= 10 psi
0
= 20 psi
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

q
=
(
p
s
i
)
4
3
2
1
0.01 0.02
Axial strain,
1
P
o
r
e

p
r
e
s
s
u
r
e
,

u

(
p
s
i
)
5
6
0.03 0.04
4
3
2
1
0.01 0.02
5
6
0.03 0.04
Axial strain,
1
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

(
p
s
i
)
8
6
4
2
0.01 0.02
Axial strain,
1
P
o
r
e

p
r
e
s
s
u
r
e
,

u

(
p
s
i
)
10
12
0.03 0.04
8
6
4
2
0.01 0.02
10
0.03 0.04
Axial strain,
1
q
=
509
CHAPTER 7
Material Library
0
= 30 psi
Figure 7-86 Hydrostatic Compression Test: C
c
= 0.04, C
s
= 0.06
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

(
p
s
i
)
8
6
4
2
0.01 0.02
Axial strain,
1
P
o
r
e

p
r
e
s
s
u
r
e
,

u

(
p
s
i
)
10
16
0.03 0.04
14
12
10
8
0.01 0.02
16
0.03 0.04
Axial strain,
1
0
12
14
6
4
2
q
=
V
o
i
d

r
a
t
i
o
,

e
1.0
0.9
0.8
20 30
Pressure, p (psi) (logarithmic scale with base 10)
A
40 50
0.7
1.1
1.2
C
c
C
s
10
B
C

Figure 7-87 Critical State Line in q-p (psi)
Determination of parameters and : The values of gamma and kappa can be related to the commonly known
quantities such as compression index ( ) and swelling index ( ). The compression index, , is defined as the
slope of virgin loading line on e-log
10
p plot while the swelling index, , is defined as the unloading-reloading curves
on the same plot. Usually, the compression index and swelling index are defined with respect to a one-dimensional
consolidation test. However, it can be shown that the e-ln(p) curve for any constant stress ratio test, that is, for constant
q/p ratio, is parallel to that obtained from a hydrostatic test, (Figure 7-88).
Figure 7-88 Isotropic Consolidation (Data from 5-8)
D
e
v
i
a
t
o
r
i
c

s
t
r
e
s
s
,

q

(
p
s
i
)
30
20
10
50 20
Mean Pressure, p (psi)
30 40
0
40
45
10
Slope = M = 1.0
Drained
Constant pressure
Undrained
M
C
c
C
s
C
c
C
s
V
o
i
d

r
a
t
i
o
,

e
1.0
0.9
0.8
60 30
Pressure (psi) (logarithmic scale with base 10)
40 50
0.7
1.1
1.2
20
Hydrostatic loading
At critical state
70 80 90
e
0
vs. p
0
e
f
vs. p
f
511
CHAPTER 7
Material Library
In fact, one-dimensional hydrostatic test is parallel to that obtained under critical state conditions. The values of
gamma and kappa can be related to and as follows. The virgin compression line can be expressed as
(7-306)
or
(7-307)
and the swelling line (unloading-reloading) can be expressed as
(7-308)
or
(7-309)
Therefore, comparing Equations (7-306) and (7-307), we have
(7-310)
and comparing Equations (7-308) and (7-309) yields
(7-311)
The value of can be computed by considering two points, A and B, in Figure 7-86 as
(7-312)
Here the subscript denotes the value at that point. The swelling index, , can be computed by considering points B
and C of the same figures
(7-313)
C
c
C
s
e e
0
C
c
log
10
p
p
0
-----
\ .
| |
=
e e
0

p
p
0
-----
\ .
| |
ln =
e e
0
C
s
log
10
p
p
0
-----
\ .
| |
=
e e
0

p
p
0
-----
\ .
| |
ln =
C
c
10 ln
-------------
C
c
2.303
------------- = =
C
s
10 ln
-------------
C
s
2.303
------------- = =
C
c
C
c
e
A
e
B
log p
B
log p
A
-------------------------------------- =
1.08 0.08
log 50 log 10
------------------------------------- =
0.40 =
C
s
C
s
e
C
e
B
log p
B
log p
C
-------------------------------------- =
0.842 0.80
log 50 log 10
------------------------------------- =
0.06 =

Hence, the values of gamma and kappa can be computed from Equations (7-310) and (7-311) as
(7-314)
Damage Models
In many structural applications, the finite element method is used to predict failure. This is often performed by
comparing the calculated solution to some failure criteria, or by using classical fracture mechanics. Previously, we
discussed two models where the actual material model changed due to some failure, see Progressive Composite
Failure on page 379 and the previous section on Low Tension Material on page 498. In this section, the damage
models appropriate for ductile metals and elastomeric materials will be discussed.
Ductile Metals
In ductile materials given the appropriate loading conditions, voids will form in the material, grow, then coalesce,
leading to crack formation and potentially, failure. Experimental studies have shown that these processes are strongly
influenced by hydrostatic stress. Gurson studied microscopic voids in materials and derived a set of modified
constitutive equations for elastic-plastic materials. Tvergaard and Needleman modified the model with respect to the
behavior for small void volume fractions and for void coalescence.
In the modified Gurson model, the amount of damage is indicated with a scalar parameter called the void volume
fraction f. The yield criterion for the macroscopic assembly of voids and matrix material is given by:
(7-315)
as seen in Figure 7-89.
Figure 7-89 Plot of Yield Surfaces in Gurson Model
The parameter was introduced by Tvergaard to improve the Gurson model at small values of the void volume
fraction. For solids with periodically spaced voids, numerical studies [Ref. 10] showed that the values of
and were quite accurate.
0.40
2.303
------------- 0.174 = =
F

y
------
\ .
| |
2
2q
1
f
q
2
kk
2
y
---------------
\ .
|
| |
1 q
1
f
( )
2
+ [ ] cosh + 0 = =
1.0
0.5
0
0 1 2 3 4
0.9
0.6
0.3
0.1
f* f
u
*
0.01
=
f* 0
e

M
kk
3
M
q
1
q
1
1.5 =
q
2
1 =
513
CHAPTER 7
Material Library
The evolution of damage as measured by the void volume fraction is due to void nucleation and growth. Void
nucleation occurs by debonding of second phase particles. The strain for nucleation depends on the particle sizes.
Assuming a normal distribution of particle sizes, the nucleation of voids is itself modeled as a normal distribution in
the strains, if nucleation is strain controlled. If void nucleation is assumed to be stress controlled in the matrix, a normal
distribution is assumed in the stresses. The original Gurson model predicts that ultimate failure occurs when the void
volume fraction f, reaches unity. This is too high a value and, hence, the void volume fraction f is replaced by the
modified void volume fraction in the yield function.
The parameter is introduced to model the rapid decrease in load carrying capacity if void coalescence occurs.
(7-316)
where is the critical void volume fraction, and is the void volume at failure, and . A safe choice
for would be a value greater than namely, . Hence, you can control the void volume
fraction, , at which the solid loses all stress carrying capability.
Numerical studies show that plasticity starts to localize between voids at void volume fractions as low as 0.1 to 0.2.
You can control the void volume fraction , beyond which void-void interaction is modeled by Marc. Based on the
classical studies, a value of can be chosen.
The existing value of the void volume fraction changes due to the growth of existing voids and due to the nucleation
of new voids.
(7-317)
The growth of voids can be determined based upon compressibility of the matrix material surrounding the void.
(7-318)
As mentioned earlier, the nucleation of new voids can be defined as either strain or stress controlled. Both follow a
normal distribution about a mean value.
In the case of strain controlled nucleation, this is given by
(7-319)
where is the volume fraction of void forming particles, the mean strain for void nucleation and the standard
deviation.
f
f =
if f f
c
f
f
c
f
u
*
f
c
f
F
f
c
----------------
\ .
|
| |
f f
c
( ) + =
if f > f
c
f
c
f
F
f
u
*
1 q
1
=
f
F
1 q
1
( ) f
F
1.1 q
1
=
f
F
f
c
f
c
0.2 =
f
gr owt h
f
nucl eat i on
+ =
f
gr owt h
1 f ( )
kk
p
=
f
nucl eat i on
f
N
S 2
-------------- exp
1
2
---

m
p
S
-------------------
\ .
| |
2

m
p
=
f
N

n
S

7
Material Library
In the case of stress controlled nucleation, the rate of nucleation is given by:
(7-320)
where .
If the second phase particle sizes in the solid are widely varied in size, the standard deviation would be larger than in
the case when the particle sizes are more uniform. The Marc user can also input the volume fraction of the nucleating
second phase void nucleating particles in the input deck, as the variable .
A typical set of values for an engineering alloy is given by Tvergaard for strain controlled nucleation as
. (7-321)
It must be remarked that the determination of the three above constants from experiments is extremely difficult. The
modeling of the debonding process must itself be studied including the effect of differing particle sizes in a matrix. It
is safe to say that such an experimental study is not possible. The above three constants must necessarily be obtained
by intuition keeping in mind the meaning of the terms.
When the material reaches 90 percent of , the material is considered to be failed. At this point, the stiffness and the
stress at this element are reduced to zero.
Elastomers
Under repeated application of loads, elastomers undergo damage by mechanisms involving chain breakage, multi-
chain damage, micro-void formation, and micro-structural degradation due to detachment of filler particles from the
network entanglement. Two types of phenomenological models namely, discontinuous and continuous, exists to
simulate the phenomenon of damage.
1. Discontinuous Damage:
The discontinuous damage model simulates the Mullins effect as shown in Figure 7-90.
Figure 7-90 Discontinuous Damage
f
nucl eat i on
f
N
S 2
--------------
1
2
---

n
S
----------------
\ .
| |
2
exp =

1
3
---
kk
+ =
f
N
n
0.30 ; f
N
0.04 ; S 0.01 = = =
f
F
515
CHAPTER 7
Material Library
This involves a loss of stiffness below the previously attained maximum strain. The higher the maximum
attained strain, the larger is the loss of stiffness. Upon reloading, the uniaxial stress-strain curve remains
insensitive to prior behavior at strains above the previously attained maximum in a cyclic test. Hence, there is
a progressive stiffness loss with increasing maximum strain amplitude. Also, most of the stiffness loss takes
place in the few earliest cycles provided the maximum strain level is not increased. This phenomenon is found
in both filled as well as natural rubber although the higher levels of carbon black particles increase the
hysteresis and the loss of stiffness. The free energy, , can be written as:
(7-322)
where is the nominal strain energy function, and
(7-323)
determines the evolution of the discontinuous damage. The reduced form of Clausius-Duhem dissipation
inequality yields the stress as:
(7-324)
where is the deformation tensor.
Mathematically, the discontinuous damage model has a structure very similar to that of strain space plasticity.
Hence, if a damage surface is defined as:
(7-325)
The loading condition for damage can be expressed in terms of the Kuhn-Tucker conditions:
(7-326)
The consistent tangent can be derived as:
(7-327)
2. Continuous Damage:
The continuous damage model can simulate the damage accumulation for strain cycles for which the values of
effective energy is below the maximum attained value of the past history as shown in Figure 7-91.
This model can be used to simulate fatigue behavior. More realistic modeling of fatigue would require a
departure from the phenomenological approach to damage. The evolution of continuous damage parameter is
governed by the arc length, , of the effective strain energy as:
(7-328)
W
W K , ( )W
0
=
W
0
max W
0
( ) =
S 2K , ( )
W
0
C
----------- =
C
W 0 =
0
0 =
D 4 K
2
W
0
CC
---------------
K
W
0
-----------
W
0
C
-----------
W
0
C
----------- + =
s

s
----- W
0
s ( ) s d
0
t
}
=

Hence, accumulates continuously within the deformation process.
Figure 7-91 Continuous Damage
The Kachanov factor is implemented in Marc through both an additive as well as a multiplicative
decomposition of these two effects as:
(7-329)
(7-330)
You specify the phenomenological parameters and . If is not defined, it is
automatically determined such that, at zero values of and , the Kachanov factor . If, according to
Equation (7-328) or Equation (7-329) the value of exceeds 1, is set back to 1.
The above damage model is available for deviatoric behavior and is flagged by means of the OGDEN and DAMAGE
model definition options. If, in addition, viscoelastic behavior is desired, the VISCELOGDEN option can be included.
Finally, the UELDAM user subroutine can be used to define damage functions different from Equations (7-326) to
(7-329).
The parameters required for the continuous or discontinuous damage model can be obtained using the experimental
data fitting option in Mentat and Patran.
Cohesive Zone Modeling
Marc has a library of so-called interface elements (186, 187, 188, 189, 190, 191, 192,and 193), which can be used to
simulate the onset and progress of delamination. The constitutive behavior of these elements is expressed in terms of
tractions versus relative displacements between the top and bottom edge/surface of the elements (see Figure 7-92).
K , ( )
K , ( ) d
d
n
n
------
\ .
| |
d
n
n
-----
\ .
| |
exp
n 1 =
2
+ exp
n 1 =
2
+ =
K , ( ) d
d
n

n
+
n
--------------------
\ .
| |
exp
n 1 =
2
+ =
d
n
d
n
n

n
d
n

n
, , , , , d
K 1 =
K K
517
CHAPTER 7
Material Library
Figure 7-92 3-D Linear Interface Element
Considering a 3-D interface element, the relative displacement components are given by one normal and two shear
components, expressed with respect to the local element system (see Marc Volume B: Element Library for the
definition of the local element systems):
(7-331)
Based on the relative displacement components, the effective opening displacement is defined as:
(7-332)
Later on, some modifications of this definition will be discussed.
The effective traction is introduced as a function of the effective opening displacement and is characterized by an
initial reversible response followed by an irreversible response as soon as a critical effective opening displacement
has been reached. The irreversible part is characterized by increasing damage ranging from 0 (onset of delamination)
to 1 (full delamination). This damage value is available for postprocessing using element post code 80, Damage.
Three standard functions are currently available; namely, a bilinear, an exponential, and a linear-exponential function
(see Figure 7-93):
(7-333)
2
4
3
v
2
v
1
5
6
7
8
v
3
top face
bottom face
1
v
n
u
1
t op
u
1
bot t om
=
v
s
u
2
t op
u
2
bot t om
=
v
t
u
3
t op
u
3
bot t om
=
v v
n
2
v
s
2
v
t
2
+ + =
t
v
c
t
2G
c
v
m
----------
v
v
c
----- =
if 0 v v
c

t
2G
c
v
m
----------
v
m
v
v
m
v
c
-------------------
\ .
|
| |
=
if v
c
v < v
m
t 0 = if v v
m
>
Bilinear

(7-334)
(7-335)
in which is the energy release rate (cohesive energy), is the maximum effective opening displacement (which
is only used by the bilinear model), and is the exponential decay factor (which is only used by the linear-exponential
model).
Figure 7-93 Bilinear (left), Exponential (middle), and Linear-exponential (right)
Cohesive Material Model
It can easily be verified that the maximum effective traction , corresponding to the critical effective opening
displacement is given by:
Bilinear (7-336)
Exponential (7-337)
Linear-exponential (7-338)
t G
c
v
v
c
2
----- e
v v
c
=
Exponential
t
2qG
c
v
c
q 2 + ( )
------------------------
v
v
c
----- =
if 0 v < v
c
Linear-exponential
t
2qG
c
v
c
q 2 + ( )
------------------------ e
q 1 v v
c
( )
=
if v v
c
>
G
c
v
m
q
v v
t t
v
c
v
c
v
m
t
v
c
v
t
c
v
c
t
c
2G
c
v
m
---------- =
t
c
G
c
ev
c
-------- =
t
c
2qG
c
v
c
q 2 + ( )
------------------------ =
519
CHAPTER 7
Material Library
So if the maximum effective traction is known, the critical or maximum effective opening displacement can be
determined by:
Bilinear (7-339)
Exponential (7-340)
Linear-exponential (7-341)
Note that for the bilinear model, the critical effective opening displacement does neither affect the cohesive energy nor
the maximum effective traction.
Until now, the behavior in the normal and shear direction is treated similarly. However, sometimes the behavior of an
interface material may be different in tension and shear. The first method to include such differences is incorporated
by the shear-normal stress ratio , which defines the ratio of the maximum stress in shear and the maximum stress
in tension [Ref. 26]. This ratio is used to redefine the effective opening displacement according to:
The effect of is depicted in Figure 7-94 for the bilinear model.
Figure 7-94 Response in Tension and Shear for a Shear-normal Stress Ratio (Bilinear Model)
Although the use of the shear-normal stress ratio offers some flexibility, it assumes that the cohesive energy in tension
and shear is the same. If one wants to define a different value of the cohesive energy in shear than in tension, the shear-
normal energy ratio can be used. In a general state of deformation, when , the curve defining the effective
traction versus the effective opening displacement is defined as a linear combination of the response in pure tension
and pure shear. Using and , Figure 7-95 shows the response in tension and shear for the bilinear
model.
v
m
2G
c
t
c
---------- =
v
c
G
c
et
c
------- =
v
c
2qG
c
t
c
q 2 + ( )
---------------------- =
1
v v
n
2
1
2
v
s
2
1
2
v
t
2
+ + =
1
0.5 =
v
t
v
c
v
m
v
t
v
c
v
m
tension
only
shear
only
G
c
G
c
1
0.5 =
2

2
1
1
0.5 =
2
0.75 =

Figure 7-95 Response in Tension and Shear for a Shear-normal Stress Ratio and a Shear-normal Energy
Ratio (Bilinear Model)
In order to avoid convergence problems in a finite element simulation of delamination, one may activate so-called
viscous energy dissipation. The basic idea of the dissipation model is that when delamination starts, the rate of
deformation may suddenly increase. This increase is used to augment the constitutive behavior with a viscous
contribution being equivalent to this rate of deformation:
(7-342)
in which is the viscous energy factor, is effective opening displacement rate and is the reference value of the
effective opening displacement rate. This reference value can either be user-defined or calculated by the program. In
the latter case, the reference value is given by the maximum effective opening displacement rate in any interface
element, as long as the response in all the interface elements is reversible. The viscous energy dissipation model does
not directly have a physical background, but is basically numerical in nature.
In the equations discussed above, no distinction has been made between tensile and compressive loading in the normal
direction. Assuming that in compression the behavior will remain reversible, Equation (7-343) will be adapted as:
(7-343)
Since, irrespective of the damage level, the interface elements should be able to sustain ongoing loading in
compression (so that inter-penetration is prohibited), it is possible to make the stiffness in compression a function of
the corresponding (negative) opening displacement. By default, the stiffness in compression is constant and given by
the slope of the traction versus opening displacement curve at the origin. If a non-default value of the stiffening factor
in compression is given, the stiffness at is given by:
v
t
v
c
v
m
v
t
v
c
v
m
tension
only
shear
only
0.75G
c
G
c
1
0.5 =
2
0.75 =
t
vi s
t
c
v
r
---------- =
v
0
v max v
n
0 , ( ) [ ]
2
v
s
2
v
t
2
+ + =
F v v
c
=
t
n
v
n
--------
v
c
F
t
n
v
n
--------
0
=
521
CHAPTER 7
Material Library
So far, the constitutive behavior has been discussed in terms of an effective traction versus an effective opening
displacement. The traction components follow from the effective traction according to:
; (7-344)
As an alternative to the above mentioned standard models, the UCOHESIVE user subroutine can be used to enter a
user-defined material behavior.
There is a corresponding thermal cohesive model to what was discussed above which uses heat transfer cohesive
elements. In this model, the thermal conductivity can be a function of the displacement opening. The thermal gradient
through the thickness is based upon the temperatures along the top and bottom surfaces. The UCOHESIVET user
subroutine may be used to provide a more complex thermal conductivity.
Lemaitre Model
The Lemaitre damage model is a phenomenological approach to ductile damage in ferrous materials that are subject
to large plastic deformations as they occur in the manufacturing processes. The model is based on the thermodynamic
dissipation potential of the material where ductile damage is considered as a specific energy that is released when
macroscopic fracture occurs. Without going further into detail the mathematical derivation can be found in: Lemaitre,
J.: A Course on Damage Mechanics, 2nd Ed., Springer Verlag, Berlin, 1996. In the following paragraphs, some basic
formulas are given that serve for the calculation and the interpretation of the damage values. The material parameters
can be derived by uniaxial tensile tests for the assumed forming conditions (strain rate, temperature). Some standard
values are given in this context, more information can be found in the mentioned literature.
The Lemaitre damage model calculates three damage values which have different meanings. Macroscopic damage is
characterized by plastic deformation that leads to pore growth, pore coalescence and final rupture of the material
matrix. The damage growth begins approximately after an equivalent plastic strain threshold, . This is the first
material parameter to be defined in an experiment. For mild steels, it is assumed to be between 0.1 and 0.2. The so-
called absolute damage represents the ductile damage growth in the material. The incremental damage law is given
as follows:
,
where the triaxiality function contains information about the state of stresses and is defined as follows:
is the von Mises stress, the Youngs modulus, the Poissons ratio, the mean normal stress, D the current
integrated value of the absolute damage at that material point and the effective plastic strain increment. For most
steels one can assume a maximum value of from 0.15 to 0.4 at fracture. Copper might even reach . The
more ductile the material, the higher becomes. is called the damage resistance factor, a material parameter to be
determined from tensile tests. is from 1 to 8 according to the ductility of the material (1 low, 8 high). This parameter
t
n
t
v
v
n
--------- = t
s t ,
t
v
v
s t ,
----------- =
d
D
dD
f ( )
2
2E S 1 D ( )
2

------------------------------------------- d
p
= 0 D 1
f ( )
f ( )
2
3
--- 1 v + ( ) 3 1 2v ( )
2
+ =

m
------- =
E v
m
d
p
D D 0.9 =
D S
S

influences mostly the growth of and can also be determined by data correlation (for example, simulate the material
test then derive the correct material parameter).
The critical damage is used to compare the ductile damage with the state of the material; such as, whether
the actual conditions (stresses, strains, state of stresses, already reached damage, etc.) might be critical for macroscopic
failure:
is the critical damage in the uniaxial loadcase, the third material parameter to be determined by tensile tests. Most
steels show a from 0.15 to 0.4. is the ultimate stress during the tensile test (before necking begins). The lower
, the more likely a material damage is. Note that behaves contrary to : for the safe material and
low for possible damaged regions.
In fact the comparison between and is necessary to identify critical forming zones: as long as is (much) lower
than , the forming operation is safe. When reaches , the damage probability tends to be 100%. The
comparison is done by the relative damage value (reflected in the postprocessing):
When approaches 1 in a specific region, fracture is highly possible whereas small values indicate a safe region.
Simplified Model
For elastic, elastic-plastic, or rigid-plastic materials, there is the option for you to define a simplified damage model.
You define the damage factor (df) in the UDAMAG user subroutine.
If model 9 is used, then:
If model 10 is used, then:
The normal data for a specific material are defined with the ISOTROPIC, WORK HARD, and OGDEN options. Cross-
reference to this material is made with the material number.
D
D
c
D
D
c
D
1c
U
2
f ( ) ( )
2
--------------------------- 1 D ( )
2
= 1 D
c
0
D
1c
D
1c

v
D
c
D
c
D D
c
1 =
D D
c
D
D
c
D D
c
D
r el
D
r el
D
D
c
------ = 0 D
r el
1
D
r el
y

y

p
p
T , , ( )* 1.0 df ( ) =
y

y

p
p
T , , ( )* 1.0 df ( ) and E = E T ( )* 1.0 df ( ) =
523
CHAPTER 7
Material Library
Cockroft-Latham Damage Indicator
Cockroft-Latham damage indicator does not affect the yield stress. It is a postprocessing value to indicate a possible
damage area. It can also be used to initiate crack by removing elements in the area.
where is the maximum principal stress, is the effective von Mises stress and is the effective plastic strain
rate. is material constant threshold for damage.
Principal-tension Damage Indicator
This damage indicator does not affect the yield stress.
where is the maximum principal stress and is the effective von Mises stress. is material constant threshold
for damage.
Figure 7-96 shows the application of the Cockroft-Latham damage criterion used to predict cracks known as Chevron
cracks in the drawing process.
Oyane Damage Indicator
Similar to Cockraft-Latham, Oyane damage indicates a possible damage area where is the mean or average stress,
and are both material constants.
The UDAMAGE_INDICATOR user subroutine can be used to define unsupported damage criteria. The UACTIVE user
subroutine can be used for element removal.
Figure 7-97 show the application of the Oyane damage criterion used to predict cracks known as Chevron cracks in
the extrusion process.
max
-------------
t C d
}
max

max
------------- t C d
}
max
C
------- B +
\ .
| |
t C d
}
m
B C

Figure 7-96 Chevron Crack Prediction with Cockroft-Latham Damage Criterion
Figure 7-97 Chevron Crack Prediction with Oyane Damage Criterion
Nonstructural Materials
In addition to stress analysis, Marc can be used for heat transfer, coupled thermo-electrical heating (Joule heating),
coupled electrical-thermal-mechanical analysis (Joule-mechanical), hydrodynamic bearing, fluid/solid interaction,
electrostatic, magnetostatic, magnetodynamic, piezoelectric, acoustic, and fluid problems. Material properties
associated with these analyses and Marc options that control these analyses are described below.
Heat Transfer Analysis
In heat transfer analysis, use the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition options to input
values of thermal conductivity, specific heat, and mass density. If the latent heat effect is to be included in the analysis,
the value of latent heat and associated solidus and liquidus temperatures must be entered through the LATENT HEAT
525
CHAPTER 7
Material Library
or TEMPERATURE EFFECTS model definition option. Both the thermal conductivity and specific heat can be
dependent on temperatures. The mass density must be constant throughout conventional heat transfer analysis. In
addition, the ANKOND user subroutine can be used for the input of anisotropic thermal conductivity. For radiation
analysis, you must enter the Stefan-Boltzmann constant in the RADIATION parameter and the emissivity through the
EMISSIVITY or ISOTROPIC option. The emissivity can be temperature dependent.
For high temperature analysis where pyrolysis occurs, the THERMO-PORE option is used to define the advanced
material behavior.
In a piezoelectric analysis, the material properties contain a mechanical part, an electrostatic part, and a part connecting
these two. The material properties for the mechanical part are the same as for an elastic stress analysis. The material
properties for the electrostatic part, the permittivity, and the material properties for the piezoelectric coupling can be
defined through the PIEZOELECTRIC option. The coefficients for the piezoelectric coupling can be either stress based
or strain based.
Thermo-Electrical Analysis
In addition to thermal conductivity, specific heat and mass density, the value of electric resistivity must be entered
using the ISOTROPIC and ORTHOTROPIC model definition options in a coupled thermo-electrical analysis. Input the
variation of electric resistivity with temperature through the TEMPERATURE EFFECTS option.
Coupled Electrical-Thermal-Mechanical Analysis
Material properties for coupled electrical-thermal-mechanical analysis are the same for stress analysis in addition to
those of Joule heating analysis (thermal conductivity, specific heat and mass density and electric resistivity).
Hydrodynamic Bearing Analysis
In a hydrodynamic bearing analysis, the ISOTROPIC and ORTHOTROPIC model definition options can be used for
entering both the viscosity and specific mass. Define the temperature dependency of viscosity through the
TEMPERATURE EFFECTS option.
Fluid/Solid Interaction Analysis Added Mass Approach
In a fluid/solid analysis, the density of a fluid is given in the first field of the ISOTROPIC option. The fluid is assumed
to be nonviscous and incompressible.
In an electrostatic analysis, the permittivity can be defined through either the ISOTROPIC or ORTHOTROPIC options.
In addition, the UEPS user subroutine can be used for the input of anisotropic permittivity.

In a magnetostatic analysis, the permeability can be defined through either the ISOTROPIC or ORTHOTROPIC
options. A nonlinear relationship can be entered via the B-H RELATION option. The UMU user subroutine can be used
for the input of isotropic permeability.
In an magnetodynamic analysis, the permittivity, permeability and conductivity can be defined using either the
ISOTROPIC or ORTHOTROPIC options. An assumption made is that the permittivity is a constant (does not vary with
time) in the analysis. A nonlinear permeability can be entered via the B-H RELATION option. The UEPS, UMU, and
USIGMA user subroutines can be used for the input of anisotropic permittivity, permeability, and conductivity,
respectively. Both the dependent and harmonic analysis can be performed.
Coupled Electrostatic-Structural
Material properties for a coupled electrostatic-structural are the same as for a stress analysis in addition to those for an
electrostatic analysis (permittivity).
Acoustic Analysis
In an acoustic analysis of a cavity with rigid boundaries, the bulk modulus and the relative density of the medium can
be entered through the ISOTROPIC option.
The ACOUSTIC parameter is used to indicate that a coupled acoustic-structural analysis is performed. In addition to
the CONTACT option, the ACOUSTIC and REGION model definition options are used to define the material properties
of the acoustic medium and to set which elements correspond to the solid and the fluid region.
Fluid Analysis
In the fluid analysis, viscosity and density can be defined using the ISOTROPIC option. In addition, conductivity and
specific heat are defined for coupled fluid-thermal analysis. Non-Newtonian fluid behavior can be defined using the
STRAIN RATE option. The TEMPERATURE EFFECTS option can be used to define the temperature dependence of
the properties above.
References
1. Auricchio, F. and Taylor, R.L., Shape-memory alloy: modeling and numerical simulations of the finite-strain
superelastic behavior, Comput. Methods Appl. Mech. Engrg., Vol. 143, pp.175-194 (1997).
2. Auricchio, F., A robust integration-algorithm for a finite-strain shape-memory-alloy superelastic model, Int.
J. Plasticity, Vol.17, pp.971-990 (2001).
3. Arruda, E. M. and Boyce, M. C. A three-dimensional constitutive model for the large stretch behaviour of
rubber elastic materials, J. Mech. Phys. Solids, Vol.41, No. 2, 1993.
4. Buyukozturk, O., Nonlinear Analysis of Reinforced Concrete Structures, Computers and Structures, 7, 149-
156 (1977).
527
CHAPTER 7
Material Library
5. Gent, A. N., A new constitutive relation for rubber, Rubber Chem. Tech., 69, 1996.
6. Barlat, F., Lege, D.J. and Brem, J.C., A six-component yield function for anisotropic metals, Int. J. Plasticity,
7, 693-712 (1991).
7. Chung, K. and Shah, K., Finite element simulation of sheet metal forming for planar anisotropic metals, Int.
J. Plasticity, 8, 453-476 (1992).
8. Yoon, J.W., Yang, D.Y. and Chung, K. and Barlat. F., A general elasto-plastic finite element formulation
based on incremental deformation theory for planar anisotropy and its application to sheet metal forming, Int.
J. Plasticity, 15, 35-67 (1999).
9. Yoon, J.W., Barlat, F., Chung, K., Pourboghrat, F. and Yang, D.Y., Earing predictions based on asymmetric
nonquadratic yield function, Int. J. Plasticity, 16, 1075-1104 (2000).
10. Current Recommended Constitutive Equations for Inelastic Design Analysis of FFTF Components. ORLN-
TM-360Z, October 1971.
11. Finite Element Calculation of Stresses in Glass Parts Undergoing Viscous Relaxation, J.Am.Ceram.Soc. Vol.
70[2], pp. 90-95, 1987.
12. Mooney, M. J. Appl Phys., Vol. II, p. 582, 1940.
13. Naghdi, P. M. Stress-Strain Relations in Plasticity and Thermoplasticity. In Plasticity, Proceedings of the
Second Symposium on Naval Structural Mechanics, edited by E. H. Lee and P. S. Symonds. Pergamon Press,
1960.
14. Narayanaswamy, O.S., A Model of Structural Relaxation in Glass, J.Am.Ceram.Soc., Vol. 54[10], pp. 491-
498, 1971.
15. Prager, W. Introduction to Mechanics to Continua. New York: Dover Press, 1961.
16. Proposed Modification to RDT Standard F9-5T Inelastic Analysis Guidelines. ORNL, October 1978.
17. Rivlin, R. S. Phil Trans Roy Soc (A), Vol. 240, 459, 1948.
18. Rivlin, R. S. Phil Trans Roy Soc (A), Vol. 241, 3-79, 1948.
19. Timoshenko, S. P., and J. N. Goodier. Theory of Elasticity. Third Ed. New York: McGraw-Hill, 1970.
20. Simo, J. C. On a Fully Three-dimensional Finite Strain Viscoelastic Damage Model: Formulation and
Computational Aspects, Computer Methods in Applied Mechanics and Engineering, Vol. 60, 1987,
pp. 153-173.
21. Tvergaard, V., Influence of Voids on Shear Band Instabilities under Plane Strain Conditions, Int. J. Fracture,
Vol. 17, pp. 389-407, 1981.
22. Tvergaard, V., Material Failure by Void Coalescence in Localized Shear Bands, in Int. J. Solids Struct., Vol.
18, No. 8, pp. 652-672, 1982.
23. Chaboche, J. L., Constitutive Equations for Cyclic Plasticity and Cyclic Viscoplasticity, International
Journal of Plasticity, Vol. 5, pp. 247-302, 1989
24. van den Boogard, A.H., Implicit integration of the Perzyna viscoplastic material model, TNO Building and
Construction Research, The Netherlands, TNO-report, 95-NM-R711, 1995
25. Puck, A. and Schrmann, H., Failure Analysis of FRP Laminates by Means of Physically Based
Phenomenological Models; Composite Science and Technology, 58, pp. 1045-1067, 1998.

26. Camacho, G.T. and Ortiz, M., Computational modelling of impact damage in brittle materials, Int. J. Solids
Struct., Vol. 33, pp 2899-2938, 1996.
27. Greve, L. and Picket, A. K., Modelling Damage and Failure in Carbon/Epoxy Non-crimp Fabric Composites
Including Effects of Fabric Pre-shear, Composites: Part A, Vol. 37, pp. 1983-2001, 2006.
28. Knops, M. and Bgle, C. Gradual Failure in Fibre/Polymer Laminates, Composites Science and Technology,
Vol. 66, pp. 616-625, 2006.
29. Buchanan, David L., Gosse, Jonathan H., Wollschlager, Jeffrey A., Ritchey, Andrew, and Pipes, R. Byron,
Micromechanical enhancement of the macroscopic strain state for advanced composite materials,
Composites Science and Technology, pp. 1974-1978 (2009).
30. Gosse, Jon H., Boeing Company, Strain Invariant Failure Criteria for Fiber Reinforced Polymeric Composite
Materials.
31. Brown, S. B., Kim, K. H., and Anand, L., 1989, An Internal Variable Constitutive Model for Hot Working of
Metals, Int. J. Plast., 5, pp. 95130.
Chapter 8 Contact
8
Contact
Definition of Contact Bodies 530
Numbering of Contact Bodies 532
Motion of Bodies 536
Detection of Contact 538
Implementation of Constraints 542
Friction Modeling 544
Separation 558
Coupled Analysis 561
Thermal Contact 563
Element Considerations 565
Dynamic Impact 568
Global Adaptive Meshing and Rezoning 569
Adaptive Meshing 569
Result Evaluation 570
Tolerance Values 571
Numerical and Mathematical Aspects of Contact 572
References 595

The simulation of many physical problems requires the ability to model the contact phenomena. This includes analysis
of interference fits, rubber seals, tires, crash, and manufacturing processes among others. The analysis of contact
behavior is complex because of the requirement to accurately track the motion of multiple geometric bodies, and the
motion due to the interaction of these bodies after contact occurs. This includes representing the friction between
surfaces and heat transfer between the bodies if required. The numerical objective is to detect the motion of the bodies,
apply a constraint to avoid penetration, and apply appropriate boundary conditions to simulate the frictional behavior
and heat transfer. Several procedures have been developed to treat these problems including the use of Perturbed or
Augmented Lagrangian methods, penalty methods, and direct constraints. Furthermore, contact simulation has often
required the use of special contact or gap elements. Marc allows contact analysis to be performed automatically
without the use of special contact elements. A robust numerical procedure to simulate these complex physical
problems has been implemented in Marc.
Definition of Contact Bodies
There are two types of contact bodies in Marc deformable and rigid. Deformable bodies are simply a collection of
finite elements as shown below.
Figure 8-1 Deformable Body
This body has three key aspects to it:
1. The elements which make up the body.
2. The nodes on the external surfaces which might contact another body or itself. These nodes are treated as
potential contact nodes.
3. The edges (2-D) or faces (3-D) which describe the outer surface which a node on another body (or the same
body) might contact. These edges/faces are treated as potential contact segments.
Note that a body can be multiply connected (have holes in itself). It is also possible for a body to be composed of both
triangular elements and quadrilateral elements in 2-D or tetrahedral elements and brick elements in 3-D. Beam
elements and shells are also available for contact. Both lower-order (linear) and higher-order (quadratic) elements may
be used. One should not mix continuum elements, shells, and/or beams in the same contact body.
Each node and element should be in, at most, one body. The elements in a body are defined using the CONTACT
option. It is not necessary to identify the nodes on the exterior surfaces as this is done automatically. The algorithm
used is based on the fact that nodes on the boundary are on element edges or faces that belong to only one element.
Each node on the exterior surface is treated as a potential contact node. In many problems, it is known that certain
nodes never come into contact; in such cases, the CONTACT NODE option can be used to identify the relevant nodes.
As all nodes on free surfaces are considered contact nodes, if there is an error in the mesh generation such that internal
holes or slits exist, undesirable results can occur.
531
CHAPTER 8
Contact
The potential segments composed of edges or faces are treated in potentially two ways. The default is that they are
considered as piece-wise linear (PWL). As an alternative, a cubic spline (2-D) or a Coons surface (3-D) can be placed
through them. The SPLINE option is used to activate this procedure. This improves the accuracy of the calculation of
the normal to the surface.
Rigid bodies are composed of curves (2-D) or surfaces (3-D) or meshes with only thermal elements in coupled
problems. The most significant aspect of rigid bodies is that they do not deform. Deformable bodies can contact rigid
bodies, but contact between rigid bodies is not considered.
They can be created either in CAD systems and transferred through Mentat into Marc, created within Mentat, or
created directly through the Marc input. There are several different types of curves and surfaces that can be entered
including:
Within the Mentat GUI, all curves or surfaces are mathematically treated as NURBS curves or surfaces. This allows
the greatest level of generality. Within the analysis, these rigid curves or surfaces can be treated in two ways discrete
piecewise linear lines (2-D) or patches (3-D), or as analytical NURBS curves or surfaces. When the discrete approach
is used, all geometric primitives are subdivided into straight line segments or bilinear patches. You have control over
the density of these subdivisions to approximate a curved line or surface within a desired degree of accuracy. This
subdivision is also relevant when the corner conditions (see Detection of Contact on page 538) are determined. Too
few subdivisions might lead to a very bad surface description and too many subdivisions might lead to a more
expensive analysis due to more complex contact checking. When using the analytical description, the number of
subdivisions given is less relevant. It should be large enough to roughly visualize the shape of the NURBS. In 3-D, a
too large number is automatically reduced so that the analysis cost is not influenced too much. The treatment of the
rigid bodies as NURBS surfaces is advantageous because it leads to greater accuracy in the representation of the
geometry and a more accurate calculation of the surface normal. Additionally, the variation of the surface normal is
continuous over the body which leads to a better calculation of the friction behavior and better convergence.
Note:
These problems can be visualized by using Mentat to create an outline plot and fixed by using the sweep
command in Mentat.
Table 8-1 Geometrical Entities Used in Modeling Contact
2-D 3-D
line 4-node patch
circular arc ruled surface
spline surface of revolution
NURBS Bezier
poly-surface
cylinder
sphere
NURBS
trimmed NURBS

To create a rigid body, you can either read in the curve and surface geometry created from a CAD system or create the
geometry in Mentat or directly enter it in Marc. You then use the CONTACT option to select which geometric entities
are to be a part of the rigid body. An important consideration for a rigid body is the definition of the interior side and
the exterior side. For two-dimensional analysis, the interior side is formed by the right-hand rule when moving along
the body.
This is easily visualized with Mentat by activating ID CONTACT.
Figure 8-2 Orientation of Rigid Body Segments
For three-dimensional analysis, the interior side is formed by the right-hand rule along a patch. The interior side is
visualized in Mentat as the front surface (pink); whereas, the exterior side is visualized in Mentat as the back surface
(gold).
It is not necessary for rigid bodies to define the complete body. Only the bounding surface needs to be specified. You
should take care, however, that the deforming body cannot slide out of the boundary curve in 2-D (Figure 8-3). This
means that it must always be possible to decompose the displacement increment into a component normal and a
component tangential to the rigid surface.
Figure 8-3 Deformable Surface Sliding Out of Rigid Surface
Numbering of Contact Bodies
When defining contact bodies for a deformable-to-deformable analysis, it is important to define them in the proper
order. As a general rule, a body with a finer mesh should be defined before a body with a coarser mesh.
Deformable
Rigid
X
Y
Z
1
top/front
top/front/inside
bottom/back/outside
X
Y
Z
1
Incorrect Correct
533
CHAPTER 8
Contact
If one has defined a body numbering which violates the general rule, or if the rule is violated upon remeshing, then a
CONTACT TABLE model or history definition option can be used to modify the order in which contact is established.
This order can be directly user-defined or decided by the program. In the latter case, the order is based on the rule that
if two deformable bodies might come into contact, searching is done such that the contacting nodes are on the body
having the smallest element edge length and the contacted segments are on the body having the coarser mesh. It should
be noted that this implies single-sided contact for this body combination, as opposed to the default
double-sided contact.
For deformable-deformable contact, the CONTACT option offers the following three global contact search controls:
SINGLE-SIDED
DOUBLE_SIDED
OPTIMIZE CONTACT CONSTRAINT EQUATIONS
If flexible region A comes into contact with flexible region B, the following rules will be satisfied by the program to
determine the constraint equations:
1. Region A and region B belong to the same contact body:
a. If OPTIMIZE CONTACT CONSTRAINT EQUATIONS is active, the finer meshed region contacts the
coarser meshed region.
b. If OPTIMIZE CONTACT CONSTRAINT EQUATIONS is not active, the region containing the lower
node number coming into contact contacts the other region.
2. Region A and region B do not belong to the same contact body:
a. If SINGLE-SIDED contact is active, the region belonging to the lower contact body number contacts the
region belonging to the higher contact body number.
b. If DOUBLE_SIDED contact is active, the region containing the lower node number coming into contact
contacts the other region. Because of the search order for contact is from lower body numbers to higher body
numbers, this often leads to the same solution as mentioned above.
c. If OPTIMIZE CONTACT CONSTRAINT EQUATIONS is active, the region with the softer material
contacts the region with the harder material, or the region with the finer mesh contacts the region with the
coarser mesh (soft-hard criterion has priority over mesh-density criterion).
d. All the above settings can be overruled by the CONTACT TABLE model or history definition option where
you can specify that a certain body AA should contact body BB and not vice-verse, or that the body with the
smallest element edge should contact the other body.
As an example, two identical bodies shown in Figure 8-4, but created using two different approaches so that
the node numbering order is different. Each U-Section is one contact body. In particular, for the model on
the left, the left leg of the U-section has lower node ids that the right leg; while for the model on the right,
the opposite is true.
Note:
For problems involving adaptive meshing or automated remeshing, care must be taken to satisfy this rule
before as well as after the mesh change.

Figure 8-4 Node Numbers on Models
A uniform distributed load is placed on the right leg, such that self-contact occurs. Without using optimized
contact, the behavior is shown in Figure 8-5.
Figure 8-5 Contact Based upon Original Model.
The dark squares indicate nodes that are in contact based upon the Contact Status. Observe that the model
on the left gives good behavior, while that on the right does not. This is because the lower node numbers on
the left leg detect contact with the right leg before the corner node on the right leg (which has a higher id)
detects contact with the left leg. Then, activate the Optimized Contact Constraint option, and the results
are shown in Figure 8-6. The node number IDs are no longer controlling the solution.
535
CHAPTER 8
Contact
Figure 8-6 Contact using Optimized Contact Constraints
Each model is then subdivided into two contact bodies as shown in Figure 8-7, the contact table option is
not used and the optimized procedure is not used.
Figure 8-7 Contact Based upon Contact Body Numbering
Observe now that in the left model, the left leg is a lower body number, so the nodes on this body try to
contact the segments on the right leg. In the model on the right then, the right leg has the lower body number,
and so the nodes on this leg contact (successfully) with the segments on the left leg.
When either the Optimized Contact Constraint Procedure or a well-selected Contact Table utilizing
Single-Sided Contact, the consistent results may be obtained for either model as shown in Figure 8-8

Figure 8-8 Contact Using either Optimized Contact Constraint or Contact Table.
Motion of Bodies
The motion of deformable bodies is prescribed using the conventional methods of applying displacements, forces, or
distributed loads to the bodies. It is advantageous to not apply displacements or point loads to nodes which might come
into contact with other rigid bodies. If a prescribed displacement is to be imposed, it is better to introduce another rigid
body and apply the motion to the rigid body. Symmetry surfaces are treated as a special type of bodies which have the
property of being frictionless and where the nodes are not allowed to separate. If a distributed load is applied to an
edge or face, it is possible for this load to be deactivated if all nodes on the edge or face are in contact with another
body.
There are four ways to prescribe the motion of rigid surfaces:
Prescribed velocity
Prescribed position
Prescribed load
Prescribed scaling
Associated with the rigid body is a point labeled the centroid. When the first two methods are chosen, you define the
translational motion of this point, and the angular motion about an axis through this point. The direction of the axis
can be defined for three-dimensional problems. For two-dimensional problems, it is a line normal to the plane. For
complex time-dependent behavior, the MOTION user subroutine can be used to prescribe the motion as an alternative
to the input. The motion during a time increment is considered to be linear. The position is determined by an explicit,
forward integration of the velocities based upon the current time step. A time increment must always be defined even
if a static, rate-independent analysis is performed.
When load controlled rigid bodies are used, two additional nodes, called the control nodes, are associated with each
rigid body. In 2-D problems, the first node has two translational degrees of freedom (corresponding to the global x-
and y-direction) and the second node has one rotational degree of freedom (corresponding to the global z-direction).
In 3-D problems, the first node has three translational degrees of freedom (corresponding to global x-, y-, and
z-direction) and the second node has three rotational degrees of freedom (corresponding to the global x-, y-, and z-
direction). In this way, both forces and moments can be applied to a body by using the POINT LOAD option for the
control nodes. Alternatively, one may prescribe one or more degrees of freedom of the control nodes by using the
FIXED DISP or DISP CHANGE options. Generally speaking, load-controlled bodies can be considered as rigid bodies
537
CHAPTER 8
Contact
with three (in 2-D) or six (in 3-D) degrees of freedom. These degrees of freedom may be in a user-defined system of
either the TRANSFORMATION or COORD SYSTEM option is used. The prescribed position and prescribed velocity
methods (see Figure 8-9) have less computational costs than the prescribed load method (see Figure 8-10).
Figure 8-9 Velocity Controlled Rigid Surface
Figure 8-10 Load Controlled Rigid Surface
If the second control node is not specified, the rotation of the body is prescribed to be zero.
It should be noted that the nodal coordinates of the first control node define the center of rotation for the body. The
nodal coordinates of the second control node are arbitrary.
It is also possible to either expand or contract the size of the rigid body through the use of specified scale factors
multiplied by a table to make the scaling time dependent. Alternately, the UGROWRIGID user subroutine may be used
to provide scale factors. The scale factors must be initialized to 1.0 but can change afterwards. Abrupt or discontinuous
changes in the scale factors should be avoided while the rigid surface is in contact with other bodies. If no rotations
are applied, the scale factors can be different in the x-, y-, and z-directions.
1
2
3
1
2
V
Centroid
Extra Node
F
x
F
y
M
z

Initial Conditions
At the beginning of the analysis, bodies should either be separated from one another or in contact. Bodies should not
penetrate one another unless the objective is to perform an interference fit calculation. Rigid body profiles are often
complex, making it difficult for you to determine exactly where the first contact is located. Before the analysis begins
(increment zero), if a rigid body has a nonzero motion associated with it, the initialization procedure brings it into first
contact with a deformable body. No motion or distortion occurs in the deformable bodies during this process. In a
coupled thermal mechanical analysis, no heat transfer occurs during this process. If more than one rigid body exists in
the analysis, each one with a nonzero initial velocity is moved until it comes into contact. Because increment zero is
used to bring the rigid bodies into contact only, you should not prescribe any loads (distributed or point) or prescribed
displacements initially. For multistage contact analysis (often needed to simulate manufacturing processes), the
APPROACH and SYNCHRONIZED options in conjunction with the CONTACT TABLE and MOTION CHANGE options
allow you to model contact bodies so that they just come into contact with the workpiece.
In assembly analysis, it is possible that multiple bodies are initially in contact. Because of mesh discretization, it is
possible that the contact is not perfect, which would result in inducing stresses due to overclosure. The CONTACT
TABLE option may be used to specify that the initial contact should be stress free. In such cases, the coordinates of the
nodes are relocated on the surface so that initial stresses do not occur.
Detection of Contact
During the incremental procedure, each potential contact node is first checked to see whether it is near a contact
segment. The contact segments are either edges of other 2-D deformable bodies, faces of 3-D deformable bodies, or
segments from rigid bodies. By default, each node could contact any other segment including segments on the body
that it belongs to. This allows a body to contact itself. To simplify the computation, it is possible to use the CONTACT
TABLE option to indicate that a particular body will or will not contact another body. This is often used to indicate that
a body will not contact itself. During the iteration process, the motion of the node is checked to see whether it has
penetrated a surface by determining whether it has crossed a segment.
Because there can be a large number of nodes and segments, efficient algorithms have been developed to expedite this
process. A bounding box algorithm is used so that it is quickly determined whether a node is near a segment. If the
node falls within the bounding box, more sophisticated techniques are used to determine the exact status of the node.
During the contact process, it is unlikely that a node exactly contacts the surface. For this reason, a contact tolerance
(Figure 8-11) is associated with each surface.
Figure 8-11 Contact Tolerance
2
x Tolerance
539
CHAPTER 8
Contact
If a node is within the contact tolerance, it is considered to be in contact with the segment. The default contact tolerance
is calculated by the program (see Tolerance Values in this chapter). It is also possible to define the contact tolerance
through the input, either globally via the CONTACT option or per pair of contact bodies via the CONTACT TABLE
option.
During an increment, if node A moves from to , where is beyond the contact tolerance, the node
is considered to have penetrated (Figure 8-12). In such a case, a special procedure is invoked to avoid this penetration.
More details are discussed in Numerical and Mathematical Aspects of Contact in this chapter.
Figure 8-12 Trial Displacement with Penetration
The size of the contact tolerance has a significant impact on the computational costs and the accuracy of the solution.
If the contact tolerance is too small, detection of contact and penetration is difficult which leads to higher costs.
Penetration of a node happens in a shorter time period leading to more recycles due to iterative penetration checking
or to more increment splitting and increases the computational costs. If the contact tolerance is too large, nodes are
considered in contact prematurely, resulting in a loss of accuracy or more recycling due to separation. Furthermore,
the accepted solution might have nodes that penetrate the surface less than the error tolerance, but more than desired
by the user. The default error tolerance is recommended.
Many times, areas exist in the model where nodes are almost touching a surface (for example, in rolling analysis close
to the entry and exit of the rolls). In such cases, the use of a biased tolerance area with a smaller distance on the outside
and a larger distance on the inside is advised. This avoids the close nodes from coming into contact and separating
again and is accomplished by entering a bias factor. The bias factor should be given a value between 0.0 and 0.99. The
default is 0.0 or no bias. Also, in analyses involving frictional contact, a bias factor for the contact tolerance is
recommended. The outside contact area is (1. - bias) times the contact tolerance on the inside contact area (1. + bias)
times the contact tolerance (Figure 8-13). The bias factor recommended value is 0.95.
In some instances, you might wish to influence the decision regarding the deformable segment a node contacts (or does
not contact). This can be done using the EXCLUDE option.
A
t ( )
A
t+ ( )
A
t+ ( )
A
(t)
A
(t + t)

Figure 8-13 Biased Contact Tolerance
Shell Contact
A node on a shell makes contact when the position of the node plus or minus half the thickness projected with the
normal comes into contact with another segment (Figure 8-14). In 2-D, this can be shown as:
Figure 8-14 Default Shell Contact
If point x or y falls within the contact tolerance distance of segment S, node A is considered in contact with the segment
S. Here and are the position vectors of a point on the surfaces 1 and 2 on the shell, A is the position vector of
a point (node in a discretized model) on the midsurface of the shell, is the normal to the midsurface, and is the
shell thickness.
As the shell has finite thickness, the node (depending on the direction of motion) can physically contact either the top
surface, bottom surface, or mathematically contact can be based upon the midsurface. You can control whether
detection occurs with either both surfaces, the top surface, the bottom surface, or the middle surface. In such cases,
either two or one segment will be created at the appropriate physical location. Note that these segments will be
dependent, not only on the motion of the shell, but also the current shell thickness (Figure 8-15).
Note:
Shell elements should be oriented consistently. Also, a shell segment cannot be in contact simultaneously
on both the Top and Bottom sides. If simultaneous contact is required, use the solid shell element.
(1 - Bias) tolerance
(1 + Bias) tolerance
x
1
A n t 2 + =
x
2
A n t 2 =
x
x
2
x
to
le
ra
n
ce
A
S
Shell
Midsurface
t
1
2
x
1
x
2
n t
541
CHAPTER 8
Contact
Figure 8-15 Selective Shell Contact
are segments associated with shell consisting of node 1 and 2.
Neighbor Relations
When a node is in contact with a rigid surface, it tends to slide from one segment to another. In 2-D, the segments are
always continuous and so are the segment numbers. Hence, a node in contact with segment n slides to segment
or to segment (Figure 8-16). This simplifies the implementation of contact.
Figure 8-16 Neighbor Relationship (2-D)
In 3-D, the segments are often discontinuous (Figure 8-17). This can be due to the subdivision of matching surfaces
or, more likely, the CAD definition of the under lying surface geometry.
Figure 8-17 Neighbor Relationship (3-D)
n
1
2
S
1
S
2
1
n
2
S
1
S
2
2
1
n
1
S
1
2
Include Both Segments Top Segment Only
Bottom Segments Only Ignore Shell Thickness
S
1
S
2
,
n 1
n 1 +
n - 1
n
n + 1
Hole Physically
Discontinuous
Nonmatching
Segments
Continuous Surface Segments
Discontinuous Surface Geometry

Continuous surface geometry is highly advantageous as a node can slide from one segment to the next with no
interference (assuming the corner conditions are satisfied). Discontinuous surface geometry results in additional
operations when a node slides off a patch and cannot find an adjacent segment. Hence, it is advantageous to use
geometry clean-up tools to eliminate small sliver surfaces and make the surfaces both physically continuous and
topologically contiguous.
Implementation of Constraints
For contact between a deformable body and a rigid surface, the constraint associated with no penetration is
implemented by transforming the degrees of freedom of the contact node and applying a boundary condition to the
normal displacement. This can be considered solving the problem:
where represents the nodes in contact which have a local transformation, and represents the nodes not in contact
and, hence, not transformed. Of the nodes transformed, the displacement in the normal direction is then constrained
such that is equal to the incremental normal displacement of the rigid body at the contact point.
Figure 8-18 Transformed System (2-D)
As a rigid body can be represented as either a piecewise linear or as an analytical (NURBS) surface, two procedures
are used. For piecewise linear representations, the normal is constant until node P comes to the corner of two segments
as shown in Figure 8-19. During the iteration process, one of three circumstances occur. If the angle is small
, the node P slides to the next segment. In such a case, the normal is updated based upon
the new segment. If the angle is large ( or ) the node separates from the surface if it is a
convex corner, or sticks if it is a concave corner. The value of is important in controlling the computational
costs. A larger value of reduces the computational costs, but might lead to inaccuracies. The default values
are 8.625 for 2-D and 20 for 3-D. These can be reset using the PARAMETERS option in the model definition or
history definition section.
K
a
K
a
b
K
ba
K
bb
u
a
u
b )
`

f
a
f
b )
`

=
a
b
u
a
n
n
t
P
smooth

smooth
< < ( )

smooth
>
smooth
<
smooth
smooth
543
CHAPTER 8
Contact
Figure 8-19 Corner Conditions (2-D)
In 3-D, these corner conditions are more complex. A node (P) on patch A slides freely until it reaches the intersection
between the segments. If it is concave, the node first tries to slide along the line of intersection before moving to
segment B. This is the natural (lower energy state) of motion.
These corner conditions also exist for deformable-to-deformable contact analysis. Because the bodies are continuously
changing in shape, the corner conditions (sharp convex, smooth or sharp concave) are continuously being re-evaluated.
When a rigid body is represented as an analytical surface, the normal is recalculated at each iteration based upon the
current position. This leads to a more accurate solution, but can be more costly because of the NURBS evaluation.
Figure 8-20 Corner Conditions (3-D)
When a node of a deformable body contacts a deformable body, a multipoint constraint (called tying) is automatically
imposed. Recalling that the exterior edges (2-D) or faces (3-D) of the other deformable bodies are known, a constraint
expression is formed. For 2-D analysis using lower-order elements, the number of retained nodes is three two from
the edge and the contacting node itself. For 3-D analysis, the number of retained nodes is five four from the patch
and the contacting node itself. When using higher-order elements and true quadratic contact, the number of retained
nodes for 2-D becomes four, for 3-D, (hexahedrals) nine, and for 3-D (tetrahedrals) seven. The constraint equation is
such that the contacting node should be able to slide on the contacted segment, subject to the current friction
conditions. This leads to a nonhomogeneous, nonlinear constraint equation. In this way, a contacting node is forced to
be on the contacted segment. This might introduce undesired stress changes, since a small gap or overlap between the
node and the contacted segment will be closed (note that using the Single Step Houbolt or generalized alpha/HHT
operator in dynamics, forcing a node to be on the contacted segment is switched off by default, but can be activated
via the PARAMETERS option). During initial detection of contact (increment 0), the stress-free projection option
avoids those stress changes for deformable contact by adapting the coordinates of the contacting nodes such that they
are positioned on the contacted segment. This stress-free projection can be activated using CONTACT TABLE. A
similar option exists for glued contact; however, in this case, overlap will not be removed.
Convex Corner Concave Corner

A
B
P
P

During the iteration procedure, a node can slide from one segment to another, changing the retained nodes associated
with the constraint. A recalculation of the bandwidth is automatically made. Because the bandwidth can radically
change, the bandwidth optimization is also automatically performed.
A node is considered sliding off a contacted segment if is passes the end of the segment over a distance more than the
contact tolerance. As mentioned earlier, the node separates from the contacted body if this happens at a convex corner.
For deformable contact, this tangential tolerance at convex corners can be enlarged by using the delayed sliding off
option activated via CONTACT TABLE.
Friction Modeling
Friction is a complex physical phenomenon that involves the characteristics of the surface such as surface roughness,
temperature, normal stress, and relative velocity. An example of this complexity is that quite often in contact problems
neutral lines develop. This means that along a contact surface, the material flows in one direction in part of the surface
and in the opposite direction in another part of the surface. Such neutral lines are, in general, not known a priori.
The actual physics of friction and its numerical representation continue to be topics of research. Currently, in Marc the
modeling of friction has basically been simplified to two idealistic models. The most popular friction model is the
Coulomb friction model. This model is used for most applications with the exception of bulk forming as encountered
in e.g. forging processes. For such applications the shear friction model is more appropriate.
Coulomb Friction
The Coulomb model can be characterized by:
(stick) and (slip)
where
Similarly, the Coulomb model can also be written in terms of nodal forces instead of stresses:
(stick) and (slip)
is the tangential (friction) stress
is the normal stress
is the friction coefficient
is the tangential vector in the direction of the relative velocity:
, in which is the relative sliding velocity.
t

n
<
t

n
t =
t
t
v
r
v
r
----------- = v
r
f
t
f
n
< f
t
f
n
t =
545
CHAPTER 8
Contact
where
When Coulomb friction is used with the stress-based model, the integration point stresses are first extrapolated to the
nodal points and then transformed, so that a direct component is normal to the contacted surface. Given this normal
stress and the relative sliding velocity, the tangential stress is then evaluated and a consistent nodal force is calculated.
For shell elements, the nodal force based Coulomb model must be used, since a reduced stress state is assumed with
. For continuum elements, both the stress-based and force-based Coulomb model can be used.
For a given normal stress or normal force, the friction stress or force has a step function behavior based upon the value
of the relative sliding velocity or the tangential relative incremental displacement , as outlined in Figure 8-21
for a 2-D case, where the relative velocity and incremental displacement are scalar values.
Figure 8-21 Coulomb Friction Model
Since this discontinuity in the friction value may easily cause numerical difficulties, different approximations of the
step function have been implemented. They are graphically represented in Figure 8-22 and they will be successively
discussed.
Figure 8-22 Different Approximations for the Coulomb Friction Model
is the tangential (friction) force
is the normal force
f
t
f
n
n
0 =
v
r
u
t
Stick
Slip
t
f
t
v
r
u
t
or
or
t
f
t
or
v
r
f
t
f
t
u
t
u
t
arctangent model stick-slip (modified step bilinear model
function) model

Arctangent Model
The first procedure is based on a continuously differentiable function in terms of the relative sliding velocity:
for the friction stress, and:
for the friction force.
Physically, the value of can be seen as the value of the relative velocity below which sticking occurs. The
value of is important in determining how closely the mathematical model represents the step function, as
shown in Figure 8-23. A very large value of results in a reduced value of the effective friction. A very
small value may result in poor convergence. It is recommended that the value of be 1% to 10% of a typical
relative sliding velocity, .
The friction implementation not only affects the external force vector, but also the stiffness matrix of the final set of
equations to be solved on a global level. This stiffness matrix contribution follows from:
This later contribution, if fully implemented, would lead to a non-symmetric system. Because of the additional
computational costs, both in terms of memory and CPU costs, the contribution to the stiffness matrix is symmetrized.
Figure 8-23 Step Function Approximation for Different Values of RVCNST (-10<v
r
<10)
t

n

2
--- arctan
v
r
RVCNST
-------------------------
\ .
| |
t =
f
t
f
n

2
--- arctan
v
r
RVCNST
-------------------------
\ .
| |
t =
RVCNST
RVCNST
RVCNST
RVCNST
v
r
K
i j
f
t
i
v
r
k
----------
v
r
k
u
t
j
------------- =
f
t
v
r
f
n
f
n
RVCNST 10 =
RVCNST 0.1 =
RVCNST 1 =
10 10
547
CHAPTER 8
Contact
During iteration of the Newton-Raphson process, the change of the friction force is related to the change of
the relative sliding velocity . Since the latter can be written as:
the resulting equation can still be expressed in displacement degrees of freedom. During the very first iteration of an
increment, it turns out that the convergence can be significantly improved by using:
where results in a contribution to the right hand side of the global set of equations.
As a result of the smoothing procedure, a node in contact always has some slipping. Besides the numerical reasons,
this ever slipping node model has a physical basis. Oden and Pires pointed out that for metals, there is an elasto-
plastic deformation of asperities at the microscopic level (termed as cold weld) which leads to a nonlocal and
nonlinear frictional contact behavior. The arctangent representation of the friction model is a mathematical idealization
of this nonlinear friction behavior.
A complicating aspect of the arctangent model is that it may be difficult to estimate a priori the typical relative sliding
velocity. This is particularly true in quasi-static deformable contact analyses (where the velocity is not an input
quantity) and in cases where the sliding velocity varies strongly during the analysis.
Stick-slip (Modified Step Function) Model
The second procedure is based on a slightly modified step function and can be used to simulate true stick-slip behavior.
In this procedure, each node in contact gets a friction status, being either stick or slip. Depending on this status,
different constraints are applied and after each iteration in the iterative solution process, the correctness of the friction
status is checked and if necessary adapted.
The typical parameters used (see also Figure 8-24) are the friction coefficient multiplier , the slip to stick transition
region and the friction force tolerance .
The multiplier (with a default value of ) can be seen as an overshoot parameter used to simulate the difference
between static and dynamic friction.
The slip-to-stick transition region (with a default value of ) can be seen as a tolerance on the friction
solution. If a node is in slipping mode and moves in the direction of the friction force (which is physically incorrect),
but the corresponding relative displacement magnitude is within the slip-to-stick transition region, then this will not
cause the increment to be restarted with modified friction conditions.
i ( ) f
t
i ( )
v
r
i ( )
v
r
i ( )
u
t
i ( )
u
t
i 1 ( )
t
-----------------------------------------
u
t
i ( )
t
------------ = =
v
r
1 ( )
u
t
1 ( )
t
--------------
u
t
previous increment ( )
t
------------------------------------------------
u
t
1 ( )
t
------------- v
r
= =
v
r
e
1.05
1
6
10

Related to the parameter , a reduction factor is used. This fixed factor causes the product of and
to be very small, so that it can be used as a criterion to estimate if a node is initially in slipping or sticking mode.
This initial assumption serves as the starting point of the above mentioned process of checking and adapting the
friction status, which, for a 2-D analysis, is given in Figure 8-25 (note that implies ).
Figure 8-24 Stick-slip or Modified Step Function Friction Parameters
Figure 8-25 Flow Diagram for the Stick-slip or Modified Step Function Friction Model
1.0
6
10 =
u
t
0 u
t
<
f
t
u
t
f
n
f
n
: friction coefficient multiplier

: slip to stick transition region
: reduction factor ( )
1
6
10
Initial Contact
u
t
0
Assume Slipping
Mode
Assume Sticking
Mode
Determine Solution
of Next Iteration
Remain in Slipping Mode if:
f
t
u
t
0 < and u
t
>
Remain in Sticking Mode if:
f
t
f
n
Change to Sticking Mode if:

f
t
u
t
0 > and u
t
>
or if u
t
0
Change to Slipping Mode if:
f
t
f
n
>
No Yes
549
CHAPTER 8
Contact
8
Contact
When a node is in slipping mode, the contribution to the stiffness matrix is:
while in sticking mode constraints are applied which exactly enforce a zero relative tangential displacement. It should
be noted that in slipping mode the current friction force is not only a function of the current normal force and relative
displacement, but also of the friction force during the previous iteration. This generally improves the stability of the
nonlinear solution process.
As opposed to the procedure based on the arctangent model, this procedure requires additional friction-based testing.
Changes in the friction status as well as too large changes in the friction forces compared to the previous iteration may
cause the program to perform an additional iteration. In 2-D, the latter implies that the friction force should fulfil1:
where is the friction force of the previous iteration and is the above mentioned friction force tolerance, with a
default value of 0.05. In 3-D, in addition to this requirement on the magnitude of the friction force, a check is
performed on the change in direction of the friction force vector:
in which is the angle between the friction vector in the current and the previous iteration.
In order to limit the amount of friction-related iterations, the above mentioned requirements will not cause extra
iterations if the number of nodes not fulfilling the friction requirements compared to the total number of nodes in
contact is small. This can be expressed in the friction force tolerance as:
and must be true for all contact bodies in the model.
If the maximum number of iterations as defined on the CONTROL model definition option has been reached, additional
friction iterations will not be enforced and the analysis will continue with the next increment.
The stick-slip model is always based on nodal forces. This model is not recommended for quadratic elements in 3-D,
because of the varying magnitude and sign of nodal forces of such elements.
When using the PRINT parameter to get extra contact-related information in the output file, the program will echo the
contact status, the ratio of the magnitude of the current and previous friction force vector (and in 3-D also the cosine
of the angle between the current and previous friction force vector), as well as the total number of nodes with a friction
contribution and the number of nodes with a friction contribution which do not fulfil the friction tolerance criteria.
K
i j
f
t
i
u
t
j
------------- =
1 e
f
t
f
t
p
----- 1 e +
f
t
p
e
1 e
f
( ) cos 1 e +
f
number of nodes which do not fulfil friction tolerances
total number of nodes in contact
----------------------------------------------------------------------------------------------------------------------------------- e <

Bilinear Model
The third procedure is called the bilinear model. Similar to the modified step function model, it is based on relative
tangential displacements. Instead of defining special constraints to enforce sticking conditions, the bilinear model
assumes that the stick and slip conditions correspond to reversible (elastic) and permanent (plastic) relative
displacements, respectively. The clear resemblance with the theory of elasto-plasticity will be used to derive the
governing equations.
First, Coulombs law for friction is expressed by a slip surface :
The stick or elastic domain is given by , while is physically impossible.
Next, the rate of the relative tangential displacement vector is split into an elastic (stick) and a plastic (slip)
contribution according to:
(8-1)
and the rate of change of friction force vector is related to the elastic tangential displacement by:
(8-2)
in which matrix is given by:
with the slip threshold or relative sliding displacement below which sticking is simulated (see Figure 8-26). The
value of is by default determined by the program as 0.0025 times the average edge length of the finite elements
defining the deformable contact bodies.
Now attention is paid to the case that, given a tangential displacement vector, the evolution of would result in a
physically impossible situation, so . This implies that the plastic or slip contribution must be determined.
Equations (8-1) and (8-2) yield:
f
t
f
n
=
0 < 0 >
u
t
u
t
e
u
t
p
+ =
f
t
Du
t
e
=
D
D
f
n
-------- 0
0
f
n
--------
=
f
t
0 >
f
t
D u
t
u
t
p
( ) =
551
CHAPTER 8
Contact
Figure 8-26 Bilinear Model
It is assumed that the direction of the slip displacement rate is given by the normal to the slip flow potential , given
by:
So that, by indicating the slip displacement rate magnitude as :
with the slip surface, , different from the slip flow potential, an analogy to nonassociative plasticity can be
observed.
Since a force point can never be outside the slip surface, it is required that:
In this way, the magnitude of the slip rate can be determined. To this end, the equations above can be combined to:
or:
f
t
u
t
f
n
slip threshold
f
t
=
u
t

f
t
------- =

f
t
------
\ .
| |
T
f
t
0 = =

f
t
------
\ .
| |
T
D u
t

f
t
-------
\ .
| |
0 =

f
t
------
\ .
| |
T
Du
t

f
t
------
\ .
| |
T
D

f
t
-------
---------------------------- =

Utilizing this result, the final set of rate equations reads:
Similar to non-associative plasticity, matrix will generally be non-symmetric. In Marc, a special procedure has
been used which results in a symmetric matrix, while maintaining sufficient numerical stability and rate of
convergence.
The bilinear model also uses an additional check on the convergence of the friction forces, which has been achieved
if the following equation is fulfilled:
where is the current total friction force vector (the collection of all nodal contributions), is the total friction
force vector of the previous iteration and is the friction force tolerance, which has a default value of 0.05.
In the output file one can find, per deformable contact body, information about the current total friction force, the
current total normal force and the total friction force change, compared to the previous iteration.
If a node comes into contact and the structure is still stress-free, then the friction stiffness matrix according to the
derivation above will still be zero. This could result in an ill-conditioned system during the next solution of the global
set of equations. To avoid this problem, the initial friction stiffness will be based on the average contact body stiffness
(following from the trace of the matrix defining the material behavior), which is determined during increment 0 of the
analysis. Alternatively, the user can define this initial friction stiffness via the PARAMETERS model definition or
history definition option.
Limitations of the Coulomb Model
When the normal force or stress becomes large, the Coulomb friction model may not correlate well with experimental
observations. This is caused by the fact that the Coulomb model predicts that the frictional shear stresses increase to
a level that can exceed the flow stress or the failure stress of the material. As this is not physically possible, a different
friction calculation should be applied. The choices are either to have a nonlinear coefficient of friction, or to introduce
the friction stress limit in the bilinear model or to use the shear based friction model.
f
t
D
D

f
t
-------

f
t
------
\ .
| |
T
D

f
t
------
\ .
| |
T
D

f
t
-------
---------------------------------
\ .
|
|
|
|
| |
u
t
D D
*
( )u
t
= =
D
*
F
t
F
t
p
F
t
--------------------------------- e
F
t
F
t
p
e
553
CHAPTER 8
Contact
Figure 8-27 Linear Coulomb Model versus Observed Behavior
A nonlinear friction coefficient can be defined via UFRIC user subroutine or via table driven input.
The friction stress limit for the bilinear model ( ) can be entered via the CONTACT TABLE option and is used to
bound the maximum friction stress, based on the assumption that the extrapolated and averaged shear (friction) stress
in a node is proportional to the applied shear (friction) force (see Figure 8-28). If the shear stress reaches the limit
value, then the applied friction force is reduced, so that the maximum shear stress is given by .
Finally, instead of using an adapted Coulomb friction model, one may use the shear friction model, which is the second
idealistic friction model in Marc.
Figure 8-28 Friction Stress Limit used by Bilinear Model
Shear Friction
The shear based model states that the frictional stress is a fraction of the equivalent stress in the material:
(stick) and (slip)
where is the friction factor.
Given the similarity with the Coulomb model, this shear model is implemented using two approximations of the
theoretical step function. First, the arctangent function to smooth out the step function is applied:
Linear Coulomb Model

Observed Behavior
t
l i mi t
mi n
n

t
l i mi t
, ( )
n
Friction Stress Limit
t
m

3
------- <
t
m

3
------- t =
m
t
m

3
-------
2
--- arctan
v
r
RVCNST
-------------------------
\ .
| |
t =

This model is available for all elements using the distributed load approach.
Second, the bilinear model has been adapted based on the assumption that the shear (friction) stress in a node is
proportional to the applied shear (friction) force. Similar to the friction stress limit, the shear stress due to friction is
limited by:
Friction Coefficient
When a node contacts a rigid body, the coefficient of friction associated with the rigid body is used. When a node
contacts a deformable body, the average of the coefficients for the two bodies are used. The CONTACT TABLE option
can be used if complex situations occur.
Recalling that friction is a complex physical phenomenon, due to variations in surface conditions, lubricant
distribution, and lubricant behavior, relative sliding, temperature, geometry, and so on, it was decided to implement
the above two friction models (Coulomb and shear), and to allow you to extend them, if necessary, by means of the
UFRIC user subroutine or by referencing a table. In such a routine, you provide the friction coefficient or the friction
factor as
or
where
Anisotropic Friction Model
The anisotropic friction model implemented in Marc allows for different friction coefficients in two orthogonal
directions on the contact surface. These directions are referred to as slip directions. This model is an extension of the
existing bilinear friction model in Marc. Anisotropic friction model can be expressed in mathematical and graphical
form as shown below.
(8-3)
is the position of the point at which friction is being calculated
is the normal force at the point at which friction is being calculated
is the temperature at the point at which friction is being calculated
is the relative sliding velocity between point at which friction is being calculated and surface
is the equivalent stress at the point at which friction is being calculated
t
mi n m
n
m

3
------- ,
\ .
| |
=
x f
n
T v
r
, , , , ( ) =
m m x f
n
T v
r
, , , , ( ) =
x
f
n
T
v
r
y
f
t 1
1
------
\ .
| |
2
f
t 2
2
---------
\ .
| |
2
+
1
2
---
f
n
555
CHAPTER 8
Contact
Figure 8-29 Slip Surface for Anisotropic Friction Model in Terms of Nodal Contact Forces
For anisotropic friction, the user needs to define two friction coefficients ( and ) and the direction associated
with the first friction coefficient ( ). This direction is henceforth referred to as the first slip direction.
For any contact pair, the anisotropic friction can be activated using CONTACT TABLE option. It should be noted that,
with anisotropic friction model, the friction coefficients specified through the CONTACT TABLE option are only used
in the simulation. The friction coefficients defined on the contact table can be associated with a table ID or can be
modified by using the UFRIC user subroutine.
The first slip direction definition is specified for an individual body through the CONTACT option. For the anisotropic
friction behavior, at least one of the contact bodies should have orientations defined. If the orientations are defined on
both the contact bodies, then the touching body orientations are used. To facilitate the definition of the slip direction,
five orientation options are provided in Marc:
1. COORDINATE SYSTEM:
In this option, the user can define a coordinate system and select the axis which defines the first slip direction.
2. CURVE:
In this option, the user can assign a list of NURBS curves to the contact body on which slip directions are to be
computed. To compute the slip direction at a node, the closest point on any of the given curves is automatically
detected and the first slip direction is given by the direction of the curve at this point.
3. 1-D ELEMENT DIRECTION:
For a deformable touching body made up of 1-D elements like beams or cables, the first slip direction is
computed as the vector from node one to node two. The slip direction is updated based on the displacements
of the two nodes.

1
fn
2
fn
1

2
1

4. SHARP EDGE DIRECTION:
This orientation option is for special applications wherein a higher sliding resistance is expected at the sharp
edges of a structure due to its penetration in underlying soft material. For example, a underwater cable resting
on the sea floor. In this option, the set of element edges with discontinuous outward normals define the first
slip direction. The second slip direction is computed by the cross product of the first slip direction with the
contact normal direction. As shown in fig. 8-30 the first slip direction is selected such that the second slip
direction is aligned with the average normal of faces attached to the sharp edge. With this option, the second
coefficient of friction is used only for the sharp edge moving outwards (i.e., in positive second direction).
Figure 8-30 First and Second Slip Direction
5. USER SUBROUTINE (UCONORT):
In this option the user can write a subroutine to return the unit vector in the first slip direction. It allows the user
to define the first slip direction at a touching node based on the updated nodal positions. All the required inputs
are passed through the argument.
Updating Slip Directions
The slip direction defined using the first two options are mapped on the contact segments at the start of the analysis.
As the associated contact segment deforms or rotates, the slip direction is automatically updated based on the current
orientation of the segment. For orientations defined on rigid contact bodies, the orientations are updated based on the
current position of the rigid body. For option 3, the slip direction is updated based on the displacement of the two
nodes. For option 4, the slip directions are updated based on the displacement of the nodes belonging to the sharp
edges. For option 5, it is the users responsibility to specify any updates to the slip direction.
Additional Output
In addition to standard contact output, additional post quantities like relative slip, frictional forces and frictional
stresses along the first and second slip directions are made available using nodal post codes 82, 83 and 84 respectively.
A vector plot of the slip directions can be visualized using nodal post codes 84 and 85.
(b) Second Slip Direction: Perpendicular to Sharp Edge (a) First Slip Direction: Along Sharp Edge
557
CHAPTER 8
Contact
Limitations
Currently, the anisotropic friction model can be used with node-to-segment contact algorithm only.
Anisotropic friction is only available with the bilinear friction model (for example, IFRIC = 6 or 7).
For every contact pair, we have two surfaces coming into contact. The anisotropic friction is defined by the
properties of one of these surfaces; if both surfaces have anisotropic properties, the properties of the touching
surface will be used.
With the ADAPT GLOBAL or APADTIVE (local) remeshing option, the update in slip direction is not possible.
In that case, only a user subroutine option can be used.
Glue Model
A special type of friction model is the glue option, which imposes that there is no relative tangential motion. The glue
motion is activated through the CONTACT TABLE option.
A novel application of contact is to join two dissimilar meshes. In such a case, by specifying that the glue motion is
activated, the constraint equations are automatically written between the two meshes. Unless a flag has been set, nodes
in contact via the glue option are not allowed to separate.
When a node of a shell or beam element is glued to a load controlled rigid body or to the face of a shell or solid element,
the rotation of the shell or beam can be suppressed. This allows a true moment carrying glued connection. For the case
that a connection is made to the face of a solid element, the rotations of the touching node are connected to the
translations of the nodes of the contacted patch by a constraint relation based upon the large rotation RBE3 theory. The
moment carrying feature is optional and is activated through the CONTACT TABLE option.
Deact Glue
Another option related to the glue option id DEACT GLUE. Suppose two bodies are glued together and a part of the
interface should have standard contact (in order to model, for example, a crack), the nodes in the region that should
have standard contact are then identified with DEACT GLUE. Without this option, one would have to split up the
contact bodies in order to model this case.
Breaking Glue
There is a capability on the CONTACT TABLE option to break up the glued connection using a stress criterion. When
the following criterion is fulfilled at a node, the glued contact is released:
(8-4)
Here is the contact normal stress, the contact tangential stress, and , , m, and n are user-defined
parameters. A user-defined criterion can be given with the UBREAKGLUE user subroutine.
When a node is released due to breaking glued, it changes status from being glued to standard contact permitting
separation and friction.
The contact stresses are calculated using extrapolated stresses for solid elements and as contact force divided by
equivalent area for shell elements.
n
S
n
------
\ .
| |
m

t
S
t
-----
\ .
| |
n
+ 1 >
n

t
S
n
S
t

Nodal post codes are available for postprocessing the separate terms in Equation (8-4) as well as the sum. A nodal post
code for the deact glue status is also available.
Separation
After a node comes into contact with a surface, it is possible for it to separate in a subsequent iteration or increment.
Mathematically, a node should separate when the reaction force between the node and surface becomes tensile or
positive. Physically, you could consider that a node should separate when the tensile force or normal stress exceeds
the surface tension. Rather than use an exact mathematical definition, you can enter the force or stress required to cause
separation.
Separation can be based upon either the nodal forces or the nodal stresses. The use of the nodal stress method is
recommended as the influence of element size is eliminated.
When using higher-order elements and true quadratic contact, the separation criteria must be based upon the stresses
as the equivalent nodal forces oscillate between the corner and midside nodes.
In many analysis, contact occurs but the contact forces are small; for example, laying a piece of paper on a desk.
Because of the finite element procedure, this could result in numerical chattering. Marc has some additional contact
control parameters that can be used to minimize this problem. As separation results in additional iterations (which
leads to higher costs), the appropriate choice of parameters can be very beneficial.
When contact occurs, a reaction force associated with the node in contact balances the internal stress of the elements
adjacent to this node. When separation occurs, this reaction force behaves as a residual force (as the force on a free
node should be zero). This requires that the internal stresses in the deformable body be redistributed. Depending on
the magnitude of the force, this might require several iterations.
You should note that in static analysis, if a deformable body is constrained only by other bodies (no explicit boundary
conditions) and the body subsequently separates from all other bodies, it would then have rigid body motion. For static
analysis, this would result in a singular or nonpositive definite system. This problem can be avoided by appropriate
boundary conditions. When the direct solvers are used, the AUTOSPC parameter may also be used to suppress singular
modes.
Release
A special case of separation is the intentional release of all nodes from a rigid body. This is often used in manufacturing
analysis to simulate the removal of the workpiece from the tools. After the release occurs in such an analysis, there
might be a large redistribution of the loads. It is possible to gradually reduce the residual force to zero, which improves
the stability, and reduces the number of iterations required. The RELEASE history definition option allows the release
(separation) of all the nodes in contact with a particular surface at the beginning of the increment. The rigid body
should be moved away using the MOTION CHANGE option or deactivated using the CONTACT TABLE option to
ensure that the nodes do not inadvertently recontact the surface they were released from.
Glue Contact Conditions
In a contact analysis of the node-to-segment type, a distinction is made between touching and glue conditions. In a
structural analysis, a touching condition triggers the local application of a nonpenetration constraint still allowing
relative sliding of the bodies in the contact interface. The nonpenetration constraint is applied through a tying on the
559
CHAPTER 8
Contact
displacement components normal to the contact surfaces. A glue condition suppresses all relative motions between
bodies through tyings applying them to all displacement degrees of freedom of the nodes in contact. For elements with
nodes that also have rotational degrees of freedom the rotations may be coupled additionally to provide a moment
carrying glue capability. The activation of the glue conditions and their type is controlled by appropriate settings on
the CONTACT TABLE input.
In most nonstructural analysis types, such as thermal or mass diffusion, the touching condition by default does not
trigger the application of tyings, but allows interaction in the contacting interface through the use of a contact interface
coefficient. For small or zero values of the interface coefficient the bodies practically behave as being insulated. For
high values the interface coefficient acts as a penalty constraint. Alternatively the user may request glue conditions
meaning the degrees of freedom in the contact interface are being coupled through tyings as in the structural case. This
has the advantage that the user does not have to choose a high interface coefficient if ideal transport conditions are
desired over the contact interface (e.g., a high heat transfer coefficient for ideal heat transport in case of a heat transfer
analysis or a high mass transfer coefficient for ideal mass transport in case of a mass diffusion analysis) and it becomes
relatively easy to connect dissimilar meshes. The activation of the glue conditions is controlled by appropriate settings
on the CONTACT TABLE input.
The electrostatic analysis type does not support contact interface coefficients and, by default tyings, are applied to the
electric potential degrees of freedom of the nodes in the contacting interfaces. For this case, there is no difference
between touching or glue conditions. It is possible, however, to disconnect bodies electrically by making appropriate
settings on the CONTACT TABLE.
In a three-dimensional magnetostatic analysis, the touching condition means that the tangential components of the
magnetic vector potential in the contact interface are coupled through tyings, but the normal components are not. The
glue condition means that all components of the magnetic vector potential are coupled through tyings. In a 2-D
analysis, the one nonzero component of the magnetic potential is always coupled through a tying constraint and there
is no difference between touching or glue conditions.
In a weakly coupled multi-physics analysis where the equations are being solved in a staggered way, it may be
desirable to have touching conditions in one physics pass, but glue conditions in the other. For example, in a coupled
thermo-mechanical analysis, it may be desirable to have touching conditions in the structural pass allowing relative
sliding between the contact bodies, but glue conditions in the thermal pass simulating ideal heat transfer over the
contact interface. In such cases, the touching or glue conditions may be requested on the CONTACT TABLE for each
physics pass separately.
In the strongly coupled analysis types that are supported in Marc (i.e., piezoelectric analysis and soil/pore pressure
analysis), the glue condition simultaneously couples all degrees of freedom of both physics types through tyings and
it is not possible to treat the different physics types separately.
Table 7-2 summarizes the meaning of touching and glue conditions for each possible physics type that is supported in
Marc. The exceptions are fluid and electromagnetic analysis where the cells are blank indicating that these analysis
types do not support contact.

Table 7-2 Contact Condition Summary per Physics Type
Physics Type Pass # Touching Condition Glue Condition
Structural 1 Tyings on normal displacements.
Tangential displacements are free,
unless there is friction.
Tyings on all displacement degrees of
freedom.
Moment carrying if desired and
applicable.
Thermal 2 No tyings. Heat transfer may be
controlled through a contact heat
transfer coefficient.
Tyings on the temperatures.
Fluid 3
Electrical in Joule
heating
4 No tyings. Electric current may be
controlled through a contact
electrical coefficient.
Tyings on the voltages.
Mass diffusion 5 No tyings. Mass transfer may be
controlled through a contact mass
transfer coefficient.
Tyings on the pressures.
Electrostatic 6 Tyings on the electric potential
degrees of freedom. There is no
contact interface coefficient.
Tyings on the electric potential
degrees of freedom.
Magnetostatic 7 2-D: Tyings on the non-zero
magnetic potential degrees of
freedom. There is no contact
interface coefficient.
2-D: Tyings on the nonzero magnetic
potential degrees of freedom.
3-D: Tyings on the tangential
magnetic vector potential degrees of
freedom. Normal components are
free.
There is no contact interface
coefficient.
3-D: Tyings on all magnetic vector
potential degrees of freedom.
Electromagnetic 8
Piezoelectric 1 Tyings on normal displacements.
Tangential displacements are free.
degrees of freedom.
freedom.
degrees of freedom.
Soil/pore pressure 1 Tyings on normal displacements.
Tangential displacements and
pressure are free.
freedom.
Tyings on the pressure degrees of
freedom.
561
CHAPTER 8
Contact
Coupled Analysis
Contact analysis has several consequences on performing a coupled analysis. When performing a coupled contact
analysis between multiple deformable bodies, each body deforms due to mechanical and thermal loads and undergoes
heat conduction. When the bodies come into contact, there is heat flux across the surfaces. You need to provide a
coefficient of heat transfer between the surfaces. This is often quite significant as a hot workpiece might come into
contact with a cold tool set. Additionally, if friction is present, heat is generated based upon:
where is the mechanical equivalent of heat. Half of the heat generated due to friction is contributed to each body.
If the model allows a body to be rigid, two options are available. In the first case, the rigid body is considered to have
a constant temperature (heat source or heat sink). Heat flux is exchanged between the rigid body and the deformable
body. In the second case, it is required to perform a heat transfer analysis in the rigid body. In this case, the rigid body
must be constructed of finite elements, but they are chosen as heat transfer types. Deformation does not occur in this
body, but a heat transfer analysis is performed. This is computationally more efficient than performing a coupled
analyses where all bodies are deformable. Symmetry bodies have different characteristics as they are frictionless
(hence, no heat generated) and because there is no heat flux across a symmetry plane.
Heat fluxes (FILMS) are automatically created on all the boundaries of the deformable-bodies. Three types of heat
fluxes can be distinguished, where each flux has its own physical characteristic. The type of flux Marc uses, depends
on the distance between the contacting bodies (see Figure 8-31). We define the distance, , as the distance
below which the two bodies are touching each other. For a distance smaller than , the first flux type is used.
The second flux type is used when the distance is between and , where is called the near contact
distance. This distance represents a small gap between the contact bodies, and should not be chosen larger than the
smallest element size. If this distance is not set, this type of flux is disregarded. The third flux type is used when the
distance is larger then .
Figure 8-31 Contact Distance
If , the bodies touch and conduction takes place between the two contacting bodies. The heat flux is
written as:
q f
f r
v
r
M
eq
=
M
eq
d
cont act
d
cont act
d
cont act
d
near
d
near
d
near
d
T
1
T
2
d d
cont act
<
q H
TC
T
2
T
1
( ) =

where
If , a small gap between the bodies exists. The heat flux that exists through this gap can be
decomposed in the following physical processes: convection, natural convection, radiation, and a distance dependent
term. This flux is given by:
where
If , or has not been defined and , the bodies are not touching and a heat flux to the
environment exists in the following form,
where
is the surface temperature,
is the film coefficient between the two surfaces,
is the temperature of the same contact location, as obtained from interpolation of nodal
temperatures of the body being contacted.
is the convection coefficient for near field behavior.
is the natural convection coefficient for near field behavior.
is the exponent associated with natural convection.
are the coefficients for radiation, the Stefan-Boltzman coefficient, the emissivity, and the viewfactor,
respectively. The Stefan-Boltzman constant is given via the PARAMETERS option.
is the separation distance dependent heat transfer coefficient.
is the temperature in absolute scale. The offset between the user temperature and the
absolute temperature is given in the PARAMETERS option.
is the heat transfer coefficient to the environment
is the environment sink temperature
T
1
H
TC
T
2
d
cont act
d d
near
< <
q H
CV
T
2
T
1
( ) H
NC
T
2
T
1
( )
B
NC
f T
A2
4
T
A1
4
( )
H
CT
1
d
d
near
------------ -
\ .
| |
H
BL
d
d
near
------------ - +
)
`

T
2
T
1
( )
+ +
+
=
H
CV
H
NC
B
CN
, , and f
H
BL
T
A2
T
2
T
ABSREF
+
d d
near
> d
near
d d
cont act
>
q H
CTVE
T
SI NK
T
1
( ) f T
A2
4
T
A1
4
( ) + =
H
CTVE
T
SI NK
563
CHAPTER 8
Contact
When a deformable-body contacts a rigid-body, the coefficients associated with the rigid-body are used. When two
deformable bodies are in contact, the average value of the film coefficient specified for the bodies is used. This only
works for the coefficients discussed in the first and third heat flux type. A better way to prescribe the coefficients is to
use the CONTACT TABLE option. With this option, all the previously discussed coefficients can be set for each
interface separately. For the second heat flux type, this is the only way to specify the coefficients. As with all other
coupled problems, heat generated by plastic deformation can be calculated and applied as a volumetric flux. The heat
generated by friction is also calculated and applied as a surface flux.
Three user subroutines are available to facilitate the creation of more sophisticated boundary flux definitions (such as
radiation and convections with variations in space, temperature, pressure, etc.). UHTCON allows you to specify a film
coefficient while the surface is in contact ( ), UHTNRC allows you to specify a film coefficient while the
surface is almost in contact ( ), and UHTCOE allows you to specify a film coefficient while the
surface is free ( ).
Thermal Contact
A thermal contact analysis is performed when no structural pass is present, this results in a computationally more
efficient analysis. Possible analyses types are Heat Transfer, Coupled Thermal-Electrical, Magnetostatics and
Electrostatics. For Heat Transfer and Coupled Thermal-Electrical the analysis is comparable to a coupled analysis. The
two types of rigid bodies that are described in the coupled analysis are allowed. So one is a rigid which is considered
to have a constant temperature (heat source or heat sink), and the other is a finite element body with thermal properties.
The three types of heat fluxes as described in Coupled Analysis can also be used here.
The coefficients are prescribed in the same way as in a Coupled Analysis where it is advised to use the CONTACT
TABLE option.
The three user subroutines UHTCON, UHTNRC, and UHTCOE as previously discussed in Coupled Analysis can also
be used in a Heat Transfer and Coupled Thermal-Electrical analysis.
Joule Heating
In a Joule heating analysis similar conditions as thermal contact apply regarding the flow of current between
contacting bodies.
If , the bodies touch and electrical conduction takes place between the two contacting bodies. The current
is written as:
or
where
is the surface voltage,
is the electrical contact conductivity,
d d
cont act
<
d
cont act
d d
near
< <
d d
near
>
d d
cont act
<
i H
EC
V
2
V
1
( ) = H
EC
V
body
V
1
( )
V
1
H
EC

If , a small gap between the bodies exists. The current that exists through this gap can be
decomposed into:
where
If , or has not been defined and , the bodies are not touching and a current leakage to
the environment exists in the following form,
where
Electrical contact between bodies does contribute a term to the heat generation.
Note that near electrical contact for Joule heating does not represent electrical arcing between bodies, which is a much
more complex phenomena.
In a Joule heating analysis, the THERMAL CONTACT model definition option is also used to identify elements which
are in a conducting body to perform a resistance calculation using EMRESIS history definition option.
Three user subroutines are available to facilitate the creation of more sophisticated boundary definitions. UVTCON
allows you to specify a film coefficient while the surface is in contact ( ), UVTNRC allows you to specify
a film coefficient while the surface is almost in contact ( ), and UVTCOE allows you to specify a
film coefficient while the surface is free ( ).
Contact in an Electrostatic or Piezoelectric Analysis
In an electrostatic analysis, if two bodies are in contact, the scalar potential is tied between the bodies composed of
elements. For an electrostatic analysis the THERMAL CONTACT option is used. When a node is in contact with a
symmetry surface, it will be treated as a free surface, that is no current transmitted through the symmetry surface. The
is the voltage of the same contact location, as obtained from interpolation of nodal voltages of the body
being contacted.
is the voltage of the rigid body being touched.
is the coefficient for near field behavior.
is the separation distance dependent conductivity coefficient.
is the coefficient to the environment
is the environment sink voltage
V
2
V
Body
d
cont act
d d
near
< <
i H
NE
V
2
V
1
( ) H
EC
1
d
d
near
-------------
\ .
| |
H
BLE
d
d
near
------------ - +
)
`

V
2
V
1
( ) + =
H
NE
H
BLE
d d
near
> d
near
d d
cont act
>
q H
VE
V
SI NK
V
1
( ) =
H
VE
V
SI NK
d d
cont act
<
d
cont act
d d
near
< <
d d
near
>
565
CHAPTER 8
Contact
THERMAL CONTACT option is also used to identify the elements which are in a conducting body to perform a
capacitance calculation using EMCAPAC history definition option. In such cases, the conducting bodies should not be
in contact. Cyclic symmetry may not be used in a pure electrostatic analysis in this release.
Contact in Magnetostatic Analysis
For 3-D magnetostatics, by default the tangential components of the magnetic vector potential are tied for nodes in
contact. When the glue flag is set on the CONTACT TABLE input all the components of the magnetic vector potential
are tied. For 2-D magnetostatics the one non-zero component of the magnetic vector potential gets tied in contact
irrespective of the value of the glue flag.When a node is in contact with a symmetry surface, a constraint is imposed
such that only the normal component of the potential is zero. For a magnetostatic analysis, the THERMAL CONTACT
option is used. Cyclic symmetry may not be used in a pure magnetostatic analysis in this release.
Contact in Magnetodynamic Analysis
Contact is not currently supported in harmonic or transient magnetodynamic analysis.
Contact in Soil Analysis
When two deformable bodies are in contact, the displacements are treated in the conventional manner. The pressure
degree of freedom associated with the fluid hydrostatic pressure is tied when the glue flag is set on the CONTACT
TABLE input. If a node contacts a rigid body, nothing is done with the pressure degree of freedom.
Contact in an Acoustic Analysis
In an acoustic analysis with flexible and/or rigid walls composed of finite elements and/or rigid bodies, a coupling
interface between the acoustic medium and the walls is created to connect the fluid pressure to the wall movement
taking into account accelerations and reactive boundary conditions.
Element Considerations
Marc allows contact with almost all of the available elements, but the use of certain elements has a consequence on
the analysis procedure. Contact analysis can be performed with all of the structural continuum elements, either lower
order or higher order, including those of the Herrmann (incompressible) formulation, except axisymmetric elements
with twist. Friction modeling is available in all of these elements except the semi-infinite elements. Higher-order
isoparametric elements use shape functions which, when the elements are loaded by a (for example) uniform pressure,
lead to equivalent nodal loads that oscillate between the corner and midside nodes. This has a detrimental effect on
determining contact separation and two procedures have been implemented to eliminate this problem:
1. Linearized Contact: The midside nodes on the exterior surface are automatically tied to the corner nodes in this
case. This effectively results in a linear variation of both the geometry and the displacement on the exterior
element edges. Hence, the elements with edges on the exterior do not behave as full bi-quadratic (2-D) or tri-
quadratic (3-D) elements.

All elements in the interior of the body behave in the conventional higher-order manner, but the constraints on
the exterior easily cause the behavior of the complete structure to be too stiff; while in the area of contact, the
stress distribution might be irregular.
2. True Quadratic Contact: No special constraints introduced on the exterior surface other than coming from
contact in this case. Both the midside and the corner nodes may come into contact and when contact is
established with another deformable body consisting of quadratic elements, a constraint equation
corresponding to the complete quadratic shape function is automatically incorporated. Since the above
mentioned oscillating nodal loads cannot be used for separation, the decision whether or not a node should
separate is based on the contact normal stress rather than the contact normal force.
In many manufacturing and rubber analyses, the lower-order elements behave better than the higher-order elements
because of their ability to represent the large distortion; hence, these lower-order elements are recommended.
The constraints imposed on the nodal degrees of freedom are dependent upon the type of element.
1. When a node of a continuum element comes into contact, the translational degrees of freedom are constrained.
2. When a node of a shell element comes into contact, the translational degrees of freedom are constrained and
no constraint is places on the rotational degrees of freedom. The exception to this is when a shell contacts a
symmetry surface. In this case, the rotation about the element edge is also constrained.
Additionally, beams and shells contact is governed by the rules outlined below.
2-D Beams
All nodes on beams are potential contact nodes. Beam elements can be used in contact in two modes.
1. The two-dimensional beams can come into contact with rigid bodies composed of curves in the same x-y plane.
The normal is based upon the normal of the rigid surface.
2. The two-dimensional beams can come into contact with deformable bodies either of continuum elements or
other beam elements. As the beams are in two dimensions, they do not intersect one another.
3-D Beams
Three-dimensional beam elements can be used in contact in three methods.
1. The nodes of the beams can come into contact with rigid bodies composed of surfaces. The normal is based
upon the normal of the rigid surface.
2. The nodes of the beam and truss elements can also come into contact with the faces of three-dimensional
continuum elements or shell elements. The normal is based upon the normal of the element face.
3. The three-dimensional beam and truss elements can also come into contact with other beam or truss elements
(beam-to-beam contact).
The first two methods are activated by default if contact bodies consisting of beam or truss elements are defined using
the CONTACT option. The third method must be activated explicitly by additionally switching on the beam-to-beam
contact flag on the CONTACT option.
567
CHAPTER 8
Contact
In the beam-to-beam contact model, a beam or truss element is viewed as a conical surface with a circular cross-
section. The radius of the cross-section can vary linearly between the start and end node of a beam element. For each
beam or truss element of a contact body, a contact radius must be entered via the GEOMETRY option. The contact
radius at a node follows from the average contact radius of the elements sharing that node. Hence, the start and end
node of an element may have different contact radii.
Contact is detected between two beam or truss elements if the associated conical surfaces touch each other; that is, if
the distance between the closest points on the conical surfaces is smaller than the distance below which bodies are
considered touching each other. This is outlined in Figure 8-32, where beam elements and their contact body
representation are given. It should be emphasized that the contacting points are points on the conical surfaces and not
nodes of the finite element model.
Figure 8-32 Beam-to-Beam Contact: Finite Element Model (top) and Contact Body Representation (bottom)
If two beam or truss elements are in contact, a multipoint constraint equation (tying) is automatically imposed to
ensure that the conical surfaces will not interpenetrate. This constraint equation involves the displacements of the
begin and end nodes of both elements. The tied node in that equation is automatically selected, taking into account the
location of the contacting points with respect to the elements, any boundary conditions applied to the nodes and any
contact between the nodes and rigid surfaces or faces of continuum or shell elements.
During the iteration procedure, the contacting points of two beam or truss elements can change if the elements slide
with respect to each other. In addition, the points in contact can move from one element to another. In that case, the
nodes involved in the multipoint constraint equations are automatically updated. During sliding, friction may be taken
into account. Since for beam or truss elements the normal stress in the contact points is not available, only Coulomb
friction based on nodal forces (either the arctangent or the bilinear model) is supported for beam-to-beam contact. The
glue model, which imposes that there is no relative tangential motion, is also available.
A limitation of the beam-to-beam contact model is that a contact body cannot contain branches, that is, every element
in the contact body must have a unique successor and predecessor.
d
d
X
Y
Z

Shell Elements
All nodes on shell elements are potential contact nodes. As the midside nodes of shell elements are automatically tied,
the high-order shell element (type 22) has no benefit. Shell elements can contact either rigid bodies, continuum
elements, or other shell elements. Shell-shell contact involves a more complex analysis because it is necessary to
determine which side of the shell contact occurs.
By default, the top and bottom face of a shell element (which follow from the mid face by using the thickness offset
vector) are separately taken into account. This implies that a node may be found to be in contact based on the top or
the bottom face of an element. However, due to the nature of the shell formulation, it is not possible to simultaneously
use contact conditions following from the top and the bottom face. Touching a shell element may also occur at its top
or bottom face.
The user has the option to set per pair of contact bodies on the CONTACT TABLE option how the geometry of the shell
elements should be handled. This means that both for the contacting and the contacted body it can be defined that the
contact description will be based on one of the following options: both top and bottom (default), top only (with or
without the thickness offset vector) or bottom only (with or without the thickness update vector). For glued contact it
is possible to use the top and bottom face, but without the thickness update vector.
The beam-to-beam contact model discussed above can also be used to model shell edge-to-edge contact. This requires
switching on the beam-to-beam contact flag on the CONTACT option and, for a shell contact body combination, the
edge-to-edge contact flag on the CONTACT TABLE option. For edge-to-edge contact, half of the thickness of the shell
is used to set the contact radius.
Dynamic Impact
The capability is available for both implicit time integration via the Newmark-beta or Houbolt operators, and the
explicit time integration via the (fast) central difference operator.
The Newmark-beta, Single Step Houbolt and Generalized Alpha procedure have the capability to allow variable time
steps and, when using the user-defined fixed time step procedure, the time step is split by the algorithm to satisfy the
contact conditions. Since the central difference operator is constrained to small time steps by the stability requirement,
time splitting is not used by the contact algorithm.
For most dynamic impact problems, the Single Step Houbolt method is recommended, as this procedure possesses
high-frequency dissipation. This is often necessary to avoid numerical problems by contact-induced high-frequency
oscillations. If the other dynamic operators are used, it is recommended that numerical damping is used during the
analysis, or, when using the Generalized Alpha method, that a small spectral radius is defined. Note that for a zero
spectral radius, the Generalized Alpha method is identical to the Single Step Houbolt method with default parameters
( , ).
In dynamic analysis, the requirement of energy conservation is supplemented with the requirement of momentum
conservation. In addition to the constraints placed upon the displacements, additional constraints are placed on the
velocity and acceleration of the nodal points in contact, except for the Single Step Houbolt and Generalized Alpha
method.
When a node contacts a rigid surface, it is given the velocity and acceleration of the rigid surface in the normal
direction. The rigid surfaces are treated as if they have infinite mass, hence, infinite momentum.
1
1.5 = 0.5 =
569
CHAPTER 8
Contact
Global Adaptive Meshing and Rezoning
In manufacturing simulations the objective is to deform the workpiece from some initial (simple) shape to a final, often
complex shape. This deformation of the material, results in mesh distortion. For this reason, it is often required to
perform a rezoning/remeshing step. At this point, a new mesh is created; the current state of deformation, strains,
stresses, etc. is transferred to the new mesh, and the analysis is continued. This has several consequences for contact
analysis.
1. When manual rezoning is used, it is necessary to redefine the elements associated with the deformable body
that is being rezoned. This is done through the CONTACT CHANGE option.
2. If a contact tolerance was not defined by you, a new contact tolerance is calculated based upon the new mesh.
This distance can be smaller than the previously calculated tolerance, leading to more iterations.
3. After the new mesh is created, the program redetermines the potential contact nodes. The memory is
automatically allocated to accommodate the contact data in the new mesh.
4. At the first increment after remeshing, Marc determines which nodes are in contact. If the new mesh does not
reflect the exterior surface of the old model, it is possible that a region that was previously in contact is no
longer in contact. This is usually not so serious; as most likely, the node subsequently comes into contact. A
more significant problem is if a node in the new mesh is created which, in fact, has penetrated another body. If
the newly created node is beyond another bodys surface by more than the contact tolerance, it is possible that
a poor contact analysis results. With automatic global remeshing, steps are taken to made sure contact
conditions are preserved and incorrect penetrations are removed in the new mesh. This is achieved by
correcting the nodal position after a new mesh is created.
If rezoning occurs, rigid surfaces must be kept the same and cannot be changed.
Marc auto-rezoning feature, activated by the parameter REZONING,1 for 2-D problems and REZONING,2 for 3-D
problems, automatically creates new meshes and performs rezoning to resolve mesh distortion problems. In such
cases, points (1) and (4) are treated automatically.
Adaptive Meshing
Contact between a deformable-body and a rigid body is insured such that the nodes do not penetrate the rigid surface.
It is possible that an edge of an element penetrates a rigid surface, especially where high curvature is present because
of the finite element discretization. The use of the adaptive mesh generation procedure can be used to reduce these
problems.
In addition to the traditional error criteria such as Zienkiewicz-Zhu or maximum stress, etc., you can request that the
mesh be adaptively refined due to contact. In such a case, when a node comes into contact, the elements associated
with that node are refined. This results in a greater number of elements and nodes on the exterior region where contact
occurs. This can lead to a substantial improvement in the accuracy of the solution.

Figure 8-33 Contact Closure Condition
The adaptive meshing procedure has consequences to the contact procedure similar to the rezoning procedure.
If a contact tolerance was not defined by you, a new contact tolerance is calculated based upon the new mesh. This
distance can be smaller than the previously calculated tolerance, leading to more iterations.
Result Evaluation
The Marc post file contains the results for both the deformable bodies and the rigid bodies. In performing a contact
analysis, you can obtain three types of results. The first is the conventional results from the deformable body. This
includes the deformation, strains, stresses, and measures of inelastic behavior such as plastic and creep strains. In
addition to reaction forces at conventional boundary conditions, you can obtain the contact forces and friction forces
imparted on the body by rigid or other deformable bodies. By examining the location of these forces, you can observe
where contact has occurred, but Marc also allows you to select the contact status as a post file variable:
A value of 0 means that a node is not in contact.
A value of 0.5 means the node is in near thermal contact.
A value of 1 means that a node is in contact.
A value of 2 means the node is on a cyclic symmetry boundary.
It is also possible to obtain the resultant force following from contact on the deformable bodies and the resultant force
and moment on the rigid bodies. The moment is taken about the user-defined centroid of the rigid body. The time
history of these resultant forces are of significant issues in many engineering analysis. Of course, if there is no resultant
force on a rigid body, it implies that body is not in contact with any deformable body.
Finally, if the additional print is requested via the PRINT parameter, the output file reflects information on when a
node comes into contact, what rigid body/segment is contacted, when separation occurs, when a node contacts a sharp
corner, the displacement in the local coordinate system, and the contact force in the local coordinate system. For large
problems, this can result in a significant amount of output.
The motion of the rigid bodies can be displayed as well as the deformable bodies. Rigid bodies which are modeled
using the piecewise linear approach are displayed as line segments for flat patches. When the rigid surfaces are
modeled as analytical surfaces, the visualization appears as piecewise linear in older versions (pre-K7/Mentat 3.1) or
as trimmed NURBS in subsequent versions.
Penetration of Element Edge due to Mesh Discretization
Result of Adaptive Meshing
571
CHAPTER 8
Contact
Tolerance Values
On the third data block for contact, five tolerances can be set for determination of the contact behavior. Not entering
any values here means that Marc calculates values based on the problem specification.
Relative Sliding Velocity Between Surfaces Below Which Friction Forces Drop
As discussed in the Friction Modeling section, the equations of friction are smoothed internally in the program to avoid
numerical instabilities. The equations are inequalities whenever two contacting surfaces stick to each other and
equalities whenever the surfaces slide (or slip). Thus, the character of contact constraints change depending on
whether there is sticking or slipping. The smoothing procedure consist of modifying it in such a way, that there is
always slip; the amount is a function of the relative velocity and a constant RVCNST. The value of this constant must
be specified here. It actually means, that if we specify a small value in comparison to the relative velocity, the jump
behavior is better approximated, but numerical instabilities can be expected. A large value means, that we need a large
relative velocity before we get the force at which the slip occurs.
It is suggested to use values between 0.1 and 0.01 times a typical surface velocity.
Distance Below Which a Node is Considered Touching a Surface
In each step, it is checked whether a (new) node is in contact with other surfaces. This is determined by the distance
between the nodes and the surfaces. Since the distance is a calculated number, there are always roundoff errors
involved. Therefore, a contact tolerance is provided such that if the distance calculated is below this tolerance, a node
is considered in contact. A too large value means that a high number of body nodes are considered to be in contact
with the surface and are consequently all moved to the surface, which can be unrealistic in some applications. A too
small value of this number means that the applied deformation increment is split into a high number of increments,
thus increasing the cost of computation.
The tolerance must be provided by the analyst or can be calculated by Marc. In general, the contact tolerance should
be a small number compared to the geometrical features of the configuration being analyzed. The value calculated by
Marc is determined as 1/20 of the smallest element size of all elements in any contact body or 1/4 of the smallest
thickness of all shell or beam elements in any contact body, whichever is the smallest. Note that for a continuum
element the element size is defined here as the smallest edge of the surrounding rectangle (2D) or box (3D), set up in
the global coordinate system, while for a shell element it is the smallest element edge and for a beam element it is the
element length.
Tolerance on Nodal Reaction Force or Nodal Stress Before Separation Occurs
If a tensile force occurs at a node which is in contact with a surface, the node should separate from the surface. Rather
than using any positive value, a threshold value can be specified. This number should theoretically be zero. However,
because a small positive reaction might be due only to errors in equilibrium, this threshold value avoids unnecessary
separations. A too small value of this force results in alternating separation and contact between the node and the
surface. A too large value, of course, results in unrealistic contact behavior.
Marc calculates this value as the maximum residual force in the structure. Note that the maximum residual force can
be specified in the CONTROL option. Default for this value, a 10 percent of the maximum reaction force is used.
Consequently, if locally high reaction forces at a particular point are present, the separation force is large as well. In
most cases, however, the default value is a good measure.

If you indicate that separation is to be based upon stresses, a value of the separation stress should be entered. The
default value is the maximum residual force at a node divided by the contact area of node n.
Numerical and Mathematical Aspects of Contact
Lagrange Multipliers
In performing contact analysis, you are solving a constrained minimization problem where the constraint is the no
penetration constraint. The Lagrange multiplier technique is the most elegant procedure to apply mathematical
constraints to a system. Using this procedure, if the constraints are properly written, overclosure or penetration does
not occur. Unfortunately, Lagrange multipliers lead to numerical difficulties with the computational procedure as their
inclusion results in a nonpositive definite mathematical system. This requires additional operations to insure an
accurate, stable solution which leads to high computational costs. Another problem with this method is that there is no
mass associated with the Lagrange multiplier degree of freedom. This results in a global mass matrix which cannot be
inverted. This precludes the used of Lagrange multiplier techniques in explicit dynamic calculations which are often
used in crash simulations. The Lagrange multiplier technique has often been implemented in contact procedures using
special interface elements such as the Marc gap element (type 12). This facilitates the correct numerical procedure,
but puts a restriction on the amount of relative motion that can occur between bodies. The use of interface elements
requires an apriori knowledge of where contact occurs. This is unachievable in many physical problems such as crash
analysis or manufacturing simulation.
Penalty Methods
The penalty method or its extension, the Augmented Lagrangian method, is an alternative procedure to numerically
implement the contact constraints. Effectively, the penalty procedure constrains the motion by applying a penalty to
the amount of penetration that occurs. The penalty approach can be considered as analogous to a nonlinear spring
between the two bodies. Using the penalty approach, some penetration occurs with the amount being determined by
the penalty constant or function. The choice of the penalty value can also have a detrimental effect on the numerical
stability of the global solution procedure. The penalty method is relatively easy to implement and has been extensively
used in explicit dynamic analysis although it can result in an overly stiff system for deformable-to-deformable contact
since the contact pressure is assumed to be proportional to the pointwise penetration. The pressure distribution is
generally oscillatory.
Hybrid and Mixed Methods
In the hybrid method, the contact element is derived from a complementary energy principle by introducing the
continuity on the contact surface as a constraint and treating the contact forces as additional elements. Mixed methods,
based on perturbed Lagrange formulation, usually consist of pressure distribution interpolation which is an order less
than the displacement field, have also been used to alleviate the difficulties associated with the pure Lagrange method.
Direct Constraints
Another method for the solution of contact problems is the direct constraint method. In this procedure, the motion of
the bodies is tracked, and when contact occurs, direct constraints are placed on the motion using boundary conditions
both kinematic constraints on transformed degrees of freedom and nodal forces. This procedure can be very accurate
573
CHAPTER 8
Contact
if the program can predict when contact occurs. This is the procedure that is implemented in Marc through the
CONTACT option. No special interference elements are required in this procedure and complex changing contact
conditions can be simulated since no apriori knowledge of where contact occurs is necessary. The details of the
implementation are presented later.
Contact can be defined as finding the displacement of points A and B such that
where A is on one body and B is on another body, n is the direction cosine of a vector between the two points, and
TOL is the closure distance.
Figure 8-34 Normal Gap Between Potentially Contacting Bodies
Lagrange Multiplier Procedure
In the Lagrange Multiplier Procedure, the variational procedure is augmented by the constraint. The virtual work
principle leads to the traditional systems equation:
The constraint conditions can be expressed as
Through the minimization of the augmented functional
you obtain
This equation can be solved simultaneously for both the displacement ( ) and the Lagrange multiplier ( ). Note that
the introduction of Lagrange multipliers results in a zero on the diagonal. Hence, even well posed physical problems
no longer have positive definite systems. This often results in numerical difficulties and restricts the type of linear
equation solver that can be used.
u
A
u
B
( ) n TOL <
A
B
n
Ku f =
Cu 0 =
1
2
--- u
T
Ku u
T
f
T
Cu + =
K C
T
C 0
u
)
`
f
0
)
`

=
u

Direct Constraint Procedure
Marc divides contact problems into two domains; the first is when a deformable body makes contact with a rigid
surface and the second is when a deformable body makes contact with another deformable body or itself.
Two-dimensional or three-dimensional bodies are treated conceptually in the same manner.
Deformable-Rigid Contact
In such a problem, a target node on the deformable body has no constraint while contact does not occur. Once contact
is detected, the degrees of freedom are transformed to a local system and a constraint is imposed such that
where v is the prescribed velocity of the rigid surface. This local transformation is continuously updated to reflect
sliding of point A along the rigid surface.
If the glue option is activated, an additional displacement constraint is formed as
The determination of when contact occurs and the calculation of the normal vector are critical to the numerical
simulation. Historically in Marc, three procedures are used for detecting penetration and imposing contact, but now
two of them are preferred and are discussed here. The first (and most general one) is the Iterative Penetration Checking
Procedure. The second one (Time Step Reduction Procedure) is recommended only for a transient dynamic analysis
with automatic time stepping.
In Figure 8-36, node A is not in contact with the rigid body at the beginning of an iteration. After time step , node
A penetrates rigid surface if no constraint is imposed during the iteration which is not an acceptable solution.
Figure 8-35 Contact Coordinate System
u
nor mal
v n =
u
tangential
v t =
t
A
n
t y
x
z
Before Contact After Contact
B
A
B
575
CHAPTER 8
Contact
Figure 8-36 Contact Constraint Requirement
The Iterative Penetration Checking Procedure checks for every iteration in the Newton-Raphson process if the
displacement solution found would cause nodes to penetrate. If this is the case, the iterative displacement solution is
scaled such that at the start of the subsequent iteration, new contact will be established. Some mathematical details can
be found at the end of this chapter.
When in a transient dynamic analysis, the adaptive time stepping procedure AUTO STEP is used, Marc reduces the
time step of the current increment such that nodes, which would penetrate based on the original time step, will just
come into contact. In the current increment, no contact constraints are applied to these nodes. In the next increment,
the nodes are given the appropriate contact constraints and a new time step is determined based upon the convergence
criteria.
Analytical Contact
Utilizing the analytical representation of rigid surfaces influences the numerical procedure in several ways:
1. The rigid surface is represented by a nonuniform rational B-spline surface (NURBS) which provides a precise
mathematical form capable of representing the common analytical shapes circle, conic curve, free-form
curves, surfaces of revolution, and sculptured surfaces.
The surface can be expressed in simple mathematical form to model complex multiple surfaces with advantage
of continuity in first and second derivative. Those analytical forms are expressed in four-dimensional
homogeneous coordinate space by:
For curves, it simplifies to:
A
Start of Iteration, No Contact
A
End of Iteration, If No Constraint
P u v , ( )
B
i j ,
j 1 =
m 1 +
h
i j ,
N
i k ,
u ( )M
j l ,
v ( )
i 1 =
n 1 +
h
i j ,
N
i k ,
u ( )M
j l ,
v ( )
j 1 =
m 1 +
i 1 =
n 1 +
----------------------------------------------------------------------------------------- =

where the B are 4-D homogeneous defining polygon vertices, and are nonrational B-spline basis
functions, h
i,j
is homogeneous coordinates. For given parameters and in local system, the location (x, y,
z) in three-dimensional space, first derivative and second derivative (if required), are calculated. Given a point
with x, y and z Cartesian coordinates, there is no explicit mathematical form to find out the parameters u and
v in the local space, so they are obtained using an iterative procedure.
2. The location of the closest point on the contact surface corresponding to node A needs to be determined. This
process, determined analytically when using flat patches, requires an iterative approach for rigid bodies
modeled with NURBS.
Figure 8-37 Closest Point Projection Algorithm
3. Once the point P is known, the normal is calculated based upon the NURBS. This normal is used in a manner
consistent with the PWL procedure.
4. Because the normal can change from iteration to iteration, the imposition of the kinematic boundary conditions
is continuously changing.
In the PWL approach:
In the analytical approach:
where the subscripts represent the iteration number.
5. In the PWL approach, when a node slides from one segment to another, the corner condition logic is activated.
The advantages of analytical surfaces is that this logic is not necessary until you come to the end of the spline.
This reduces the amount of iterations required.
P u ( )
B
i
h
i
N
i k ,
u ( )
i 1 =
n 1 +
h
i
N
i k ,
u ( )
i 1 =
n 1 +
------------------------------------------ =
N
i k ,
M
j l ,
u v
A
P
u
n
( )
0
v n =
u
n
( )
i
0 =
u
n
( )
0
v n
0
=
u
n
( )
i
v n
1
v n
0
=
577
CHAPTER 8
Contact
6. The friction calculation is dependent upon the surface normal and tangent. When using the analytical approach,
the friction calculation includes the effect of changes in the direction of the normal vector from iteration to
iteration. This improves the accuracy and convergence behavior.
7. Because the normal is continuous, the calculation of nodal forces is more accurate. This allows for the
determination of nodal separation to occur beginning with the second iteration as opposed to when equilibrium
convergence has been achieved using the PWL approach.
Solution Strategy for Rigid Contact
Step 1: At the start of an increment, all boundary nodes are checked for contact with surfaces and flagged if so. If
the Iterative Penetration Checking Procedure has detected potential penetration, this is repeated at the start
of an iteration.
Step 2: Next, transformations and imposed displacements are determined for each touching node. With the
knowledge of the time increment and surface velocity, the configuration of the surface at the end of the
increment is found. Then, the distance from the node starting position to the current surface segment is
determined and is the probable normal displacement to be imposed. If there is a previous solution for
displacement increments, these are used to estimate the tangential displacement increment (Figure 8-38). If
this estimate still puts the node at the end of the increment in the same surface segment, then this one is used
to determine the transformation matrix between local and global coordinates, and the normal displacement
is accepted. Otherwise, the procedure is repeated for the adjacent segment. If a concave corner is hit, the
boundary node is fixed to such corner. If the node goes out of a sharp convex corner, it is immediately
released from the surface.
Figure 8-38 Node P Already In Contact (2-D)
SEG
B
P
u
u
test
u
n
SEG
E
SEG
B
- Die segment at beginning of increment
SEG
E
- Die segment at end of increment
u - Displacement increment of previous increment
u
n
- Imposed normal displacement
u
test
- Estimate of tangential displacement
- Angle used in transformation matrix

8
Contact
Step 3: One increment of the problem is solved iteratively. During each iteration, the Iterative Penetration Checking
Procedure checks if the current iterative displacement solution must be scaled to avoid penetration
(Figure 8-39). If needed, Step 1 is repeated to establish new contact constraints. Step 2 is repeated at each
iteration for possible changes in surface segment.
Figure 8-39 New Node in Contact P
Step 4: Once convergence is obtained, nodal forces are checked. Whenever a positive normal force (or stress) is
detected which is larger than the tolerance, a node is released from the surface segment at which such force
(or stress) corresponds and Step 3: is repeated. This tolerance can be modified using the SEPFOR or
SEPSTR user subroutine.
Step 5: Go to next increment (Step 3:).
Figure 8-40 Node P Already in Contact (3-D)
SEG
B
SEG
E
A
B
u
P
SEG
B
- Die segment at beginning of iteration
SEG
E
- Die segment at end of iteration
u - Iterative Displacement calculated
- Factor to scale iteration
PB
PA
--------
PAT
B
PAT
E
P
u
u
n
u
test
2
3
1
PAT
B
- Patch at beginning of increment
PAT
E
- Patch at end of increment
u - Previous displacement increment of Node P
u
n
- Imposed normal displacement increment (local direction 3)
u
test
- Estimated tangential displacement
579
CHAPTER 8
Contact
Deformable-Deformable Contact
When a node contacts a deformable body, a tying relation is formed between the contacting nodes and the nodes of the
contacted segment on the other body. This constraint relationship uses information regarding the normal vector to the
segment and the closest point projection of the contacting node on the contacted segment. For lower-order elements,
the normal vector and the projection are calculated based upon the piecewise linear representation of the element. This
has the consequence that the constraint relation may not be accurate because the normal vector is constant over the
complete segment, whereas the actual structure may be curved with a varying normal vector. When a node slides from
one segment to another, there may be a discontinuity of the normal vector, which leads to potential numerical
difficulties or an inaccurate solution. This is especially important for contact analyses where the deformations are
small.
When higher-order elements are used, the normal vector and projection are created based upon the true quadratic
element description. Provided that the midside nodes are defined to be on the actual curved geometry, this improves
the accuracy, but does not necessarily result in continuity of the normal vector at nodes shared by multiple segments.
Figure 8-41 2-D Deformable Contact; Piece wise Linear FE Description and Cubic Spline Representation
Both for lower- and higher-order elements, you can fit a smooth curve (2-D) or surface (3-D) through the finite element
segments. For 2-D, a cubic spline is used (see Figure 8-41). This spline is calculated based upon the tangent and
position vectors of the nodes on the segment. For lower-order elements, each segment is internally replaced by one
cubic spline; where for higher-order elements, each segment is replaced by two cubic splines. This gives a more
accurate representation of the actual physical geometry and a more accurate calculation of the normal vector. You must
identify where actual discontinuities exist. This is done by defining nodes where the normal vector is discontinuous
or by activating the option that the program will determine such nodes based on the angle spanned by the normal
vectors to adjacent segments. You can use the SPLINE option to define deformable bodies in 2-D or 3-D where the
smoothness is required.
For 3-D problems, the concept of a cubic spline can be expanded to generate a Coons surface to improve the geometric
representation of a segment (see Figure 8-42). When calculating the closest point projection, it is advantageous to
replace the Coons surface description by a cubic serendipity surface. For 2-D problems, you have to define nodes
where the actual normal vector has a discontinuity; for 3-D problems, you have to specify edges where the normal
vector is discontinuous. The SPLINE option allows you to enter a list of nodes (where each edge is defined by its start
and end node) or to activate the option that the program will determine such edges based on the angle spanned by the
normal vectors to adjacent segments. At edges with a discontinuous normal vector, the smooth surface description of
Actual Geometry Finite Element Approximation Cubic Spline Representation
: Nodes with a normal vector discontinuity

adjacent segments can be set up independently of each other, which usually results in C
0
-discontinuity. Alternatively,
the smooth surface description of adjacent segments can be forced to include information from the interface they have
in common, which results in C
0
-continuity. You can force C
0
-continuity in 3-D also on the SPLINE option.
If the SPLINE option is used for curved contact bodies defined by higher-order elements, the coordinates of the
midside nodes can be recalculated based on the cubic splines defined by the position and tangent vectors of the corner
nodes. This can be advantageous if the structure to be modeled is available only based on lower-order elements, so
that a change from lower- to higher-order elements yields midside nodes which are not on the curved boundary, but
on the straight edges between the corner nodes.
Figure 8-42 3-D Deformable Contact; Piece wise Linear FE Description and Coons Surface Representations
Solution Strategy for Deformable Contact
1. Upon initiation, the program determines the boundary nodes on each body. For 2-D, these nodes are ordered in
sequence in order to describe the boundary in a counterclockwise manner. Linear geometrical entities are then
created out of subsequent pairs of nodes, to define a surface profile as in a rigid surface. If higher-order
elements are used in full quadratic mode, quadratic entities are created from the three nodes. For 3-D using
lower-order elements, 4-node patches are created. For 3-D higher-order hexahedral elements, a quadratic
surface is created.
2. Once contact between a node and a deformable surface is detected, a tie is activated. The tying matrix is such
that the contacting node follows the shape of the surface (Figures 8-43 and 8-44); it can slide along it or be
stuck according to the general contact conditions.
Actual Geometry
Finite Element Approximation
C
0
-discontinuous Coons
C
0
-continuous Coons
Surface Representation
Surface Representation
581
CHAPTER 8
Contact
3. Contact occurs between all deformable bodies unless the CONTACT TABLE option is used. In deformable
contact, there is no master or slave body; each body is checked against every other body, except single-sided
contact or the CONTACT TABLE option is used.
4. During the iteration process of finding an incremental solution, a node can slide from one segment to another,
changing the retained nodes of its tie. A recalculation of the bandwidth is, therefore, made every iteration. From
increment (or subincrement) to increment, the number of nodes in contact can change. In such cases, an
optimization of the bandwidth is automatically available to cope with the possible drastic changes in bandwidth
that a new tie produces.
When the glue option is used between two deformable bodies, a simpler tying relationship is formed, such that
no relative motion occurs.
Figure 8-43 Tyings in Deformable Contact (2-D) using Lower-order Elements
Figure 8-44 Tying in Deformable Contact (3-D) using Lower-order Elements
seg
t+t
seg
t
U
B
U
A
U
C
A
B
C
Tied Node A
Retained Nodes A B C
T/R1
R2
R5
R3
R4
t
t+ t
Tied Node T
Retained Nodes T, R2, R3,R4,R5

Iterative Penetration Checking
Except for a transient dynamic analysis with automatic time stepping where the Time Step Reduction Procedure is
used, the commonly used method to prevent penetration in a contact analysis is the Iterative Penetration Checking
Procedure. Using this procedure, the iteration process is done simultaneously to satisfy both the contact constraints
and global equilibrium using the Newton-Raphson procedure. This procedure is accurate and stable. This procedure
is automatically activated when the AUTO STEP procedure is used in static analysis or when beam-beam contact is
present.
In a conventional iteration process, the finite element system calculates for each iteration:
where is the tangent stiffness matrix and are the residuals based on displacements from the
previous iteration.
Using the iterative penetration method is now based upon the contact status at this iteration, both from a
constraint and friction perspective. Furthermore, after the solution for is obtained, the contact procedure is used
to determine if new penetration will occur. If, at least, one node penetrates a contact surface, a scale factor is applied
to the change in displacements such that the penetrating nodes are moved back to the contact surface. This procedure
can be considered a type of line search. Given that is the fraction of such that no new penetration occurs, the
displacement increment then becomes
and the total displacement is
The strains, stresses, and residuals are based upon these quantities. Separation is based upon these values. When
global equilibrium is achieved, based upon the users criteria, the solution proceeds to the next increment. Because
the procedure can reduce the change in displacements, it may require more iterations to complete an increment. It is
important to ensure that the maximum allowable number of iterations to complete an increment is set to a sufficiently
large value.
Instabilities
In some analyses, because of instabilities such as buckling or a loss of contact, large displacement increments may
occur. In such cases, it is possible to specify a maximum iterative displacement component allowed . If the
magnitude of any iterative displacement component turns out to be larger than this maximum value, the iterative
displacement solution is scaled using a factor , such that:
for any node in the model. If is not specified, the value used defaults to the average of the maximum
dimension of the global rectangular boxes enveloping the contact segments. In case of fine meshes, this default value
can be small and scaling of the solution vector can be reduced by specifying a more physically meaningful length scale.
K
T
u
i
R
i 1
=
K
T
R
i 1
K
T
u
i
s u
i
U
i
U
i 1
su
i
+ =
U
n
U
n 1
U
i
+ =
u
al l owed
s
s u
i
u
al l owed
u
al l owed
583
CHAPTER 8
Contact
Finally, if the solution is still not able to converge and the cutback feature is activated, the time step is reduced in
magnitude and the increment is repeated.
Segment-to-Segment Contact
The contact algorithm discussed above is based on nodes being in contact with a segment (a curve, surface, element
edge, or element face), and can thus be called a node-to-segment algorithm. This contact algorithm has matured and
has been successfully applied to a large variety of contact problems. However, there are a few weaknesses of this
algorithm especially for deformable contact, which call for the development of an alternative contact algorithm:
When contact has been detected, the nonpenetration constraint is enforced on a nodal basis. Because of this
point-wise application of constraints, the node-to-segment algorithm does not generally maintain stress
continuity across the contact interface of deformable contact bodies. This can be easily illustrated using the
contact patch test, where a uniform stress distribution should be obtained in two deformable contact bodies
with dissimilar meshes (see [Ref. 7]).
Since the nonpenetration constraints are enforced using multi-point constraint equations, there is a potential
dependency of the solution on the selection of the master and the slave nodes. In other words, the solution
depends on which nodes are touching and which nodes correspond to a touched contact segment. Although
there are various options to optimize the multi-point constraint equations, it is not always possible to achieve
the best results everywhere in the model and to completely eliminate the dependency on the contact body
numbering.
If there would be contact detected at the top and the bottom face of a shell element, it is not possible to only use
the nodes of the shell elements to apply both multi-point constraint equations. This implies that there are
limited modeling options for so-called double sided shell contact. A similar problem occurs when the edge of a
shell element is in contact with the face of another element. If this happens, then the contact constraints are
based on either the shell top or the shell bottom face. Consequently, the footprint of the shell edge is not
directly related to the shell thickness.
In Marc 2010, a new contact algorithm has been introduced based on a segment-to-segment algorithm, where the
nonpenetration constraints are enforced using augmented Lagrangians. In this section, the basic aspects of this
algorithm will be discussed. It should be noted that currently the segment-to-segment algorithm has the following
limitations:
Mechanical contact only, no thermal or coupled contact is supported;
The AUTO INCREMENT load stepping algorithm is not supported.
Basic Theory
For a set of deformable contact bodies, the virtual work equation commonly used as the basis for the derivation of the
standard set of equations to be solved within a nonlinear finite element analysis can symbolically be written as:
(8-5)
where is the displacement field and the kinematically admissible variation of this field. Contact conditions can
be incorporated by adjusting this virtual work equation by two surface integrals:
G u u , ( ) 0 =
u u

(8-6)
Here is that part of boundary of the bodies being in contact and the subscript is used to indicate the normal
direction to the contact boundary, where the subscript indicates the tangential direction. The function is called a
gap function and expresses the distance between a point and its closest point projection on the contact boundary. If
, a point is outside the contact boundary, if , a point is on the contact boundary and if , a point
would be beyond the contact boundary, which is physically inadmissible. The Lagrange multiplier represents the
contact normal stress. Similarly, is the tangential gap vector and is the tangential stress vector.
The tangential or frictional behavior is assumed to be governed by Coulombs friction law, which, as already stated
before for node-to-segment contact, can be expressed as:
(8-7)
where is the friction coefficient; corresponds to sticking and corresponds to slipping.
Since the gap functions and depend on the geometry of the bodies (defining the normal and tangential vectors)
and the displacement field , the contact part of Equation (8-6) can be expressed as:
(8-8)
Since Equation (8-6) has to be valid for all kinematically admissible variations , the contact contribution to the
equilibrium equation is given by:
(8-9)
The overall solution procedure will be based on an augmented Lagrangian approach. It is assumed that the solution at
step is known and that the solution at step has to be determined in an iterative fashion, using a Newton-
Raphson process. Based on the iterative displacement solution obtained at iteration , the following trial solutions for
the Lagrange multipliers are introduced (note that the superscript to indicate the step number is omitted to simplify
the notation):
; (8-10)
in which and are fixed estimates of and , and are penalty factors for the contact behavior in the
normal and tangential directions (actually is a diagonal matrix with the penalty factor on the diagonal) and
and follow from the global iterative displacement solution. The default values of the penalty factors are derived
from the contact body properties (this will be discussed later on), but can also be user defined.
G u u , ( )
n
g
n

t
T
g
t
d
}
+ d
}
+ 0 =
n
t g
n
g
n
0 > g
n
0 = g
n
0 <
n
g
t

t

t

n
0 =
0 < 0 =
g
n
g
t
u
G
c
u u , ( )
n
g
n
u
-------- u d
}

t
T
g
t
u
------- u d
}
+ =
u
F
c
T
n
g
n
u
-------- d
}

t
T
g
t
u
------- d
}
+ =
n 1 n
i
n
n
t r i al
p
n
i 1
E
n
g
n
i
+ =
t
t r i al
t
t
i 1
E
t
g
t
i
+ =
p
n
t
t

n

t
E
n
E
t
E
t
E
t
g
n
g
t
585
CHAPTER 8
Contact
Looking first at contact in the normal direction, it can easily be seen that if a contact condition is exactly fulfilled, then
the gap function and . The idea of the augmentation procedure is to better approximate the
contact conditions by adjusting the contact normal stress. The evolution of the contact normal stress during the
Newton-Raphson process is illustrated in Figure 8-45. The first displacement solution is obtained using
(which initially can be zero, but in an arbitrary increment is given by the value at the end of the previous increment).
Given the displacement solution, the gap function is calculated and the contact normal stress update is given by
. This new contact normal stress is used in a subsequent iteration. If the augmentation is not
performed, then the method reduces to a penalty method with a constant penalty factor . As for the contact
conditions in the tangential direction, the augmentation procedure can be applied also if there is sticking contact. Then
the sticking conditions (no relative tangential displacements) will be better approximated by adjusting the tangential
stresses (see Figure 8-46).
Within the Newton-Raphson solution procedure, not only the contact contributions to the global force vector (as given
by Equation (8-9)), but also the contributions to the global stiffness matrix are needed:
(8-11)
Further evaluation of Equation (8-9) and Equation (8-11) depends on whether there is sticking or slipping. At the very
beginning of an analysis, an assumption has to be made (typically sticking), but after having determined the trial
solutions, it is possible to check if there will be sticking or slipping:
; : sticking; : slipping (8-12)
g
n
i
0 =
n
t r i al
p
n
i 1
=
p
n
p
n
0
=
g
n
1
p
n
1
p
n
0
E
n
g
n
1
+ =
E
n
g
n
p
n
E
n
( ) atan
g
n
1
g
n
2
p
n
1
p
n
2
exact value of
p
n
p
n
0
Figure 8-45 Updating Contact Normal Stress according to the Augmented Lagrange Procedure
K
c
u
------ F
c
=
t r i al
t
t r i al
n
t r i al
=
t r i al
0 <
t r i al
0

and previously made assumptions on the contact state can be adjusted accordingly. Now the states of sticking and
slipping will be discussed separately.
In case of sticking, the predicted solution for the Lagrange multipliers is:
; (8-13)
For the contact contributions to the global stiffness matrix, the second order derivatives of the gap functions with
respect to the displacements are neglected, so that the following terms are needed:
; (8-14)
In case of slipping, the tangential stresses need to be adjusted in order to agree with Coulombs friction law:
; (8-15)
and the contact contribution to the global stiffness matrix now involves also changes in contact in the normal direction:
;
g
t
t
t
E
t
( ) atan
Figure 8-46 Updating Contact Friction Stress according to the Augmented Lagrange Procedure (2D)
t
t
0
t
t
1
g
t
1
g
t
2
n
n
n
t r i al
=
t
n
t
t r i al
=
u
-------- E
n
g
n
u
-------- =

t
u
------- E
t
g
t
u
------- =
n
n
n
t r i al
=
t
n

t
t ri al
t
t ri al
-------------------
n
t r i al
=
u
-------- E
n
g
n
u
-------- =

t
u
-------

t
t r i al
t
t r i al
------------------ -

n
t r i al
u
----------------
n
t r i al
u
------

t
t r i al
t
t r i al
------------------ -
\ .
|
| |
+ = =
587
CHAPTER 8
Contact
(8-16)
The first term in the expression for the frictional behavior in Equation (8-16) results in a non-symmetric stiffness
matrix. If the assumption is made that the contact normal stress is constant during a Newton-Raphson iteration, then
this term cancels and a symmetric matrix is retained. Using a non-symmetric matrix will significantly increase the
memory requirements. It should be noted that a non-symmetric matrix may only have impact on the convergence
behavior; if convergence will be obtained, the final solution wont be affected.
In the paragraphs below, it will be shown how the Equations (8-9), (8-14) and will be numerically integrated in a
2- D or 3-D contact analysis.
Contact Body Definition
The definition of contact bodies for the segment-to-segment algorithm is identical to that of the node-to-segment
algorithm. This implies that one can define deformable contact bodies consisting of finite elements, and rigid bodies
consisting of curves (2-D) or surfaces (3-D). Rigid bodies can be load, velocity or displacement controlled; load
controlled rigid bodies have one (translation only) or two (translation and rotation) control nodes.
Based on the finite elements assigned to a contact body, the program will automatically set up the outer boundary of
the deformable bodies. This boundary is defined by element edges (2-D) or element faces (3-D). Figure 8-47 shows
the basic concept for 2-D continuum and 2-D shell/beam elements. For 2-D continuum elements, the contact segments
coincide with the element edges. For 2-D shell/beam elements, not only the top and the bottom are separately included
in the contact body description, but also the free ends of the finite element model are converted into contact segments.
The same concept is applied to 3-D continuum elements and 3-D shell elements, where for the shell elements the free
edges of the finite element model are converted into contact segments. The user can also decide to ignore the shell
thickness, in which case the top and the bottom segments have the same position, but an opposite orientation. The
contact segments of the free ends will always include the shell thickness.
The SPLINE option, as discussed before for node-to-segment contact, is available for segment-to-segment contact.
When this option is selected for a shell element or a 2-D beam element, then the transition from the top and bottom
contact segments to the segment at the free ends is automatically marked to have a normal vector discontinuity.
Contact Detection
As also shown in Figure 8-47, each contact segment will get a number of auxiliary points. These auxiliary points are
only used during the contact detection phase of the analysis and they are defined at fixed positions on the contact
segments. Since the auxiliary points are defined on a per segment basis, there are multiple points with possibly
different normal vectors at adjacent contact segments. Especially at sharp corners, this allows for multi-body contact
in such a way that each (part of a) segment can touch only one other segment.
t
t r i al
t
t r i al
------------------ - E
n
g
n
u
--------

n
t r i al
t
t ri al
------------------- E
t
1

t 1
t r i al
2
t
t r i al
2
---------------------

t 1
t r i al
t 2
t r i al
t
t r i al
2
------------------------------

t 1
t r i al
t 2
t r i al
t
t r i al
2
------------------------------ 1

t 2
t r i al
2
t
t r i al
2
---------------------
g
t
u
-------
\ .
|
|
|
|
|
|
|
|
|
| |
+ =

The actual contact detection phase consists of two passes. During the first pass, there is a check whether an auxiliary
point is sufficiently close to a contact segment and if the corresponding normal vectors are almost opposite. Hence
there is a distance check followed by a direction check (see also Figure 8-48). During the distance check, the distance
between an auxiliary point and its closest point projection on a contact segment is calculated and compared with a
reference value. This reference value is determined by the contact tolerance and the bias factor , where
. The default contact tolerance is calculated by the program as the smaller of 5% of the smallest element
side or 25% of the smallest (beam or shell) element thickness, based on all the elements defined in any contact body.
Alternatively, the user can define the contact tolerance. The default bias factor is zero. A non-zero bias factor, entered
by the user, implies that the outside zone becomes smaller and the inside zone larger and usually avoids unnecessary
iterations following from physically unrealistic contact detection. If the position of an auxiliary point is and that
of its closest point projection on a segment of contact body is , then the distance check is passed if
: nodal points
: auxiliary points
Finite Elements Contact Segments Auxiliary Versus Nodal Points
re 8-47 2-D Contact Bodies: Continuum Elements (Upper); Shell/beam Elements (Lower)
n
a1
n
a2
n
s
n
s
1 B + ( )
1 B ( )
: closest point projection
Distance Check And Direction Check Passed Distance Check Passed, Direction Check Failed
Figure 8-48 Contact Detection, Phase 1: Distance And Direction Check
B
0 B 1 <
x
a
s x
s
589
CHAPTER 8
Contact
when the auxiliary point lies outside the contact body, and if , when the
auxiliary point lies inside the contact body. If the distance check is passed, then the direction check is performed. This
direction check is passed if the angle between the normal vector at the auxiliary point and the normal vector at the
potentially contacted segment is larger than a threshold value , so . The default value of is ,
but it can be user defined using the PARAMETERS option.The angle between the normal vectors is defined according
to . Once both the distance and the direction check are passed, the corresponding segments are marked as a pair of
potentially contacting segments and for each of these pairs, the projection of one segment on the other is determined.
In 2-D, this projection defines a polyline, in 3-D it defines a polygon. In , a 2-D and a 3-D example is given. Each
polyline/polygon has a number of polygon/polygon points, which will serve as the basis for the numerical integration
of Equation (8-9). It should be noted that the polyline/polygon points are not the same as the auxiliary points
introduced before. The auxiliary points only serve to find potential contact pairs, the polyline/polygon points
characterize the actual contact area between such contact pairs.
x
a
x
s
1 B ( ) x
a
x
s
1 B + ( )
n
a
n
s
, ( ) 120
n
a
n
s
n
a
n
s
, ( )
Figure 8-49 Definition of the Angle Between Normal Vectors
: polyline points
Figure 8-50 Common Areas of Contact Segments: Polylines (Left) And Polygons (Right)

During the second pass of the contact detection phase, for each polyline/polygon point it is checked whether or not the
segments are locally in contact. This check is equivalent to the check which has been described above, so a distance
check followed by a direction check. At the end of this second pass, there is a set of polyline/polygon points being in
contact and representing the contact area.
Force Vector and Stiffness Matrix Contribution
The polyline/polygon points discussed above basically define a local connection between two contact segments.
Indicating such a point by a subscript , the gap function can be evaluated as the normal displacement difference
between a point and its closest point projection on the contacted segment. Similarly, by looking at the tangential
displacements, the gap function can be evaluated. The displacement of a point on a contact segment is a function
of the displacements and (in case of shell elements) rotations of the nodes corresponding to that segment. By collecting
all the nodal displacement and rotation degrees of freedom in a vector , the gap functions can be written as:
; (8-17)
Here, vector and matrix express the dependency of the normal and tangential displacements of a point on a
contact segment on the total set of degrees of freedom. In a similar way, and are used for the closest point
projection of this point on the contacted segment. It should be noted that, if applicable, , , and also
include the effect of the shell thickness and shell offset vectors. Care has been taken to define them in such a way that
they are able to represent rigid body modes of structures being in contact without introducing spurious energies.
Equation (8-17) can be used to numerically evaluate equation (8-9). First, the contribution to the force vector is given
by:
(8-18)
in which is the total number of polyline/polygon points, and are the estimated contact normal and
tangential stresses in polyline/polygon point (see Equations (8-13) and (8-15)) and is the area corresponding
to this point. Note that a possible dependency of , , and on has been neglected.
Next, for sticking contact, the contribution to the global stiffness matrix is:
(8-19)
i g
ni
g
t i
U
g
ni
G
ni
T
G'
ni
T
( )U = g
t i
G
t i
T
G'
t i
T
( )U =
G
ni
G
t i
G'
ni
G'
t i
G
ni
G'
ni
G
t i
G'
t i
g
n
u
--------
\ .
| |
T
n
d
}
g
t
u
-------
\ .
| |
T
t
d
}
+ G
ni
G'
ni
( )
ni
i
G
t i
G'
t i
( )
t i
i
i 1 =
N
+
i 1 =
N
=
N
ni

t i
i
i
G
ni
G'
ni
G
t i
G'
t i
U
g
n
u
--------
\ .
| |
T
E
n
g
n
u
--------
g
t
u
-------
\ .
| |
T
E
t
g
t
u
------- d
}
+ d
}
G
ni
G'
ni
( )E
ni
G
ni
T
G'
ni
T
( )
i
i 1 =
N
+ =
G
t i
G'
t i
( )E
t i
G
t i
T
G'
t i
T
( )
i
i 1 =
N
+
591
CHAPTER 8
Contact
Finally, for slipping contact the contribution is found to be:
(8-20)
The right choice of the penalty factors is important to get accurate results in an acceptable number of iterations. If the
penalty factor for normal contact, , is too small, then there will be a relatively large amount of penetration, which
has to be removed by the augmentation process outlined above. However, if the penalty factor is small, then the first
order correction process illustrated in Figure 8-45 will converge slowly. On the other hand, if the penalty factor is too
large, the overall system of equations may be ill-conditioned and a non-smooth contact stress field may be found.
The default value of the penalty factor depends on the body stiffness of the two contact bodies involved and a
characteristic length (note that the dimension of the penalty factor is force per cubic length). The body stiffness is
either defined by the average trace of the initial stress-strain law of the elements defining the two contact bodies or by
the average bulk modulus for (nearly) incompressible rubber materials, whichever of the two is the largest. For
continuum elements, the characteristic length is given by one half of the average length of all the edges being part of
the contact boundary. For shell elements, the characteristic length is given by half of the average thickness of all the
shell elements being part of a contact body. When there is contact between a solid and a shell element, then the
characteristic length is defined by the shell element. If polyline/polygon point defines contact between the
deformable bodies and , with body stuffiness and , and the characteristic length of the model is , then the
default value of is given by:
(8-21)
g
n
u
--------
\ .
| |
T
E
n
g
n
u
--------
g
n
u
--------
\ .
| |
T
E
n
t
T
t
----------
g
t
u
-------
g
t
u
-------
\ .
| |
T
t
---------- E
t
1

t 1
2
t
2
-------------

t 1
t 2
t
2
-------------------

t 1
t 2
t
2
-------------------

t 2
2
t
2
-------------
g
t
u
------- d
}
+ d
}
+ d
}
=
G
ni
G'
ni
( )E
ni
G
ni
T
G'
ni
T
( )
i
i 1 =
N
G
ni
G'
ni
( )E
ni
t i
T
t i
------------ G
t i
T
G'
t i
T
( )
i
i 1 =
N
+ + =
G
t i
G'
t i
( )
t
---------- E
t
1

t 1i
2
t i
2
---------------

t 1i
t 2i
t i
2
---------------------- -

t 1i
t 2i
t i
2
---------------------- - 1

t 2i
2
t i
2
---------------
G
t i
T
G'
t i
T
( )
i
i 1 =
N
+
E
n
E
ni
i
k l S
k
S
l
L
E
ni
E
ni
0.5 S
k
S
l
+ ( )
L
------------------------------- =

In case of contact with a rigid body, the equations shown before can be applied with minor modifications. First, the
displacements and of the closest point projection directly follow from the motion of the rigid body.
Second, the issue of an ill-conditioned system does not easily occur and the default penalty factor can be chosen to be
larger. Since there is no body stiffness associated with a rigid body, the default value of is related to the deformable
body only and given by:
(8-22)
For tangential contact, the default penalty factor is related to the penalty factor according to:
(8-23)
where for deformable contact and for rigid contact.
As mentioned before, the user can define the penalty factors. This can be done either globally via the CONTACT option
or per body combination via the CONTACT TABLE option. The penalty factors are by default printed in the output file.
Iterative Penetration Checking and Contact Stress Update
After each solution in the global Newton-Raphson process, there is a check if the iterative displacement field would
cause penetration of contact bodies. This check involves first the auxiliary points on the contact segments not yet being
marked as contact pairs and next the polyline/polygon points not yet being in contact. The maximum amount of
penetration at the auxiliary and polyline/polygon points is computed and if this penetration is larger than the inside
contact tolerance distance , then the entire iterative displacement solution is scaled back to avoid this
penetration.
Once the iterative displacement solution is known, the gap function can be updated with
and , where the superscript indicates the iteration number. The
update of the contact normal stress is given by:
(8-24)
Note that the last term in Equation (8-24) results from the augmentation procedure.
Since there are in general more polyline/polygon points on a contact segment than there are nodes on such a segment,
enforcing a zero gap function everywhere would in general result in an over-constrained system. To this end, a
reduction of the penetration field has to be applied. The following options are available:
No augmentation. This results in a pure penalty method and is recommended for most analyses, since it gives
reasonably accurate results in a relative small number of iterations.
Augmentation based on a constant penetration field. This is the recommended procedure for linear finite
elements.
Augmentation based on a (bi-)linear penetration field. This procedure should be used only for quadratic finite
elements.
G'
ni
U G'
t i
U
E
ni
E
ni
1000S
k
L
------------------ =
E
t i
E
ni
E
t i
E
ni
=
1
3
10 = 1
6
10 =
1 B + ( )
U
i
g
ni
i
G
i
T
G'
i
T
( )U
i
= g
ni
i
g
ni
i 1
g
ni
i
+ = i
p
ni
i
p
ni
i 1
E
ni
g
ni
i
E
ni
g
ni
i
+ + =
593
CHAPTER 8
Contact
It is also possible to let the program decide which of the augmentation methods should be applied (constant or linear).
Then the choice is made based on the combination of elements corresponding to the contacting segments. It is noted
that in case of contact with a rigid body, the augmentation procedure is always applied.
In case of augmentation based on a constant penetration field, the average gap function per contact segment is
determined and applied to each polyline/polygon point corresponding to that segment.
In case of augmentation based on a (bi-)linear penetration field, a least squares method is used to determine the values
of the gap function in the corner nodes of the segment. Based on the position of a polyline/polygon point on this
segment, a (bi-)linear interpolation between the corner values then gives the local value in the point.
Since an exact fulfillment of a zero gap function is difficult to achieve, a threshold value is needed to determine
whether or not an extra iteration due to the augmentation procedure is needed. This penetration distance beyond which
an augmentation will be applied can be calculated by the program or entered by the user. The default penetration
distance beyond which an augmentation will be applied is defined as:
(8-25)
where the characteristic length is the same as used in Equation (8-21). Similar to the penalty factor, the user can
define the augmentation distance either globally for the analysis on the CONTACT option or per body pair on the
CONTACT TABLE option.
If an augmentation procedure is selected and the maximum penetration is beyond , then an extra Newton-
Raphson iteration will take place, as long as the maximum number of iterations has not been exceeded. If the
maximum number of iterations has been reached, then the program will continue with the next increment if the usual
convergence criteria have been fulfilled; so the analysis will not stop because of the augmentation procedure. In case
of friction with augmentation, the same logic is applied to the tangential behavior.
Separation
Since in each polyline/polygon point the contact normal stress is calculated, separation will always be based on
stresses. By default, if the contact normal stress is in tension, then the corresponding polyline/polygon point will
separate, which implies that this point will no longer contribute to the global stiffness matrix and force vector.
In order to speed up the separation process on a structural level, there is an extra check performed on a contact segment
basis. If there are points on a segment which separate, then the total contact normal force due to points on this
segment being in tension (negative) is compared with the total contact normal force due to points on this segment
being in compression (positive). These forces are obtained by integrating the contact normal pressure over the
corresponding contact area. If , then all the points on this segment will separate. This extra check is
only active if the separation stress is default (zero). As soon as the user enters a non-zero separation stress, either
globally for the analysis on the CONTACT option or per body pair on the CONTACT TABLE option, this extra check
will be skipped.
aug
0.001L =
L
aug
F
nt
F
nc
F
nc
0.75 F
nt
<

Finite Sliding for Deformable Contact
In the discussion above, the position of the polyline/polygon points with respect to the element edges/faces is based
on the geometry of the elements at the first time of touching. In case of contact between deformable contact bodies,
this implies that the equations are valid only for simulations where the relative displacements between touching
deformable bodies are small. Since in many contact applications the relative displacements are finite, the default
option is to allow for large relative displacements. If large relative displacements are allowed, the relative
displacements of the polyline/polygon points are monitored and once a threshold value is exceeded, new points
will be created. The default threshold value equals five times the contact tolerance , but it can also be defined by the
user via the CONTACT option.
For a 2-D configuration, the update process for large relative displacements is illustrated in Figure 8-51. The initial
position of the polyline points is shown on the left-hand side. The blue and green areas represent the common areas
between the upper and the left-hand and right-hand lower element. A possible relative displacement is given in the
middle of Figure 8-51. As long as , the contact simulation is based on the initial position of the polyline
points. If , new polyline points are created, resulting in a situation as shown on the right-hand side.
Redefining polyline/polygon points will be done during an analysis as often as needed. After having created new
polyline/polygon points, important contact data (like the contact stresses) is mapped from the old on the new
polyline/polygon points and used as a starting point to continue the analysis.
The CONTACT option allows the user to switch on the flag for small relative displacements, in which case there is no
update of the position of the polyline/polygon points.
Postprocessing
The following contact related results are available for postprocessing:
Total forces and moments on contact bodies.
Motion of rigid bodies.
d
max
d
d d
max
<
d d
max
Figure 8-51 Large Relative Displacements: Initial Position (left), Sliding (middle) and New Position (Right)
d
595
CHAPTER 8
Contact
Contact status. This is a nodal quantity and set to one as soon as there is contact in a polyline/polygon point on
a contact segment to which this node belongs. So for segment-to-segment contact, the contact status is shown
on all bodies that are in contact. This simplifies the identification of segments that are in contact and is
independent of the body ordering.
Contact normal force. This is a nodal vector directly following from the integration of the contact normal stress
in the polyline/polygon points.
Contact friction force. This is a nodal vector directly following from the integration of the contact friction
stress in the polyline/polygon points.
Contact normal stress. This is a nodal vector representing the normal stress in the polyline/polygon points. This
nodal representation is based on a constant field approximation per contact segment of the polyline/polygon
values.
Contact friction stress. This is a nodal vector representing the friction stress in the polyline/polygon points.
This nodal representation is based on a constant field approximation per contact segment of the
polyline/polygon values.
Since there are usually more polyline/polygon points than there are nodal points, in many cases the contact stress
vectors can only approximately represent the actual stress field calculated by the contact algorithm.
References
1. Oden, J. T. and Pires, E. B. Nonlocal and Nonlinear Friction Laws and Variational Principles for Contact
Problems in Elasticity, J. of Applied Mechanics, V. 50, 1983.
2. Ju, J. W. and Taylor, R. L. A perturbed Lagrangian formulation for the finite element solution of nonlinear
frictional contact problems, J. De Mechanique Theorique et Appliquee, Special issue, Supplement, 7, 1988.
3. Simo, J. C. and Laursen, T. A. An Augmented Lagranian treatment of contact problems involving friction,
Computers and Structures, 42, 1992.
4. Peric, D. J. and Owen, D. R. J. Computational Model for 3-D contact problems with friction based on the
Penalty Method, Int. J. of Meth. Engg., V. 35, 1992.
5. Taylor, R. L., Carpenter, N. J., and Katona, M. G. Lagrange constraints for transient finite element surface
contact, Int. J. Num. Meth. Engg., 32, 1991.
6. Wertheimer, T. B., Numerical Simulation Metal Sheet Forming Processes, VDI BERICHTE, Zurich,
Switzerland, 1991.
7. Feng, Q. and Prinja, N.K., NAFEMS Benchmark Tests for Finite Element Modelling of Contact, Gapping and
Sliding, NAFEMS, 2001.
Chapter 9 Boundary Conditions
9
Boundary Conditions
Loading 598
Kinematic Constraints 625
Mesh Independent Connection Methods 660

597
CHAPTER 9
Boundary Conditions
Marc is based on the stiffness method and deals primarily with force-displacement relations. In a linear elastic system,
force and displacement are related through the constant stiffness of the system; the governing equation of such a
system can be expressed as
(9-1)
where is the stiffness matrix and and are nodal displacement and nodal force vectors, respectively. Equation
(9-1) can be solved either for unknown displacements subjected to prescribed forces or for unknown forces (reactions)
subjected to prescribed displacements. In general, the system is subjected to mixed (prescribed displacement and
force) boundary conditions, and Marc computes both the unknown displacements and reactions. Obviously, at any
nodal point, the nodal forces and nodal displacements cannot be simultaneously prescribed as boundary conditions for
the same degree of freedom.
The prescribed force boundary conditions are often referred to as loads and the prescribed displacement boundary
conditions as boundary conditions.
In a nonlinear stress analysis problem, Marc carries out the analysis incrementally and expresses the governing
equation in terms of the incremental displacement vector and the incremental force vector .
(9-2)
Consequently, you must also define both the loads and the prescribed nodal displacements incrementally.
In addition to the prescribed displacement boundary conditions, constraint relations can exist among the nodal
displacements. For example, the first degree of freedom of node i is equal to that of node j at all times. The expression
of this constraint relation is
(9-3)
Generally, a homogenous linear constraint equation can be expressed as
(9-4)
where represents the degrees of freedom to be constrained, are other retained degrees of freedom in the
structure, and are constants.
You can enter constraints through either the TYING, SERVO LINK, RBE2, or RBE3 options.
Note:
You must prescribe at least a minimum number of boundary conditions to insure that rigid body motion
does not occur.
Note:
Boundary conditions can be prescribed in either the global or a local coordinate system. A nodal
transformation between the global and the local coordinate systems must be carried out if the boundary
condition is prescribed in a local system.
Ku F =
K u F
u F
Ku f =
u
i
u
j
=
u
t
a
1
u
1
a
2
u
2
a
n
u
n
+ + + =
u u
1
u
n
, ,
a
1
a
n
, ,

You can use linear/nonlinear springs and foundations to provide special support to the structure and the gap and friction
element (or CONTACT option) to simulate the contact problem.
Loading
Different types of analyses require different kinds of loading. For example, the loads in stress analysis are forces; those
in heat transfer analysis are heat fluxes. Force is a vector quantity defined by magnitude and direction; heat flux is a
scalar quantity defined by magnitude only. Loading can be time invariant (constant value) or time dependent.
Loading Types
You can categorize a particular type of load as either a point (concentrated) load or surface/volumetric (distributed)
load, depending on application conditions. The spatial distribution of the load can be uniform or nonuniform. Special
loading types also exist in various analyses. For example, centrifugal loading exists in stress analysis, and convection
and radiation exist in heat transfer analysis. You can add point loads directly to the nodal force vector, but equivalent
nodal forces first must be calculated by Marc from distributed loads and then added to the nodal force vector. These
distinguishing features are described below.
A point (or nodal) load of either a vector (force, moment) or a scalar (heat flux) quantity is a concentrated load that is
applied directly to a nodal point (see Figure 9-1). Mechanical point loads can be defined as fixed direction forces or
as follower forces. In a global or local coordinate system, a force vector must be defined in terms of its vector
components (see Figure 9-2). If the force vector is defined in a local coordinate system, then a global-to-local
coordinate transformation matrix must be defined for the nodal point (see Figure 9-3 and Figure 9-4). For
axisymmetric elements, the magnitude of the point load must correspond to the ring load integrated around
the circumferences.
Figure 9-1 Schematic of a Point Load
Figure 9-2 Force Components
y
x
Point (a)
Heat Flux Q (Scalar)
Q
y
x
Point (b)
Force F (Vector)
F
y
x
F
F
x
F
y
y
x
F
F
x
F
y
y
x
599
CHAPTER 9
Boundary Conditions
Figure 9-3 Special Selection of Local (x, y) Coordinate System Force Components:
Surface/volumetric loads, such as pressure, distributed heat flux, and body force, are distributed loads that are applied
to the surface (volume) of various elements (see Figure 9-4). A surface/volumetric load is characterized by the
distribution (uniform/nonuniform) and the magnitude of the load, as well as the surface to which the load is applied
(surface/volume identification). The total load applied to the surface (volume) is, therefore, dependent on the area
(interior) of the surface (volume).
Equivalent nodal forces first must be calculated from surface/volumetric loads and then added to the nodal force
vector. Marc carries out this computation through numerical integration. (See Marc Volume B: Element Library for
the numbers and locations of these integration points for different elements). The calculated equivalent nodal forces
for lower-order elements are the same as those obtained by equally lumping the uniformly distributed loads onto the
nodes. However, for high-order elements, the lumping is no longer simple (see Figure 9-5). As a result, the
surface/volumetric loads should not be lumped arbitrarily.
Figure 9-4 Schematic of Surface/Volumetric Load
y
x
F
y
= 0
y x
F = F
x
F
y''
0 =
y
x
1
P
2
4 3
2
P
x
Q
(a) Distributed Mechanical Load (b) Distributed Heat Flux
P
y
P
1
P
y
x
1
P
2
4 3
2
P
1
q
Surface 2-3:
Uniform Normal Pressure p Uniform Heat flux q
Surface 3-4:
Nonuniform Normal Pressure p
1
- p
2
Nonuniform Heat flux q
1
= q
2
Whole Volume:
Volumetric Loads P
x
P
y
Volumetric Heat Flux Q

Figure 9-5 Allocation of a Uniform Body Force to Nodes for a Rectangular Element Family
Using the table driven input format, boundary conditions may be applied to finite element entities, nodes, elements,
edges or faces, or geometric entities such as points, curves, and surfaces. The latter is advantageous because it
improves compatibility with CAD and because it facilitates adaptive meshing. The application of boundary conditions
on geometric entities is only effective if finite elements have been associated with these geometric entities using the
ATTACH NODE, ATTACH EDGE, and ATTACH FACE model definition options. As shown in Figure 9-6, Mentat
shows a 2-D model consisting of four curves around the perimeter and a curve where the circle is. The mesh was
created using the automatic mesh generator which automatically attaches the edges to the curves. Those nodes which
are attached to points are shown using the circle symbol as opposed to the square symbol. Now, if a boundary
condition, such as an internal pressure, is applied on a curve during the analysis, it will automatically be applied to
those edges attached to the curve as shown in Figures 9-8 and 9-9. In a similar manner, if a displacement is applied to
a point, then it will be applied to the node. For 3-D, the analogous situation exists; namely, a distributed load applied
to a surface will be applied to those elements attached to this face as shown in Figure 9-10. And nodal boundary
conditions will be attached to nodes attached to face.
1/4
1/4 1/4
1/4
1/2 1/2 1/4
1/4
1/4
1/4
-1/12
1/3
1/3
1/3
-1/12
-1/12
-1/12
1/3
1/6 1/6 2/3
-1/12 -1/12
-1/12 -1/12 1/3
1/3
1/3 1/3
601
CHAPTER 9
Boundary Conditions
Figure 9-6 Geometric Representation with Five Curves
Figure 9-7 Original Mesh, Red Edges Indicate that they are Attached to Curve

Figure 9-8 Boundary Conditions Applied to Curves
Figure 9-9 Boundary Conditions Applied to Attached Element Edges
603
CHAPTER 9
Boundary Conditions
Figure 9-10 Distributed Load Applied to Surfaces. In Mentat, Faces Attached to Surfaces are Indicated by a Darker
Color (usually Blue and Midnight Blue)
Now, if global or local adaptive meshing is used, then, as shown in Figures 9-11 and 9-12, the load automatically is
applied to the new element edges which are attached to the curve. If a point load of prescribed displacement is applied
to a geometric point, the mesh generator will make sure that a point/node is in this exact location after remeshing
occurs. This allows all boundary conditions to be used in conjunction with global or local adaptive meshing in 2-D. In
3-D, one can apply boundary conditions to geometric entities as well, but they are not updated with global adaptive
meshing in this release.
For the case of shells, this is complicated because, for certain types of load, there is a difference between application
to the top and bottom surfaces. In mechanical analysis, a positive pressure on the top surface is equivalent to a negative
pressure applied to the bottom surface. This is because the degrees of freedom of a mechanical shell are considered to
be at the neutral surface. For thermal analysis, distributed fluxes, convective boundary conditions and radiation, there
is a difference between application of thermal boundary conditions on the top or bottom surfaces. This is particularly
relevant for radiation, as the choice of surfaces will strongly influence the viewfactor calculation. Figure 9-13, shows
a surface, with the top (yellow) and bottom (blue) surfaces identified. To select either the top or bottom surface, use
the FILTER command. A boundary condition applied to the top and bottom surface is shown in Figures 9-14 and
9-15, respectively. If the shell mesh is automatically generated from the surface, the top face of the element is
associated with the top surface of the shell. It is also possible to display the shell elements to indicate the thickness as
shown in Figure 9-16, in which case, one can easily select the top or bottom faces for which the boundary condition
is to be applied.

Figure 9-11 Distributed Load on Top Surface after Global Adaptive Remeshing
Figure 9-12 Distributed Load to Top Surface after Local Adaptive Remeshing
Figure 9-13 Identification of Top and Bottom Surface of a Shell
605
CHAPTER 9
Boundary Conditions
Figure 9-14 Boundary Condition on Top Surface
Figure 9-15 Boundary Condition on Bottom Surface
Figure 9-16 Shell Elements shown in Expanded Mode
Figure 9-17 is an example of what can go wrong if either geometric entities or elements are not consistently defined.
Two curves are used to represent an axisymmetric shell, but as shown by the orientation vector, they are not continuous
in direction.

Figure 9-17 Oriented Curves
Figure 9-18 shows the finite element mesh obtained from the convert option, which does not indicate any potential
problem. If one displays the shell in expanded mode and identifies backfaces, then one will see an inconsistency
between the top (blue) and the bottom (brown) surfaces as shown in Figure 9-19.
Figure 9-18 Finite Element Mesh
Figure 9-20 shows the boundary condition applied to the top curve, which is not what is desired. To get a consistent
applied boundary condition, one needs to flip both the curves and elements of one of the sections.
Figure 9-19 Shell Elements Expanded showing Top (Blue) and Bottom (Brown) Surfaces
607
CHAPTER 9
Boundary Conditions
Figure 9-20 Incorrect Boundary Condition on Top Surface
Face ID for Distributed Loads, Fluxes, Charge, Current, Source, Films,
and Foundations
When using the table driven input procedure, there are three ways to identify the type of load and which edge or face
it is applied to.
In the first method. an IBODY, as defined for each element type defined in Marc Volume B: Element Library, is
associated with each distributed load list. This IBODY effectively defines the type (normal pressure, shear, gravity,
pressure, centrifugal, Coriolis) and the edge or face to which the load is applied.
For the second method, one needs to enter the type of load and the location separately, but the location of the load is
consistent for all element types of the same class. This is called the face ID and is shown in the following figures. It
should be noted that when input files are created using Mentat, the convention is to use the same ID as used in the
Marc convention minus 1.
1-D 2-Node Elements
1-D 3-Node Elements
y
x
1
2
FACE ID
1
NODES
1 2
1
2
FACE ID
1
NODES
1 2 3
3

2-D 4-Node Quadrilateral Elements
2-D 8-Node Quadrilateral Elements
2-D 3-Node Triangle
2-D 6-Node Triangle
1
FACE ID
1
2
3
4
NODES
1 2
2 3
3 4
4 1
2
3 4
Load shown on FACE ID 1
1
FACE ID
1
2
3
4
NODES
1 5 2
2 6 3
3 7 4
4 8 1
2
3 4
6
5
7
8
1
FACE ID
1
2
3
NODES
1 2
2 3
3 1
2
3
1
FACE ID
1
2
3
NODES
1 4 2
2 5 3
3 6 1
2
3
4
6 5
609
CHAPTER 9
Boundary Conditions
3-D 3-Node Shell
3-D 4-Node Shell/Membrane
3-D 6-Node Shell
3-D 8-Node Shell
1
FACE ID
1
2
NODES
1 2 3
1 3 2
2
3
z
x
y
(top)
(bottom)
1
FACE ID
1
2
NODES
1 2 3 4
1 4 3 2
2
3
4
P
(top)
(bottom)
1
FACE ID
1
2
NODES
1 2 3 4 5 6
1 3 2 6 5 4
2
3
4
6 5
P
(top)
(bottom)
1
FACE ID
1
2
NODES
1 2 3 4 5 6 7 8
1 4 3 2 8 7 6 5
2
7
5
8 6
4
(top)
(bottom)
3

3-D 4-Node Tetrahedral
3-D 6-Node Pentahedral
3-D 15-Node Pentahedral
1
FACE ID
1
2
3
4
NODES
1 2 4
2 3 4
3 1 4
1 2 3
2
3
4
1
2
6
4
5
3
FACE ID
1
2
3
4
5
NODES
1 2 5 4
2 3 6 5
3 1 4 6
1 2 3
4 6 5
8
12
2 7
11
1
9
3
15
14
5
10
13
4
6
FACE ID
1
2
3
4
5
NODES
1 2 5 4 7 14 10 13
2 3 6 5 8 15 11 14
3 1 4 6 9 13 12 15
3 2 1 8 7 9
4 5 6 10 11 12
611
CHAPTER 9
Boundary Conditions
3-D 8-Node Brick
3-D 10-Node Tetrahedral
3-D 20-Node Brick
For the third method, the element edges or faces are attached to a curve or surface, respectively. In this case, one only
needs to give the type of load as the location is inferred from the ATTACH EDGE and ATTACH FACE data.
Mechanical Loads
Marc allows you to enter mechanical loads in various forms for stress analysis. These loads can be concentrated forces
and moments, uniformly and nonuniformly distributed pressures, body forces, gravity or centrifugal loads. Table 9-1
lists input options for mechanical loads.
1
FACE ID
1
2
3
4
5
6
NODES
1 2 6 5
2 3 7 6
3 4 8 7
4 1 5 8
1 2 3 4
6 5 8 7
2
6
4
8
5
3
7
FACE ID
1
2
3
4
NODES
1 2 4 5 9 8
2 3 4 6 10 9
3 1 4 7 8 10
1 2 3 5 6 7
1
2
3
4
6
5
8
10
9
7
1
FACE ID
1
2
3
4
5
6
NODES
1 2 6 5 9 18 13 17
2 3 7 6 10 19 14 18
3 4 8 7 11 20 15 19
4 1 5 8 12 17 16 20
1 2 3 4 9 10 11 12
6 5 8 7 13 16 15 14
2
6
4
8
5
3
7
10
19
11
14
15
20
16
13
17
12
9
18

Application of centrifugal and Coriolis loadings require the ROTATION A model definition option, which defines the
data corresponding to the axis of rotation. If multiple parts are rotating, then different rotation axes may be defined.
The actual load can be invoked by specifying an IBODY load types 100, 103,104, or 105 for centrifugal and Coriolis
loadings, respectively. For load types 100 or 103, the square of rotation speed, , is entered in radians per time, for
the magnitude of the distributed load. For load types 104 or 105, the angular velocity is entered as cycles per time. The
mass density must also be defined in the ISOTROPIC, ORTHOTROPIC, or other material options.
Application of gravity load (load per unit mass) is achieved by using IBODY load type 102. The mass density must be
defined in the ISOTROPIC, ORTHOTROPIC, or other material options. The acceleration can be given independently
in the x, y, and z direction through the DIST LOADS option.
Volumetric loads (load per unit volume) may be entered by using IBODY type 106 or 107.
The FOLLOW FOR parameter is used when the direction of Distributed and/or Point Loads are not fixed and need to
follow the geometry. FOLLOW FOR automatically invokes the LOAD COR parameter, so stresses should be stored at
all integration points. Separate flags under this parameter are used to control follower forces for distributed loads and
point loads respectively. When the FOLLOW FOR parameter is flagged for distributed loads, equivalent nodal loads
due to pressures are calculated based on the current geometry. Any change of surface area or orientation results in a
change of load. You can also specify that the follower force stiffness matrix due to distributed loads can be included.
The FOLLOW FOR parameter for distributed loads is typically used when a shell structure, which can undergo large
deformation and rotations, is subjected to a pressure load.
A separate flag on the FOLLOW FOR parameter allows follower forces to be specified for point loads. When this
parameter option is flagged, the user still has the choice of specifying individual point loads as fixed direction load
vectors or as follower forces. Fixed direction load vectors are specified by the usual vector components. When defined
as a follower force, there are two techniques available for specifying the load:
The first technique is similar to the option available in Nastran. The scalar magnitudes of the load and/or
Table 9-1 Input Options for Mechanical Loads
Load Description
Input Options
Point Load: Concentrated Force/Moment POINT LOAD POINT LOAD FORCDT
Surface Load Pressure Shearing Forces,
and Distributed Moment
(Uniform/Nonuniform Distribution)
DIST LOADS DIST LOADS FORCEM
Volumetric Load Body Forces and
Acceleration Forces
DIST LOADS DIST LOADS FORCEM
Centrifugal Loading DIST LOADS
ROTATION A
DIST LOADS
Coriolis Loading DIST LOADS
ROTATION A
DIST LOADS
Fluid Loading DIST LOADS
FLUID DRAG
2
613
CHAPTER 9
Boundary Conditions
moment are specified instead of the vector components. The direction of the force is then specified through 2
or 4 independent nodes in a manner similar to the FORCE1/MOMENT1 or FORCE2/MOMENT2 options in
Nastran. In the former, two nodes A and B are to be specified by the user and the direction of the load follows
the vector directed from node A to node B, as shown in Figure 9-21(a). In the latter, four nodes A, B and C, D
are specified by the user. The direction of the load follows the cross product of the vector directed from node A
to node B and the vector directed from node C to node D, as shown in Figure 9-21(b).
Figure 9-21 Nastran Style Follower Force Definition for Point Load
The second technique offers an automated way of defining the follower force. The load/moment is specified
through vector components as usual. The initial orientation of the load with respect to an optimal mesh location
is noted. Then, as the structure deforms, the load direction is updated such that the relative orientation of the
load with respect to the mesh is maintained, as shown in Figure 9-22. Optimal nodes in the mesh are
automatically chosen for calculating and maintaining the relative orientation. One or two optimal nodes can be
used. The optimal nodes are chosen based on the following criteria: The nodes should belong to the same
element as the loaded node. If the initial vector from the loaded node to any node matches the initial load
direction, then that node is chosen as the optimal node. Else, a node whose vector from the loaded node has a
component in the load direction is chosen. If this also does not provide a suitable node, the closest node to the
loaded node is chosen. For 3-D shells, membranes, and 3-D solids, two optimal nodes are chosen if the cross
product of the vectors of the loaded node to these two nodes matches the load direction. Else, two nodes whose
vectors from the loaded node have a cross product with a component in the load direction are chosen. If none
of the above rules provide suitable nodes, the two closest nodes to the loaded node are chosen.

Figure 9-22 Geometry Based Automated Follower Force Definition for Point Load
P (Load Magnitude)
A
Load
B
A,C
B
D
(a) FORCE1 Style (b) FORCE2 Style
P (Load
Load
Direction
Magnitude)
Direction
Initial Load
Updated Load
Optimal Direction
Direction
Node

The second technique is quite convenient and can be generally applied to 2-D and 3-D beams, shells and solids. In
situations where the loaded node does not belong to an element (for example, load controlled rigid dies, etc.), the first
technique should be used. It should be noted that a follower force stiffness matrix is not currently available for the
point loads. Also, the follower force capability is not supported for a point load specified through the FORCDT user
subroutine.
A number of history definition options are available for input of multiple load increments. For example, the AUTO
LOAD option generates a specified number of increments, all having the same load increment, and is useful for
nonlinear analysis with proportional loads; the PROPORTIONAL INCREMENT option allows the previous load
increment to be scaled up or down for use in the current load increment. The AUTO INCREMENT option allows
automatic load stepping in a quasi-static analysis and is useful for both geometrically and materially
nonlinear problems.
Fluid Drag and Wave Loads
Marc provides a fluid drag and wave load capabilities that can be applied on beam type structures that are partially or
fully submerged in fluid (see Figure 9-23). Morisons equation is used to evaluate the fluid drag loads that are
associated with steady currents. Only distributed drag and buoyancy effects are considered. Marc employs Airy wave
theory to evaluate wave velocities that can be invoked for dynamic analysis option.
Fluid drag and wave loads are invoked using the DIST LOADS model definition with an IBODY load type of 11. These
loads also require the FLUID DRAG model definition to input the relevant information regarding the fluid elevation
and its flow. Table 9-2 lists input options for fluid drag and wave loads.
Table 9-2 Input Options for Fluid Drag and Wave Loads
Load Description
Input Options
Fluid Drag Load DIST LOADS
FLUID DRAG
DIST LOADS
Wave Load DIST LOADS
FLUID DRAG
DIST LOADS
615
CHAPTER 9
Boundary Conditions
Figure 9-23 Fluid Drag and Wave Loads
Cavity Pressure Loading
Marc allows the modeling of structures enclosing cavities by updating the cavity internal pressure as the cavity volume
change. For ideal gas-filled cavities, the equation of state relating the cavity pressure and volume can be written as:
(9-5)
where is the cavity total pressure, is the cavity volume, is the number of molecules of the gas, is the
Universal Gas Constant ( ), and is the absolute gas
temperature. The gas mass, , is related to the number of molecules of the gas by:
(9-6)
where is the molar mass of the gas. Substituting by from Equation (9-6) into Equation (9-5) gives:
(9-7)
where is the Specific Gas Constant. With the gas density, , defined as:
(9-8)
Steady Velocity
Wave Load
Pipe
Outside
Inside
Sea Bed
Flow Velocity
Velocity
Buoyancy
Fluid Drag
Fluid
Force
Fluid
Force
Distribution
Gradient
pV nR
o
T =
p V n R
o
R
o
8.31447 J/(molK) 1545.35 ft.lbf/(molR) = = T
M
M n =
n
pV MRT =
R R
0
=
M V =

the equation of state of an ideal gas can be finally written as:
(9-9)
The Specific Gas Constant is calculated from:
(9-10)
where is the gas reference pressure, is the gas reference temperature, and is the gas reference density. The
user must ensure that the values entered for the gas reference pressure, temperature and density are consistent. The
cavity total pressure is given by:
(9-11)
where is the ambient pressure and is the cavity gage pressure. Only the cavity gage pressure is applied to the
structure forming the cavity.
The following loading scenarios are available for cavities:
Closed Cavity
If the cavity is closed, the mass of the gas is preserved
(9-12)
where is the cavity mass from the previous increment. The gas is assumed to undergo a general polytropic process
represented by:
(9-13)
where is the polytropic exponent. The gas pressure can thus be updated using
(9-14)
where and are the cavity pressure and volume from the previous increment, respectively. Using Equation (9-7)
and Equation (9-14), the gas temperature can be updated using
(9-15)
where is the cavity temperature from the previous increment. The gas density is calculated from:
(9-16)
p RT =
R p
r

r
T
r
=
p
r
T
r

r
p p
a
p
g
+ =
p
a
p
g
M M
o
=
M
o
pV
constant =
p
p
o
-----
V
o
V
------ -
\ .
| |

=
p
o
V
o
T
T
o
------
V
o
V
-------
\ .
| |
1
=
T
o
M V =
617
CHAPTER 9
Boundary Conditions
Closed cavity processes can be set to occur:
at constant pressure, isobaric, using .
at constant temperature, isothermal, using .
with no heat transfer to the surroundings, adiabatic, using , where is the adiabatic exponent. For ideal
gases, is a constant that depends only on the number of atoms in the gas molecule (monoatomic gases:
, diatomic gases: , triatomic gases: ).
Applied Pressure
In this case the new cavity pressure is updated using
(9-17)
The gas temperature is assumed to be constant
(9-18)
The gas density is updated using
(9-19)
and the gas mass is recalculated as
(9-20)
Applied Mass
In this case, the new cavity mass is updated using
(9-21)
If , gas is pumped in or out of the cavity. The gas density is first updated as
(9-22)
the gas temperature is assumed to be constant
(9-23)
and the cavity pressure is calculated as
(9-24)
If , the cavity is assumed to be closed and Equations 9-12 to 9-16 apply with .
User Defined Loading
The UCAV user routine allows the user to define and control the cavity pressure for loading scenarios other than the
ones specified above.
0 =
1 =
k = k
k
k 1.67 = k 1.4 = k 1.33 =
p p
o
p + =
T T
o
=
p RT =
M V =
M M
o
M + =
M 0
M V =
T T
o
=
p RT =
M 0 = 1 =

Cavity Modeling
The CAVITY parameter is used to enter the number of cavities in the model (maximum 1000), an upper bound to the
number of segments in each cavity and an upper bound to the number of nodes per segment of cavity boundary.
The CAVITY model definition option is used to enter the ambient pressure, the polytropic process exponent, and the
reference pressure, temperature and density for each cavity.
The applied pressure or mass is entered through the DIST LOADS model definition and history definition options. The
distributed load type, entered on the DIST LOADS option of the elements forming the cavity, is modified for cavity
loading according to the following relation
(9-25)
where
In the first increment of the analysis, increment zero, the cavity temperature is calculated by averaging the
temperatures of the elements forming the cavity. If there are no nodal temperatures defined, the cavity temperature is
taken equal to the cavity reference temperature. For AXITO3D (or PRE STATE) analysis, the cavity pressure, mass,
and temperature for increment zero are read from the post file of the corresponding axisymmetric problem. If the
cavity is closed in increment zero, the cavity is assumed to be initially at ambient conditions. In this case, the initial
gage pressure is equal to zero and the initial mass is based on the ambient pressure and initial volume. Moreover, if
the ambient pressure is equal to zero, the cavity is assumed to be initially unloaded. Tables 9-3 and 9-4 summarize the
cavity functioning for the different cavity load types and incremental load values during the initial and subsequent load
increments, respectively.
ibody_cavity is the cavity-modified value for the distributed load type.
icavity is the cavity ID.
icavity_type is the cavity load type:
0: cavity is closed.
1: cavity is loaded with an applied pressure.
2: cavity is loaded with an applied mass.
9: cavity load is defined by the UCAV user subroutine
ibody is the original value for the distributed load type (see library element description in Marc Volume B:
Element Library.)
Table 9-3 Cavity Functioning for Increment Zero
Cavity Load Type Incremental Load Value Function
0 Ignored Cavity is initially at ambient conditions
1 Cavity is initially empty
1 Cavity is initially at the applied pressure
2 Cavity is initially empty
2 Cavity initially contains the applied mass
9 Passed to UCAV Call UCAV
i body_cavit y i cavit y
*
10000 icavi ty_t ype
*
1000 ibody + + =
p 0.0 =
p 0.0
M 0.0 =
M 0.0
619
CHAPTER 9
Boundary Conditions
In general, standard structural elements are used to define the boundaries of cavities and no extra elements are
required. However, to model the boundaries of cavities in regions where standard finite elements are not present (for
example, along rigid boundaries) cavity surface elements (elements 171-174) can be used. These elements can also
be glued to moving rigid surfaces. They are for volume calculation purposes only and do not contribute to the stiffness
equations of the model.
For 2-D problems, the volume of the cavity is calculated as the sum of the areas of all the triangles formed by the cavity
segments as bases and the coordinate system origin as the apex multiplied by the cavity thickness. Elements forming
the cavity are assumed to be of equal thickness. For 3-D problems, the volume of the cavity is calculated as the sum
of the volumes of all the tetrahedrons formed by the cavity patches as bases and the coordinate system origin as the
apex. For axisymmetric problems, the volume of the cavity is calculated as the sum of the volumes of all the cone
frustums formed by the cavity segment as the cone slant height and two parallel base circles and with an axis along
the axis of symmetry of the problem. In the axisymmetric case, it is not necessary to use cavity surface elements along
lines that are perpendicular to the axis of symmetry to close cavities.
In coupled thermo-mechanical analysis, heat transfer between the gas inside the cavity and the surrounding structure
is not supported.
Cavities are not allowed to split or to join due to deformation or self-contact. If self-contact occurs within a cavity, the
resulting cavities are still treated as a single cavity.
If the cavity is formed out of membrane and/or shell elements, the user must ensure that all elements defining the cavity
are aligned; that is, the cavity surface is defined by either all top or all bottom element faces.
Cyclic Loading
Application of cyclic loading is important when investigating fatigue issues in a structure. There are two ways of easily
doing this in Marc. In the first method, one associates a table that is time dependent with each boundary condition.
This table can be either a sawtooth function or a sinusoidal function. The boundary condition will be repeated for as
long as one has a time period assigned. note that the AUTO STEP option can either exactly reach the peaks or not be
constrained to reach the peaks.
Table 9-4 Cavity Functioning for Subsequent Increments
Cavity Load Type Incremental Load Value Function
0 Ignored Cavity is closed, constant mass, polytropic process
1 Cavity is open, constant pressure and temperature
1 Cavity is open, pressure is applied, constant temperature
2 Cavity is closed, constant mass, isothermal process
2 Cavity is open, mass is applied, constant temperature
9 Passed to UCAV Call UCAV
p 0.0 =
p 0.0
M 0.0 =
M 0.0

The alternative is to use the BEGIN SEQUENCE and END SEQUENCE options. All history definition options placed
between these two options will be repeated as often as requested. One should make sure that the value of the boundary
conditions at the beginning of the sequence is the same as the value at the end of the sequence. Multiple loadcases may
be repeated.
Thermal Loads
Element integration point temperatures are used in a thermal stress analysis to generate thermal load. The AUTO
THERM history definition option allows automatic application of temperature increments based on a set of
temperatures defined throughout the mesh as a function of time. The CHANGE STATE option presents the
temperatures to Marc which then creates its own set of temperature steps based on a temperature change tolerance
provided through this option. Table 9-5 lists input options for thermal loads. You can input either the incremental
temperature or the total temperatures as thermal loads.

The thermal strain increment is defined as
(9-26)
where is thermal strain increment, is the coefficient of thermal expansion, and is the temperature
increment. Equivalent nodal forces are calculated from the thermal strain increment and then added to the nodal force
vector for the solution of the problem.
You can input the coefficient of thermal expansion through the ISOTROPIC, ORTHOTROPIC, or other material
options and the temperature increment through various options.
Table 9-5 Input Options for Thermal Loads
Load Description
Input Options
Parameter Model Definition History Definition User Subroutine
Incremental Temperatures* THERMAL THERMAL LOADS** THERMAL LOADS** CREDE
Total Temperatures* THERMAL CHANGE STATE CHANGE STATE NEWSV
Initial Temperature* INITIAL STATE INITSV
Total Nodal Temperatures POINT TEMP POINT TEMP
Initial Nodal Temperature INITIAL TEMP USINC
Total Nodal Temperature
Alternative Mesh
MAP TEMP
* Temperatures must be specified at each integration point (or at the centroid if the CENTROID parameter is used)
of each element in the mesh.
** The THERMAL LOADS option and CREDE user subroutine should not be used with the table driven input
procedure.
t h
T
}
=
t h
T
621
CHAPTER 9
Boundary Conditions
9
Boundary Conditions
Initial Stress and Initial Plastic Strain
Marc allows you to enter a set of initial stresses that simulate the stress state in the structure at the beginning of an
analysis. A typical example is prestress in a tensioned fabric roof. The set of initial stresses must be self-equilibrating
and should not exceed the yield stress of the material. Table 9-6 shows the input options for initial stress.

Marc also provides various ways of initializing the equivalent plastic strain throughout the model. This is useful in
metal forming analysis in which the previous amount of equivalent plastic strain is often required. This history
dependent variable represents the amount of plastic deformation that the model was subjected to, and is used in the
work (strain) hardening model. The input option is shown in Table 9-5.
Heat Fluxes
In a heat transfer analysis, you can enter heat fluxes in various forms. Heat transfer analysis requires entering the total
values of flux. Table 9-7 lists the input options for heat fluxes.

There are three special heat flux conditions representing insulation, convection, and radiation as discussed in Chapter
6: Nonstructural and Coupled Procedure Library in the Boundary Conditions section.
1. Insulation
(9-27)
No input is required for the insulated case.
Table 9-6 Input Options for Initial Stress and Initial Plastic Strain
Load Description
Input Options
Parameter Model Definition User Subroutine
Initial Stress Prestress ISTRESS INIT STRESS UINSTR
Initial Plastic Strain INITIAL PLASTIC STRAIN INITPL
Table 9-7 Input Options for Heat Fluxes
Load Description
Input Options
Point Heat Flux (Sink or Source) POINT FLUX POINT FLUX FORCDT
Surface Heat Flux, Convection, Radiation DIST FLUXES
QVECT
FILMS
SURFACE ENERGY
DIST FLUXES
QVECT
FILMS
FLUX*
UQVECT
FILM*
Volumetric Flux Load Body Flux DIST FLUXES DIST FLUXES
FILMS
FLUX*
*Can be used for complicated flux loadings, convection, and radiation, allowing the input of nonuniform
temperature- and time-dependent boundary conditions.
q 0 =

2. Convection
Nondirected:
(9-28)
You must enter the film coefficient H and ambient temperature through the FILMS model definition option
or the FILM user subroutine. You can also directly input the heat flux using the FLUX user subroutine.
Directed
(9-29)
where is the orientation defined in the QVECT option and is the outward unit normal.
3. Radiation
(9-30)
You must enter either the heat flux using the FLUX user subroutine or the temperature dependent film
coefficient and ambient temperature using the FILM user subroutine. These relationships are
shown in Equation (9-31). Using the table driven input format, you can directly specify radiation to the
environment through the FILMS model definition option. The use of the FILM user subroutine is recommended
to ensure a stable solution.
(9-31)
where is the Stefan-Boltzmann constant, is emissivity, and are unknown surface and ambient
temperatures, respectively.
As an alternative for radiation between surfaces, Marc will calculate the viewfactors using the CAVITY
DEFINITION and RAD-CAVITY model definition options or the VIEW FACTOR model definition option can be
used to read in viewfactors calculated by Mentat. The RADIATION parameter is also required.
Mass Fluxes and Restrictors
In a hydrodynamic bearing analysis, you can enter mass fluxes, restrictors, and pump pressures as loads. Table 9-8
lists input options for these quantities.
q H T
s
T ( ) =
T
q
q q
0
n
n
s
=
n
n
s
q T
s
4
T
4
( ) =
q
H T
s
T , , , ( ) T
q T
s
4
T
4
( ) =
T
s
3
T
s
2
T
T
s
T
2
T
3
+ + + ( ) T
s
T
( ) =
H T
s
T
, ( ) T
s
T
( ) =
T
s
T
623
CHAPTER 9
Boundary Conditions

Electrical Currents
In coupled thermo-electrical (Joule heating) analysis and coupled electrical-thermal-mechanical analysis (Joule-
mechanical), you can prescribe electrical currents as loads for the calculation of unknown nodal voltages. In
magnetostatic analysis, you can also define the current. In such cases, as a steady state analysis is performed, there is
no time variation of the currents. Table 9-9 lists input options for electrical currents.
Electrostatic Charges
In an electrostatic analysis, the charge can be entered, noting that a steady state analysis is performed so there is no
time variation of charge. In a coupled electrostatic structural analysis there can be a variation of charge in time, but
this variation is considered to be quasi-static. Table 9-10 summarized the input options.
Table 9-8 Input Options for Mass Fluxes and Restrictors
Load Description
Input Options
Parameter Model Definition History Definition User Subroutine
Nodal Mass Fluxes BEARING POINT MASS POINT MASS FORCDT
Distributed Mass Fluxes BEARING DIST MASS DIST MASS FLUX*
Restrictors BEARING
RESTRICTOR
RESTRICTOR URESTR**
* Can be used for nonuniform mass fluxes
** Can be used for nonuniform restrictions or pump pressures.
Table 9-9 Input Options for Electrical Currents
Load Description
Input Options
Nodal Current POINT CURRENT POINT CURRENT FORCDT
Surface and Body Currents DIST CURRENT DIST CURRENT FLUX
Table 9-10 Input Options for Electrostatic Charge
Load Description
Input Options
Nodal Charge POINT CHARGE POINT CHARGE FORCDT
Distributed and Body Charges DIST CHARGES DIST CHARGE FLUX

Acoustic Sources
In acoustic analysis, you can enter a source pressure if a transient analysis by modal superposition is being performed.
Table 9-11 summarized the input options.
Piezoelectric Loads
In a piezoelectric analysis, both mechanical loads and electrostatic charges can be entered. These values can have time
variation if a transient analysis is performed or a harmonic excitation can be applied. Table 9-1 gives a summery of
the mechanical input options, and Table 9-10 gives a summary of the electrostatic input options.
Electrostatic-Structural Loads
In a coupled electrostatic-structural analysis, both mechanical loads and electrostatic charges can be entered. Table 9-1
gives a summery of the mechanical input options, and Table 9-10 gives a summary of the electrostatic input options.
Magnetostatic Currents
In a magnetostatic analysis, the current can be entered, noting that a steady state analysis is performed so there is no
time variation of current. Table 9-12 summarizes the input options.
Magnetodynamic Currents and Charges
In an magnetodynamic analysis, the current can be entered. These values can have time variation if a transient analysis
is performed or a harmonic excitation can be applied. Table 9-13 summarizes the input options.
Table 9-11 Input Options for Acoustic Sources
Load Description
Input Options
Nodal Source POINT SOURCE POINT SOURCE FORCDT
Distributed Source DIST SOURCES DIST SOURCES FLUX
Nodal Source FIXED PRESSURE PRESS CHANGE FORCDT
Table 9-12 Input Options for Magnetostatic Current
Load Description
Input Options
Distributed and Body Current 2-D DIST CURRENT DIST CURRENT FLUX
Distributed and Body Current 3-D DIST CURRENT DIST CURRENT FORCEM
625
CHAPTER 9
Boundary Conditions
Kinematic Constraints
Marc allows you to input kinematic constraints through various options that include
Boundary Conditions (prescribed nodal values)
Transformation of Degree of Freedom
Shell Transformation
Tying Constraint
Rigid Link Constraint
Shell-to-Solid Tying
Insert
AUTOMSET
Support Conditions
Bushings
Cyclic Symmetry
Nastran RBE2 and RBE3
Beam - Shell Offsets
Pin Code for Beam Elements
Boundary Conditions
Marc allows you to specify the nodal value for a particular degree of freedom. If you do not give a nodal value when
you specify the boundary condition, Marc sets the fixed nodal value to zero. An option allows boundary conditions to
be specified at the time of two-dimensional mesh generation with MESH2D. You can apply a different set of boundary
conditions for each load increment.
Table 9-14 gives the nodal values for the various analyses, and Table 9-15 lists the input options for boundary
conditions in different analyses.
Table 9-13 Input Options for Currents and Charges
Load Description
Input Options
Distributed and Body Current DIST CURRENT DIST CURRENT FLUX
Nodal Charge POINT CURRENT
POINT CHARGE
POINT CURRENT FORCDT
Distributed and Body Charge DIST CHARGE DIST CHARGE

Table 9-14 Analyses with Corresponding Nodal Values
Analysis Nodal Values
Acoustics Pressure
Coupled Fluid Thermal Velocity, Pressure, and Temperature
Coupled Thermo-electrical Voltage and Temperature
Coupled Thermo-mechanical Displacement and Temperature
Coupled Electrical-thermo-mechanical Voltage, Temperature, and Displacement
Electrostatic Potential
Piezoelectric Displacement and Potential
Fluid Velocity and Pressure
Heat transfer Temperature
Hydrodynamic Bearing Pressure
Magnetostatic Potential
Magnetodynamic Magnetic Vector Potential and Electric Scalar Potential
Rigid Plastic Flow Velocity
Stress Displacements
Table 9-15 Input Options for Boundary Conditions
Load Description
Input Options
Model Definition History Definition
User
Subroutine
Displacement FIXED DISP DISP CHANGE FORCDT
Temperature FIXED TEMPERATURE TEMP CHANGE FORCDT
Voltage FIXED VOLTAGE VOLTAGE CHANGE FORCDT
Pressure FIXED PRESSURE PRESS CHANGE FORCDT
Potential (Electrostatic) FIXED EL-POT
FIXED POTENTIAL
FORCDT
Potential (Magnetostatics) FIXED MG-POT
FIXED POTENTIAL
Potential (Magnetodynamics) FIXED EL-POT
FIXED POTENTIAL
Velocity FIXED VELOCITY VELOCITY CHANGE FORCDT
627
CHAPTER 9
Boundary Conditions
Transformation of Degree of Freedom
Marc allows transformation of individual nodal degrees of freedom from the global direction to a local direction
through an orthogonal transformation that facilitates the application of boundary conditions and the tying together of
shell and solid elements.
Transformations are assumed to be orthogonal. Once you invoke a transformation on a node, you must input all loads
and kinematic conditions for the node in the transformed system.Nodal output is in the transformed system. This
option is invoked using the TRANSFORMATION option.
The UTRANFORM option allows transformations to be entered via the UTRANS user subroutine. This allows you to
transform the degrees of freedom at an individual node from global directions to a local direction through an
orthogonal transformation. UTRANFORM allows you to change the transformation with each increment. When you
invoke this option, the nodal output is in both the local and the global system.
Shell Transformation
The SHELL TRANSFORMATION option allows you to transform the global degree of freedom of doubly curved shells
to shell degrees of freedom in order to facilitate application of forces in the shell directions, edge moments, and
clamped or simply supported boundary conditions. There are four types of shell transformations.
The SHELL TRANSFORMATION model definition option specifies information on the shell transformation. For Types
1 and 3, only the node number has to be specified. For Types 2 and 4, a boundary direction (the direction cosine of t
as shown in Figure 9-24) also has to be specified in the surface.
Figure 9-24 Boundary Directions in Shell Transformation
After transformation, the following definitions apply.
Type 1: Transformation for two-dimensional beams and shell nodes (Element Types 15, 16, and 17). The
transformation defines the degrees of freedom with respect to a local coordinate system (s, n) (see
Figure 9-32). The four degrees of freedom after transformation are:
1 =
tangential displacement
2 =
normal displacement
1

2
,
1
2
u
s
u
n

Figure 9-25 Type 1: Shell Transformation
Type 2: Transformations for doubly curved shell nodes with nine degrees of freedom (Element Type 4 corner
nodes of Element Type 24). The transformation defines a local coordinate system ( ). (See
Figure 9-26).
The nine degrees of freedom after transformation are:
3 = rotation
4 = Meridional stretch
1 =
displacement in specified (boundary) direction
2 =
displacement normal to (boundary) direction but tangential to shell surface
3 =
displacement normal to shell surface
4 =
rotation of shell around boundary
5 =
rotation of shell around normal to boundary tangential to the shell surface
6 =
rotation of boundary around normal to shell surface
7 =
stretch tangential to specified (boundary) direction
8 =
stretch normal to specified (boundary) direction
9 =
shear stretch in t-s direction
x
2
x
1
s
Positive
Direction
n
t s n , ,
u
t
u
s
u
n
t s
629
CHAPTER 9
Boundary Conditions
Figure 9-26 Types 2 and 4: Shell Transformations
Type 3: Transformations at midside nodes for doubly curved shell nodes with three degrees of freedom
(Element Type 24). The transformation defines a local coordinate system ( ), (see Figure 9-27).
The degrees of freedom after transformation are:
Figure 9-27 Type 3: Shell Transformation
Type 4: Transformation for doubly curved nodes with 12 degrees of freedom (Element Type 4). The
transformation defines a local coordinate system ( ). The first 9 degrees of freedom after
transformation are the same as 1 through 9 in Type 2 (Figure 9-26), and the remaining three are:
variation of g along the boundary (9-32)
variation of e along the boundary (9-33)
variation of f along the boundary (9-34)
1 =
rotation of normal to boundary around normal to shell
2 =
stretch normal to boundary
3 =
rotation around boundary
s
n
t
2
x
3
1
x
2
x
1
t s n , ,
u
t
s
u
s
s
u
n
s
s
n
t
x
3
x
2
x
1
t s n , ,
10
2
u
t

1
2
=
11
2
u
s

1
2
=
12
2
u
n

1
2
=

When using the SHELL TRANSFORMATION option, the displacement increments and reaction forces are output in the
local directions immediately after solution of the equations. At the end of the increment, you can print out the global
displacement increments and total displacements in the global coordinate directions.
If you invoke the FOLLOW FOR parameter, the SHELL TRANSFORMATION option defines a local coordinate system
in the current (updated) geometry of the structure. This additional feature is especially useful with the LARGE STRAIN
parameter, because you can then specify edge moments and/or large edge rotations of shells and beams.
Tying Constraint
Marc contains a generalized tying (constraint) condition option. Any constraint involving linear dependence of nodal
degrees of freedom can be included in the stiffness equations.
A tying constraint involves one tied node and one or more retained nodes, and a tying (constraint) condition between
the tied and retained nodes. The degrees of freedom (for example, displacements, temperatures) of the tied node are
dependent on the degrees of freedom of the retained nodes through the tying condition. In some special tying
conditions, the tied node can also be a retained node. The tying condition can be represented by a tying (constraint)
matrix. Note that if the tying constraint involves only one retained node, the choice of which node is to be tied to
retained is arbitrary.
As a simple example, impose the constraint that the first degree of freedom of node I be equal to that of node J at all
times (see Figure 9-28).
As a second example, the simulation of a sliding boundary condition requires the input of both the boundary conditions
and the tying constraints (see Figure 9-29).
Figure 9-28 Simple Tying Constraint
Caution: If you apply a shell transformation to a node, do not apply a standard transformation or shell tying type to
that node.
2,V
1,U
U
J
U
I
I
J
Constraint Equation:
U
I
= U
J
Tied node I, retained node J, or
Tied node J, retained node I
631
CHAPTER 9
Boundary Conditions
Figure 9-29 Tying Constraint Illustration (Sliding Boundary Conditions)
The example illustrated in Figure 9-29 enforces rigid sliding on the boundary in the local coordinates defined above.
(9-35)
(9-36)
The first equation is a set of fixed boundary conditions. The second equation is a constraint equation and can be
rewritten as three constraint equations:
(9-37)
(9-38)
(9-39)
These equations express all the displacements in terms of . In this example, node j is chosen to be the retained
node; nodes i, k, and l are tied nodes. You can use the TYING option to enter this information.
Marc has a number of standard tying constraints that can be used for mesh refinement, shell-to-shell, shell-to-beam,
beam-to-beam, and shell-to-solid intersections. Tables 9-16 through 9-19 describe these options. Tables Table 9-21
through Table 9-23 show the tying constraints for pipe elements, shell stiffeners, and nodal degrees of freedom.
Table 9-24 summarizes the rigid link constraint.
The SERVO LINK option uses the homogeneous linear constraint capability (tying) to input simple constraints of the
form
(9-40)
where is a degree of freedom to be constrained; are the other retained degrees of freedom in this
structure; and are constants provided in this option.
You can use the TYING or SERVO LINK model definition option to enter standard tying constraint information, and
the TYING CHANGE option to change tying constraints during load incrementation.
The UFORMSN user subroutine is a powerful method to specify a user-defined constraint equation. This constraint
can be nonlinear; for example, it can be dependent on time or previous deformation.
Y (v)
X (u)
Local axes
X
Y
i
j
k
l
v
i
v
j
v
k
v
1
0 = = = =
u
i
u
j
u
k
u
1
= = =
u
i
u
j
=
u
k
u
j
=
u
1
u
j
=
u u
j
u
t
a
1
u
r 1
a
2
u
r 2
+ + =
u
t
u
r 1
u
r 2
, ,
a
1
a
2
, ,

Figure 9-30 Mesh Refinement for 4-Node Quad
Figure 9-31 Mesh Refinement for 8-Node Quad
Table 9-16 Tying Constraints for Mesh Refinement
Tying
Code
Number of
Retained
Nodes Purpose Remarks
31 2 Refine mesh of first order
(linear displacement) elements in two
dimensions
Tie interior nodes on refined side to
corner nodes in coarse side (see
Figure 9-30)
32 3 Refine mesh of second order (quadratic
displacement in two dimensions
Tie interior nodes on refined side to edge
of element of coarse side (see
Figure 9-31)
33 4 Refine mesh of 8-node bricks Tie interior node on refined side to four
corner nodes of an element face on coarse
side (see Figure 9-32)
34 8 Refine mesh of 20-node bricks Tie interior nodes on refined side to eight
(four corner, four midside) nodes of
element face on coarse side (see
Figure 9-33)
T Tied Node
R Retained Node
R
T
R
T Tied Node
R Retained Node
T
T
T
T
R
R
R
633
CHAPTER 9
Boundary Conditions
Figure 9-32 Mesh Refinement for 8-Node Brick
Figure 9-33 Mesh Refinement for 20-Node Brick

Table 9-17 Tying Constraints for Shell-to-Shell Intersection
Tying
Code
Number of
Retained Nodes Purpose Remarks
22 2 Join intersecting shells, Element Type 4,
8, or 24; fully moment-carrying joint
Tied node is also the second retained
node*
8, or 24; fully moment-carrying joint
node
8, or 24; pinned point
node
24** 2 Join intersecting shells or beams, Element
Types 15-17
node
* Thickness vector must be specified at tied nodes and retained nodes.
** See Figure 9-34.
R
T
R
R
R
T
T
T
T
R
R
R
R

Figure 9-34 Standard Tying Type 24, Tie Shell-to-Shell or Beam-to-Beam; Moment-Carrying
Table 9-18 Tying Constraints for Beam-to-Beam Intersection
Tying
Code
Number of
Retained
13 2 Join Two Elements Type 13 under an
arbitrary angle; full moment-carrying
joint
Tied node also the second retained node
52 1 Pin joint for Beam Type 14, 25, 52, 76-
79, 98
53 1 Full moment-carrying joint for Beam
Types 14, 25, 52, 76-79, 98
Tie interior node on refined side to four
corner nodes of an element face on
coarse side (see Figure 9-32)
Table 9-19 Tying Constraints for Shell-to-Solid Intersections
Tying
Code
Number of
Retained
23 1 Tie axisymmetric solid node to
axisymmetric shell
(Element Type 1 or 89) node
Tied and retained nodes must be
transformed to local system and
TRANSFORMATION option invoked
(see Figure 9-35).
25 2 Join solid mesh to beam (Type 16)
Join solid mesh to axisymmetric shell
(Type 15)
Tied node also second retained node.
(see Figure 9-36)
26 2 Tie axisymmetric solid node to
axisymmetric shell
(Element Type 1 or 89) node
Similar to tying type 23 but no
transformation is required. Tied node
also second retained node. (see
Figure 9-36)
S
A B
S
2
1
635
CHAPTER 9
Boundary Conditions
Figure 9-35 Standard Tying Type 23, Tie Solid-to-Shell (Element Type 1 or 89)
Figure 9-36 Standard Tying Type 25, Tie Solid-to-Beam (Element Type 16)
Standard Tying Type 25, Tie Solid-to-Axisymmetric Shell (Element Type 15)
Standard Tying Type 26, Tie Solid-to-Axisymmetric Shell (Element Type 1 or 89)
Table 9-20 Tying Constraints for Beams to Element Edge or Face or Axisymmetric Shell to Element Edge
Tying
Code Number of Retained Nodes Purpose Remarks
44 2 is 2-D or axisymmetric
lower-order element edge
3 is 2-D or axisymmetric
higher-order element edge
3 if 3-D lower-order triangular
face
4 if 3-D lower-order
quadrilateral face
6 if 3-D higher-order
triangular face
8 if 3-D higher-order
quadrilateral face
Rigidly tie a node with
displacements and rotations
to a surface patch. This is
internally used for CWELD
and CFAST option.
The number of retained
nodes is required. This tying
type fully supports large
deformation/rotations. No
transformations are required.
See Figure 9-37.
T
2
T
1
R
x
y
B
A
T
1
T
2
R
T
3
T
4

Figure 9-37 Beam to Edge Moment Carrying Constraints
R
R
T
R
T
R
R
T
R
R
R
R
R
R
T
R
R
R
R
3-D Application Lower-order Quadrilateral Face
2-D Application Higher-order
3-D Application Higher-order Triangular Face
2-D Application Lower-order
637
CHAPTER 9
Boundary Conditions

Table 9-21 Tying Constraints for Pipe Bend Element (Elements 14 and 17)
Tying Code Number of Retained Nodes Purpose
15 One less than the number of shell nodes
in the z-r plane of the section
Special tying types for pipe bend Element 17 to remove
rigid body modes
16 Number of shell nodes in the z-r plane of
the section
Special tying types for pipe bend Element 17 to remove
rigid body modes
17 2 Special tying types for pipe bend Element 17 to couple
bend section into pipe bend
Table 9-22 Tying Constraints for Shell Stiffener (Element 13 as a Stiffener on Shell Elements 4 or 8)
Tying
Code
Number of
Retained
19 2 Use beam Element 13 as a stiffener on shell
Elements 4 or 8. Tied node is beam node; first
retained node is shell node, second is beam
node again. Beam node should be on or close
to the normal to the shell at the shell node
Tied node also second
retained node
20 3 Create an extra node in a shell Type 8 Element
tied to the interpolated shell displacements
with tying Type 21 to tie a beam Element 13 or
a stiffener across a shell element
Always used after tying Type 21
21 2 Same as Type 19, but tying beam to an
interpolated shell node not a vertex of an
element (Element Type 8 only) must be
followed by Type 20 to tie interpolated shell
node into shell mesh
Must be followed by tying Type 20

Rigid Link Constraint
Tying type 80 can be used to define a rigid link between nodes. This capability can be used for both small deformation
and large deformation, large rotation problems. In small deformations, a linear constraint equation is used. In addition
to the end points of the link, a second retained node must be given. This node is used to calculate and store the rigid
body rotations. For two-dimensional problems, this represents a rotation about the Z-axis and this is stored as the first
degree of freedom for the node. In 3-D, it should be noted that the rotations/moments about the x, y, and z axes for the
second retained node are stored and output as its first, second, and third degrees of freedom, respectively.
A complete rigid region can be modeled by using multiple tying type 80s. In this case, the same two nodes are used
for the retained nodes in all of the constraints. Transformations can be applied to the retained nodes so that any
kinematic or force boundary conditions can be defined in a local coordinate system. It is not necessary that the same
transformations should be applied to both retained nodes. It is also optional to transform the tied nodes to the same or
a different local coordinate system.
The rigid links can be used with all elements except types 4, 8, 24, 15, 16, and 17. In addition, it should not be used in
rigid plastic analysis.
To reduce linearization errors in large deformation, large rotation problems using tying type 80, an optional recycling
feature is offered through the CONTROL model and history definition options. Recycling occurs if the maximum
change over the retained nodes of rigid link 80 exceed a rotation tolerance (default is 0.001 radians). This check can
be circumvented by setting the rigid link rotation tolerance to zero (in conjunction with the FEATURE,5701 parameter)
or a negative number.
Table 9-23 Tying Constraints for Nodal Degrees of Freedom
Tying
Code
Number of
Retained
INDEG 1 Tie the Ith degree of freedom at the tied
node to the Ith degree of freedom at the
retained node
NDEG = number of degrees of freedom
per node
100 1 Tie all degrees of freedom at the tied
node to the corresponding degrees of
freedom at the retained node
>100 1 Generate several tyings of type NDEG Tying code is first degree of freedom
multiplied by 100 added to last degree of
freedom (209 means tie second through
ninth degree of freedom at tied node to
respective second through ninth degree
of freedom at retained node)
<0
(negative
integer)
User-defined User-generated tying type through the
UFORMSN user subroutine
639
CHAPTER 9
Boundary Conditions
Shell-to-Solid Tying
In many problems, a region exists that is modeled with both brick elements and shell elements. A particular case of
this is shown is Figure 9-38 and Figure 9-39.
Figure 9-38 4-Node Shell-to-Solid Automatic Constraint
Figure 9-39 8-Node Shell-to-Solid Automatic Constraint
In the first case, an 8-node brick which has been reduced to a triangular prism is connected to a 4-node shell. In the
second case, a 20-node brick is connected to an 8-node shell. An automatic constraint equation is developed between
the elements. Note that the thickness of the shell must be entered as the brick thickness.
Rigid Tying to a Surface Patch
Tying type 44 rigidly ties a node with its translation and rotation degrees of freedom to a surface patch. There are four
different tying types for 3-D analyses tying the displacements and rotations of a node (assuming it has 6 degrees of
freedom) to the motion of a 3- or 6-node triangular patch or a 4- or 8-node quadrilateral patch. There are two different
tying types for 2-D analyses, tying the displacements and rotation of a node (assuming it has three degrees of freedom)
to the motion of a 2- or 3-node line patch. The tied node can be a node of a beam or a shell element and will generally
have translation and rotation degrees of freedom. The retained nodes are the nodes of the patch and the only retained
degrees of freedom involved in the tying are the patch node translations. The exact tying type is determined from the
tying type number (44) and the number of retained nodes and, therefore, the latter is required in the input. The location
Table 9-24 Rigid Link Constraint
Tying
Code
Number of
Retained
80 2 Define a rigid link between nodes The second retained node is an unattached node
which contains the rotation

where the tied node pierces the patch is computed in terms of the local parametric coordinates that define locations
inside a patch using the usual shape functions for such a patch. The translations of this piercing point are computed
from the nodal displacements of the patch, again using the usual shape functions of the patch and these define the
translations of the tied node. The rotations can be computed by constructing a local triad to the patch at the piercing
point. The translations of the patch nodes define the change in orientation of the triad. The orientation change of the
triad defines the rotations of the tied node. The tied node must have a proper projection on the patch inside its
boundary. When it does not lie exactly on the patch, the distance will be treated as an offset. The patch does not
necessarily have to coincide with a shell or brick element face in a 3-D model or a beam or quad element edge in a 2-
D model, but the nodes constituting a patch must adhere to the conventions that are generally required for the particular
patch. The patch also does not have to be a flat surface in 3-D or a straight line in 2-D, but it is recommended to avoid
overly distorted patches. A patch is considered overly distorted when the patch normals at two of its nodes make an
angle larger than 10 degrees.
In case of a heat transfer or a thermo-mechanically coupled analysis the temperatures may be tied as well. The
temperature constraints are similar to the mechanical displacement constraints, but only one degree of freedom is
involved. When the retained nodes are shell nodes, the temperature is always tied to the first degree of freedom of the
shell nodes (i.e., the top surface of the shell), regardless of the orientation in space. When the tied node is a shell node,
its first degree of freedom is tied. In other analysis procedures, tying type 44 is not supported.
When no geometric nonlinearities are flagged in the model (i.e., neither the LARGE DISP nor the LARGE STRAIN
parameter is activated), Marc uses the small displacement method to apply the constraints. In all other cases, the large
displacement method is used fully accounting for finite rotations. The small displacement method linearizes the
constraint equations w.r.t. the reference configuration. The large displacement method takes account of all geometric
nonlinearities caused by finite displacements and rotations.
Figure 9-40 Motion of a Patch From Start to End of an Increment
The large displacement method (Figure 9-40) can be summarized as follows. The point GA is the node of, for example,
a beam and it pierces the patch at some point for which the parametric coordinates , have been computed.
A local triad is fixed to the beam cross section and another triad to the piercing point on the patch. For simplicity, it is
assumed that both triads have the same orientation in the initial configuration. The tying constraints are that the
e
3
e
2
e
1
GA
Patch at end
Patch at start
of increment
of increment
GA

GA
641
CHAPTER 9
Boundary Conditions
displacements of the beam node and the displacements of the piercing point on the patch are equal and that the
rotations of the triad connected to the beam cross-section and the triad connected to the patch are equal. The rotations
of the triad fixed to the beam cross-section are expressed in terms of the rotational degrees of freedom of the beam
node. The rotations of the triad fixed to the patch are expressed in terms of the displacement degrees of freedom of the
patch nodes. In this way, the nonlinear relations are established between the beam rotations and the patch
displacements. The relations between the displacements of the beam node and the displacements of the patch nodes
are linear when the beam node lies exactly on the patch. In case of an offset, nonlinear contributions due to the rotations
are added to the constraints.
The tied node generally has translation and rotation degrees of freedom, but this is not a requirement. In case it has
only translation and rotation degrees of freedom, the displacements are tied to the patch, possibly accounting for an
offset. No constraints are applied for rotations; hence, adequate constraints must be applied to prevent a singular
system.
The connectivity conventions for the patches are shown in Figure 9-41 for 2-D and 3-D models:
Figure 9-41 Patch Connectivity Convention
1 2
2-D Model
Line2
Line3
1 2 3
(a) Line Patch in 2-D Analysis
3-D Model
Quad4
Quad8
1 2
3
(b) Quadrilateral Patch in 3-D Analysis
1 2
3 4
8
4 7
5
6
3-D Model
Tria3
Tria6
1 2
3
(c) Triangular Patch in 3-D Analysis
1 2
3
4
5 6

Overclosure Tying
In various engineering applications, it is necessary to define a pre-stress in, for example, bolts or rivets before applying
any other structural loading. Although such a pre-stressed state is often simulated using a temperature loading, it is
rather difficult to arrive at a desired net force in the bolt or rivet. An easier way is to use overclosure tyings (type 69).
These tyings create overlaps or gaps between two parts of a model. If the motion of these parts is somehow constrained
in the direction in which the gap or overlap is being created, then an overlap will introduce an tensile (pre-) stress in
each of the parts and a gap will result in a compressive stress.
Overclosure tyings have one tied node and two retained nodes. The tied node and the first retained node are usually
nodes on the boundaries of the respective parts (see Figure 9-42). The second retained node is very often a free node
and is usually shared by all overclosure tyings which connect the parts. This node is also called the control node of the
tying, since it can be used to apply load or control the size of the gap or overlap between the parts.
Figure 9-42 Pre-stressing a Structure by Creating an Overlap Between the top and the Bottom Part Using Overclosure
Tyings
An overclosure tying imposes the following constraint on the model:
, (9-41)
in which , and are the displacement and rotation degrees of freedom (if any) of, respectively,
the tied node, the first retained node, and the control node of the tying. It immediately follows from this equation that
is the displacement difference of the tied and the retained node of the tying and is equal to the size of the
overlap or gap between the parts. Hence, by prescribing this displacement using the FIXED DISP and DISP CHANGE
options, gaps or overlaps of a particular size can be created.
Instead of prescribing the size, gaps or overlaps can also be created by prescribing the total force in the model. This
follows immediately from the fact that the work done by a constraint equation is zero. To demonstrate this, consider
the model displayed in Figure 9-42. The model is split into two disjoint parts: the top part and the bottom part. At the
split, corresponding nodes of the respective parts are connected by overclosure tyings, in which the nodes of the
bottom part are the tied nodes and the nodes of the top part are the first retained nodes of the tyings. Both tyings share
a common (free) control node. If and are the displacements of the bottom nodes and and
are the displacements of the top nodes, then the two overclosure tyings impose the following two constraints
on this model:
control node
bottom part
overclosure tyings
(tied)
(first retained)
top nodes
top part
mesh split
bottom nodes
(second retained)
undeformed
F
1,bot
F
2,bot
F
2,top
F
control
top part
bottom part
F
1,top
u
1,top
u
2,top
u
2,bot
u
1,bot
u
control
(overlap) u
control
deformed
u
tied
u
retained
u
control
+ =
u
tied
u
retained
u
control
u
control
u
1 bot ,
u
2 bot ,
u
1 top ,
u
2 top ,
643
CHAPTER 9
Boundary Conditions
, (9-42)
. (9-43)
Introducing the vectors,
and ,
and the constraint matrix
,
in which and are the unit and null matrices, Equations (9-42) and (9-43) can be summarized into the matrix
equation:
. (9-44)
Let
and
be the force vectors which are work conjugate to the displacements and , respectively. Then the zero-work
principle states that
. (9-45)
Substitution of Equation (9-44) into Equation (9-45) and requiring that the result is valid for all , it follows that
, or
, (9-46)
, (9-47)
. (9-48)
Equations (9-46) and (9-47) express the equilibrium of forces across the split, in which and can be
viewed as externally applied forces. While Equation (9-48) enables overclosure tyings to pre-stress a model with a
certain total force. It states that if the constraints described by Equations (9-42) and (9-43) are applied, the force on
u
1 bot ,
u
1 top ,
u
control
+ =
u
2 bot ,
u
2 top ,
u
control
+ =
u
u
1 bot ,
u
2 bot ,
u
1 top ,
u
2 top ,
T
= u
u
1 top ,
u
2 top ,
u
control
T
=
T
I 0 I 0
0 I 0 I
I I 0 0
T
=
I 0
u Tu
=
F
F
1 bot ,
F
2 bot ,
F
1 top ,
F
2 top ,
T
=
F
1 top ,
F
2 top ,
F
control
T
=
u u
F
T
u F
T
u
=
u
T
T
F F
=
F
1 bot ,
F
1 top ,
+ F
1 top ,
=
F
2 bot ,
F
2 top ,
+ F
2 top ,
=
F
1 bot ,
F
2 bot ,
+ F
control
=
F
1 top ,
F
2 top ,

the control node is the sum of the forces on the tied nodes of all overclosure constraints which share that control node.
Hence, the total force on the bottom part of the model is prescribed by applying a force to the control node using the
POINT LOAD option. Obviously (by the equilibrium Equations (9-46) and (9-47)) if no external force is applied to the
nodes on the split, the total force on the top part is equal but opposite is sign to the force on the control node.
The control node of an overclosure tying has the same displacement and rotation degrees of freedom as the tied and
the first retained node as given by Equation (9-41). This allows creation of overlaps or gaps in any particular direction
or combination of directions. Local coordinate systems can be defined at any of the nodes of an overclosure tying using
the TRANSFORMATION option. Since the constraint Equation (9-42) is always applied in the global coordinate
system, a local coordinate system at the control node can be used to pre-stress the model in directions other than the
global coordinate directions.
Sufficient boundary conditions must be applied on the control node to suppress any rigid body modes, if the two parts
of the structure are not constrained otherwise. Total forces on the split are available for post-processing as
reaction forces on the control node for all suppressed or prescribed displacements and rotations.
Overclosure tyings can be used in combination with the CONTACT option, that is, the tied and the first retained node
may be nodes on the boundary of a contact body.
In non-mechanical passes of a coupled analysis, the overclosure tying reduces to tying type 100 between the tied node
and the first retained node. There is no dependence on the control node in that case. This guarantees continuity of the
primary field variable (for example, temperature) across the split.
Notes:
Both methods to control the size of overlaps or gaps can be combined in one analysis. For example, in the
first part of an analysis, a pre-stress can be defined by prescribing the net force in the structure using the
POINT LOAD option on the control node of the overclosure tyings. Then in the second part, the size of the
gap or the amount of overlap which is the result of this load, can be fixed by suppressing the displacement
changes of the control node using the DISP CHANGE option.
If similar arguments are applied to tying 100 (which yields Equation (9-41) without the dependence on the
control node), then only the equilibrium Equations (9-46) and (9-47) are obtained. Equation (9-48) is solely
due to the dependence of Equation (9-41) on the control node.
Table 9-25 Overclosure Tying
Tying
Code
Number of
Retained
69 2 Create gaps or overlaps between two parts of
a model either by prescribing the total force
on the nodes on either side of the gap or
overlap or by prescribing the size of the gap
or overlap.
The second retained node is the control node
of the tying. The force on this node is equal
to the total force on the tied nodes of all
tyings that share this control node. The
displacement of the node is equal to size of
the gap or overlap between the parts. In non-
mechanical passes, the tying reduces to tying
type 100 between the tied and the first
retained node.
645
CHAPTER 9
Boundary Conditions
9
Boundary Conditions
Insert
Marc provides an INSERT model definition option which allows the definition of host bodies and lists of elements or
nodes to be inserted in the host bodies. The degrees of freedom of the nodes in the inserted node list or element list are
automatically tied using the corresponding degrees of freedom of the nodes in host body elements based on their
isoparametric location in the elements.
The INSERT model definition option can be used to place reinforcing cords or rods, such as 2-D rebar membrane
elements, into solid elements.
The INSERT model definition option can be used to apply point loads in some specific locations other than element
nodes. It also can be used to link two different meshes.
If a node to be inserted is also a node of a host body element, no tying is applied to the node.
Transformation must not be used at any nodes of host body elements and at inserted nodes, unless the same set of local
coordinate system is used for all nodes involved.
AUTOMSET
This parameter automatically re-writes the constraint equations such that a tied degree of freedom is not used as a
retained degree of freedom in another constraint equation. It also modifies constraint equations where a constrained
degree of freedom is used in a prescribed boundary condition. Because of chaining of constraint equations, this is not
always possible, in which case the analysis will terminate. The constraint equations that are considered are those
obtained from user-specified options:
TYING
SERVO LINK
RBE2
RBE3
RROD
CONRAD GAP
Constraint equations internally generated are not considered from the following options:
CONTACT
INERTIA RELIEF
INSERT
If both the MPC-CHECK and AUTOMSET parameters are included in the model, the MPC-CHECK parameter is
ignored.
Support Conditions
Marc provides linear and nonlinear springs and foundations for the modeling of support conditions. For dynamic
analysis, a dashpot can also be included. Table 9-26 lists input options for linear and nonlinear springs and
elastic foundations.

The force in the spring is
(9-49)
where is the spring stiffness, is the displacement of the degree of freedom at the second end of the spring, and
is the displacement of the degree of freedom at the first end of the spring. In a dynamic analysis, is the damping
factor and and are the velocities of the nodes.
You can specify the elements in Marc to be supported on a frictionless, elastic foundation. The foundation supports
the structure with a force per unit area (force per unit length, for beams) given by
(9-50)
where is the equivalent spring stiffness of the foundation, and is the normal displacement of the surface at a
point in the same direction as .
The same conventions apply to the elastic foundation specification as for pressure specification, in terms of the face
of the element that is used. The force is applied whether the displacement is tensile or compressive.
Nonlinear spring stiffness for mechanical, thermal, and electrical analyses can be specified through tables using the
parameter and model definition option. For more details, refer to Chapters 2 and 3 in Marc Volume C: Program Input.
The USPRNG user subroutine allows you to supply or further modify a nonlinear spring stiffness or specify a nonlinear
foundation stiffness as a function of prior displacement and force.
Bushings
A generalized spring-damper capability is provided in Marc to model bushings between various components. They are
modeled using element types 194 and 195 for 2-D and 3-D applications, respectively. The connectivity of these
elements is generally comprised of two nodes which are specified through the CONNECTIVITY model definition
option. Grounded bushing elements can be defined by providing only the first end-node for connectivity. The
properties of these bushing elements are specified through the PBUSH or PFAST model definition option. These
elements may be automatically generated using the CFAST option.
Table 9-26 Input Options for Springs and Elastic Foundations
Load Description
Input Options
Springs SPRINGS USPRNG
Elastic Foundation FOUNDATION FOUNDATION USPRNG
F K u
2
u
1
( ) D v
2
v
1
( ) + =
K u
2
u
1
D
v
2
v
1
p
n
Ku
n
=
K u
n
p
n
647
CHAPTER 9
Boundary Conditions
Bushing Coordinate System
A unique coordinate system is used to specify the bushing properties. A number of options are available to define this
coordinate system:
local element coordinate system along the bushing element
global cartesian coordinate system, or
a user-specified coordinate system specified through the COORD SYSTEM option.
The bushing coordinate system defined through global cartesian coordinates or through the user-specified COORD
SYSTEM option are maintained constant and are not updated for large displacement analysis.
The local element coordinate system can only be used for 2-node bushing elements. In this case, the local X axis is
along the element length and forms a perpendicular triad with the local Y and Z axes. In 2-D, the local Z axis is always
taken as (0,0,1) and the local Y axis is then perpendicular to the local X axis in the global x-y plane. In 3-D, the local
Y-Z axes are located by using an extra node or an orientation vector. If both are specified, the extra node takes
precedence. The local Y axis is first estimated by using the orientation vector or using the vector from the first bushing
node to the extra node. The local Z axis is then formed by taking the cross product of the local X and estimated local
Y axis. The final local Y axis is obtained by taking the cross product of the local Z and local X axes. For large
displacement analysis, the local element coordinate system is updated continuously. Mid-increment direction cosines
are used to calculate the strains while end-increment direction cosines are used to calculate the stiffnesses and internal
bushing forces.
Nodal Offsets for Bushing Elements
The exact position of the bushing element can be modified by specifying nodal offsets. This offset point is generally
defined from the first end-node of the bushing element. It serves two purposes: It allows convenient modeling of the
bushing at a location away from the actual physical location. Also, it allows coupling between translational and
rotational stiffnesses. The offset point can be located along the bushing element axis, in the global cartesian coordinate
system, or in a user-defined coordinate system. In the first case, the offset point is located along the line joining the
two end nodes of the bushing element. The distance of the offset point from the first node S is specified by the user.
In the second case, the offset vector to the bushing location from the first end-node is specified.
The treatment of offsets for bushing elements closely follows the treatment of offsets for beam elements (described
later in Beam - Shell Offsets). Once the offset point location is specified, the offset vector from each node is
determined. The equations governing the treatment and update of the offsets are given by Equations (9-62) to (9-65).
Bushing properties are used to define the stiffness/damping/mass matrix at the offset position and are then transformed
to the geometrical position.

Figure 9-43 Offset location Specification for Bushing Element
Bushing Properties
Five sets of properties can be provided for bushing elements depending on the analysis type. All properties are
provided through the PBUSH or PFAST model definition option:
mechanical stiffness - for static/dynamic/harmonic analysis
mechanical dashpot - for dynamic/harmonic analysis
mass - for dynamic/harmonic analysis
thermal conduction coefficient - for heat transfer/coupled analysis involving a thermal pass
electrical resistance coefficient - for electrical/coupled analysis involving a joule pass
Mechanical stiffness for bushing elements can be specified either directly in coefficient form or through force-
displacement curves. Note that 2-D bushing elements have two translational and one rotational degree of freedom and
separate stiffnesses can be specified for the X, Y and RZ degrees of freedom. Three-dimensional bushing elements
have three translational and three rotational degrees of freedom and separate stiffnesses can be specified for X, Y, Z
and RX, RY, RZ degrees of freedom. For nonlinear bushing elements, these stiffnesses can be varied through tables.
These tables can be a function of time, increment number, position coordinates, frequency (only for harmonic
analysis), temperature, displacement (this variable is mandatory when stiffness is specified via forces). Up to four
independent variables can be used in the tables. Based on the current stiffness value and the change in displacement
between the end-nodes, the bushing forces are updated in the specified bushing coordinate system.
Two forms of damping can be specified for the bushing elements: Nominal damping for the ith degree of freedom
of the bushing elements where the damping is specified either directly in coefficient form or through force-velocity
curves. For nonlinear dampers, the nominal damping can be varied through tables as a function of time, increment
number, position coordinates, frequency (only for harmonic analysis), temperature or velocity (this variable is
mandatory when nominal damping is specified via forces). The second form of damping is structural damping Si for
the ith degree of freedom of the bushing elements. The damping is specified in coefficient form only in this case and
can be varied through tables with similar independent variables as the nominal damping case. A composite damping
coefficient is formed for the bushing element as follows:
1
2
x
elem
S . l
(1-S) . l
Bush Location
1
2
x
elem
Bush
Location
y
elem
z
elem
v
y
elem
z
elem
(S
1
,S
2
,S
3
) OCID
CID
B
i
649
CHAPTER 9
Boundary Conditions
where is the total damping coefficient; G, W3, and W4 are structural coefficients that are specified through the
COEFFICIENT model definition option; is the nominal damping coefficient; is the structural damping
coefficient; and is the current stiffness coefficient.
For a linear bushing element, the total force is given by Equation (9-49). Note that for bushing elements defined in the
local element coordinate system, the force is positive if the bushing element is in tension and negative if it is in
compression. For bushing elements defined via the global coordinate or user-coordinate systems, the force along a
particular degree of freedom is positive if the displacements of end-node 2 along the specified degree of freedom is
greater than the displacement of end-node 1 along the specified degree of freedom. Note that if tables are used to define
the bushing properties, care should be taken to include both positive and negative values of independent variables like
displacements/velocities since these would influence the calculated stiffnesses.
For dynamic and harmonic analyses, lumped translational and rotational masses can be provided at the end-nodes of
the cbush elements. Uniform masses apply to translational and rotational degrees of freedom at the end-nodes and are
maintained constant through the analysis.
Stresses in the bushing elements can be related to the associated forces through recovery coefficients. Similarly, strains
in the bushing elements can be related to the associated deformations through recovery coefficients. Separate
coefficients are allowed for the translational components and the rotational components through the PBUSH option.
If set to 1.0, the stresses are equal to the forces and the strains are equal to the difference in displacements between the
end-nodes.
The bushing elements can also be used for thermal or electrical field analyses. The coefficients of thermal conduction
and electrical resistance are specified via the PBUSH option. The bushing element behaves like a general link in this
case and dashpot properties are not considered. Tables can be used to vary the properties as a function of temperature
(for thermal conduction), voltage (for electrical resistance), time, increment number, coordinates.
Cyclic Symmetry
A special set of tying constraints for continuum elements can be automatically generated by the Marc program to
effectively analyze structures with a geometry and a loading varying periodically about a symmetry axis. Figure 9-44
shows an example where, on the left-hand side, the complete structure is given and, on the right-hand side, a sector to
be modeled.
C
i
B
i
G
W
3
-------- K
i
1
W
4
-------- S
i
K
i
+ + =
C
i
B
i
S
i
K
i

Figure 9-44 Cyclic Symmetric Structure: Complete Model (Left) and Modeled Sector (Right)
Looking at points and on this segment, the displacement vectors should fulfil:
(9-51)
which can also be written as:
(9-52)
where the transformation matrix depends on the symmetry axis (which, in the example above, coincides with the
global Z-axis) and the sector angle (see Figure 9-44). In Marc, the input for the CYCLIC SYMMETRY option
consists of the direction vector of the symmetry axis, a point on the symmetry axis and the sector angle . The
following items should be noted:
1. The meshes do not need to line up on both sides of a sector (for example, see Figure 9-45).
Figure 9-45 Finite Element Mesh for Cyclic Symmetric Structure with Different Mesh Densities on the Sector
Sides
2. Any shape of the sector sides is allowed provided that upon rotating the sector times about the
symmetry axis over the sector angle will result in the complete model.
3. The CYCLIC SYMMETRY option can be combined with the CONTACT option. In this case, both sides of the
cyclic symmetry sector must belong to a contact body and corresponding parts of the sector sides must belong
to the same body (see Figure 9-46). Note that the contact bodies do not need to span between the sector sides.
X
Y
X
Y
x y
A
B
A B
u'
B
u
A
=
u
B
Ru
A
=
R
360
651
CHAPTER 9
Boundary Conditions
Figure 9-46 Cyclic Symmetry and Contact Bodies; upper left: one body (correct); upper right: two bodies
(correct); lower left: two bodies (right); lower right: two bodies (wrong; corresponding parts of the
sector sides are not in the same body).
4. The CYCLIC SYMMETRY option can be combined with global remeshing.
5. In a coupled thermo-mechanical analysis, the temperature is forced to be cyclic symmetric ( as in
Figure 9-44).
6. A nodal point on the symmetry axis is automatically constrained in the plane perpendicular to the symmetry
axis.
7. The possible rigid body motion about the symmetry axis can be automatically suppressed.
8. Cyclic Symmetry is valid for:
a. only the continuum elements. However, the presence of beams and shells is allowed, but there is no
connection of shells to shells, so the shell part can, for example, be a turbine blade and the volume part is a
turbine rotor. The blade is connected to the rotor and if there are 20 blades, 1/20 of the rotor is modeled and
one complete blade.
b. nonlinear static analysis including remeshing as well as coupled analysis.
c. valid for all analysis involving contact.
d. valid also for: eigenvalue analysis such as buckling or modal analysis, harmonic analysis, and transient
dynamic analysis. However, there are restrictions in the case of modal analysis which are described in more
detail in the following paragraphs.
9. Cyclic Symmetry is invalid for pure heat transfer.
10. In order to check which nodes get tying constraints due to cyclic symmetry, nodal post code 38 (contact status)
can be used; its value will be 2.
T
A
T
B
=

To prevent confusion, it must be emphasized that the cyclic symmetry feature described above is different than linear
cyclic symmetry commonly used in modal analysis where physical quantities such as, , displacements, forces,
stresses and temperature in the n-th segment are expanded in a Fourier series with terms of the cyclic components, ,
in the fundamental region, like:
(9-53)
where k is the harmonic order; N is the total number of sectors; is the fundamental inter-sector phase shift defined
as ; and K is defined as:
(9-54)
There are considerable savings in both computing time and data storage associated with the use of the linear cyclic
symmetry concept. Assuming a finite element model with a sector size of m degrees of freedom, a real-valued cyclic
symmetry approach leads, in the worst case, to one eigenvalue problem of size m and eigenvalue problems
of size ; a complex approach leads to eigenvalue problems of size m; while the full analysis leads to a single,
but very costly, eigenvalue problem of size .
Although linear cyclic symmetry can reduce the problem size greatly, it is restricted to linear analysis, and the sector
must have its surface mesh on the symmetry planes to be identical on each side of the sector. The nonlinear cyclic
symmetry implemented in Marc can be used for nonlinear problems, such as contact, and the nodes do not need to line
up on both symmetry planes of the sector (for example, see Figure 9-45).
Nastran RBE2 and RBE3
The basic formulation of the Nastran rigid body elements RBE2 and RBE3 is similar to tying type 80, which is also
known as a rigid link. However, they offer additional flexibility because the degrees of freedom may be partly
unconstrained and a list of tied/retained nodes can be entered with RBE2 and RBE3.
RBE2
The nonlinear relation of an RBE2 can be expressed as follows:
(9-55)
where is the coordinate of the tied node, is the coordinate of the reference node, and is the rotation
matrix of the retained node. The linearized relation between the incremental displacement of a tied node to a retained
node and its correction term is expressed as follows:
(9-56)
x
n
u
k
x
n
1
N
---- u
0 2
N
---- u
k c ,
n 1 ( )k cos u
k s ,
n 1 ( ) sin k + [ ]
1 ( )
n 1
N
--------------------- - u
N
2
----
+
k 1 =
K
+ =
2 N
K
N 1
2
-------------- if N i s odd
N 2
2
-------------- if N i s even
=
N 1 ( ) 2
2m N
Nm
x
t
R
r
( ) x
t
x
r
( ) =
x
t
x
r
R
r
( )
u
t
i
Su
r
i
C
non
+ =
653
CHAPTER 9
Boundary Conditions
where the tying matrix S is derived from the linearized rigid body relation. For linear analysis, is zero. For large
displacement analysis, may not match the nonlinear constraint exactly. Therefore, an error vector, , is
required to meet the constraint. It is defined as the difference between the expected coordinates ( ), using the rigid
body kinematics, and the current coordinates ( ) of the tied node.
(9-57)
where is increment number, is iteration number, and zero is the original value. In order to reduce linearization
errors, an optional recycling feature is offered through the CONTROL model and history definition options. Recycling
occurs if the maximum iterative rotation change over the retained nodes of the RBE2 exceeds a rotation tolerance
(default is 0.001 radians). This check can be bypassed by setting the rigid link rotation to zero (in conjunction with the
FEATURE,5701 parameter) or a negative number.
When the rotation is finite, then the coordinate system attached to the tied node will be co-rotated according to the
rotation of the retained node. The degrees of freedom of the tied node will be assigned with this co-rotated system. As
an illustration, consider a slider link shown in Figure 9-47. In this case, a local transformation must be defined for the
tied node in which the x-direction is coincident with the line connecting the tied and the retained node. Then, an RBE2
is defined where y-translational degree of freedom is tied but not the x-translational and any of the rotation degrees of
freedom.
.
Figure 9-47 Co-rotated Coordinate System at the Tied Node
As another example of the use of RBE2 options, consider Figure 9-48 where a cylindrical tube is shown. The tube is
clamped at one end and loaded by a torque at the other end.
If it is assumed that the loaded end remains circular with a constant radius, then a single RBE2 option can be easily
used instead of a number of tyings of type 80. With the RBE2 option, node 1 is defined as the retained or reference
node and nodes 2 to 13 are defined as the tied nodes, with all the degrees of freedom constrained.
However, if the tube is allowed to expand in radial direction, then tying type 80 cannot be used. However, RBE2 option
can still be used, but now the radial displacement is not included in the list of degrees of freedom to be constrained.
C
non
Su
r
i
C
non
x
t e
x
t
x
t e
n i +
x
r
n
R
r
n i +
( ) x
t
0
x
r
0
( ) + =
C
non
x
t e
n i +
x
t
n i +
=
n i
R
B
E
2
R
B
E
2
x
y
x
y
X
Y
(a) Increment i
(b) Increment i + 1

Figure 9-48 RBE2 Example: Cylindrical Tube loaded in Torsion
RBE3
The RBE3 option is a powerful tool to distribute applied loads and massed in a finite element model. Forces and
moments applied to reference or tied node are distributed to a set of independent degrees of freedom based on the
RBE3 geometry and local weight factors.
The way in which the forces are distributed is analogous to the classical bolt pattern analysis. Consider the bolt pattern
shown in Figure 9-49 with a force and moment acting at reference point A. The force and moment can be
transferred directly to the weighted center of gravity along with the moment produced by the force offset .
Figure 9-49 Transfer of Force and Moment on a Reference Point to the Weighted Center of Gravity (CG)
The force is distributed on the bolts proportional to their weighting factors. The moment is distributed as forces
proportional to the product of their distance from the center of gravity with corresponding weighting factors, as shown
in Figure 9-50. The total force acting on the bolts is equal to the sum of the two forces. These results apply to both
in-plane and out-of-plane loadings.
M
z
Z
R
1
2
3
4
5
6
7
8
9
10
11
12
13
Clamped end
F
A
M
A
e
F
A
M
A CG
Reference point A
F
A
M
CG
= M
A
+ F
A
e
e
655
CHAPTER 9
Boundary Conditions
Figure 9-50 RBE3 Force and Moment Distribution
As an example, consider the cantilever plate model with a single quadrilateral shell element shown in Figure 9-51. The
plate is subjected to nonuniform pressure represented by a resultant force acting at a distance of 10 mm from the center
of gravity. The easiest way to apply the pressure is to use an RBE3 option, with a reference node 5, to distribute the
resultant load to each of the four corner points. Setting the weighting factors to be equal for all nodes, results in a force
distribution based exclusively on the spatial location of the connected nodes. In this way, nodes 3 and 4 will get 3N
each and nodes 1 and 2 will get 2N each.
Figure 9-51 Using the RBE3 Option to Represent a Nonuniform Pressure Load
F
i
F
i
r
i
Force Distribution:
F
i
F
A
k
-------------
\ .
|
| |
=
Moment Distribution:
F
i
M
CG
i
r
i
k
r
k
2
-----------------------
\ .
|
| |
=
where F
i
= force at DOF i
i
= weighting factor for DOF
r
i
= radius from the weighted
i
center of gravity to point i
1
2
3
4
100 m
m
1
0
0

m
m
50 mm
40 mm
10N
5

The formulation of the RBE3 is based on the idea that the motion at a reference node is the weighted average of the
so-called error terms of the generalized displacements at a set of connected nodes. The error terms , and are
defined as follows:
(9-58)
where and are the coordinates and rotations of the l-th retained node, respectively; and are the coordinates
and rotations of the reference node, respectively; R is a rotation matrix of the reference node. The error measure for
translational and rotation degrees of freedom are weighted with weighting factors, .
(9-59)
The final nonlinear constraint relation is derived by minimizing the error terms with regard to the reference node
coordinates and rotations using least square fit.
The linearized relation between the reference node and the connected nodes and its correction term can be expressed
as:
(9-60)
with being zero for linear analyses. For large rotation analysis, the error vector, , is required to meet the
constraint. The rotational error, , is assumed to be zero. In other words, the expected incremental rotation of the
reference node is the same with the current value.
The translational error, , is derived by evaluating the difference between the expected coordinate ( ) of the
reference node with the current one ( ). The expected coordinates are calculated by introducing the coordinates of
the connected nodes at the previous increment and the incremental rotation of the reference node into the nonlinear
constraint equations.
(9-61)
x
el

el
x
l
R
t
( ) x
l
x
t
( ) x
el
+ =
l

t

el
+ =
x
l

l
x
t

t
W
l
e
t r an
x
el
T
W
l
x
el
l 1 =
N
=
e
rot

el
T
W
l
el
l 1 =
N
=
u
t
i
Su
r
i
C
non
+ =
C
non
C
non
C
non
r ot
C
non
t ran
x
t e
x
t
x
t e
W
l
l 1 =
N
\ .
|
|
| |
1
W
l
x
l
R
l

t
( ) + ( )
l 1 =
N
=
C
non
t r an
x
t e
x
t
=
657
CHAPTER 9
Boundary Conditions
Beam - Shell Offsets
In many problems, it is convenient and sometimes necessary to model beam and shell elements at a geometric location
that does not match the actual physical location. Such cases are common when flanges or surfaces for shell and beam
elements with varying cross-section properties have to be aligned. Another common instance is when beams are used
as stiffeners for shells. It is most convenient to model the beam elements by sharing the shell nodes at the midsurface
of the shell, as shown in part (a) of Figure 9-52. The fact that the beam is actually offset by half the plate thickness and
half the height of the beam section will have to be achieved by providing a suitable beam offset as shown in part (b)
of Figure 9-52.
Figure 9-52 Beam-stiffened Shell Structure
There are two methods by which beam/shell offsets can be modeled:
The first method is to place the beams and shells at the actual offset position and then tie the nodes of these
elements back to the original position through manually defined RBE2 links. While this method is accurate, it
is difficult to used for large models. Furthermore, it is not possible for the offset elements to contact other
bodies since the nodes of the offset elements are already tied.
The second method involves the beam/shell offset capability in Marc. The elements are specified at the
original position in the model and the offset vectors at the nodes of the element are specified in a variety of
ways through the GEOMETRY option. The current nodal offset is applied to each node of the offset element to
calculate the offset element position.
(9-62)
where is the increment number, is the iteration number, is the current coordinate vector of the
node at the offset position, is the original user-specified coordinate vector of the node at the original
position, and is the current offset vector. For small displacement analysis, the initial user-specified
offset vector, , is maintained through the analysis. For large displacement analysis, the offset vector is
rotated based upon the rotations at the node.
(9-63)
where is the current rotation matrix at the node and is the current rotation vector at the node.
The tying relationship between the offset nodal position and the original nodal position is then established. The
(a) Geometric model with the
beam element sharing the
nodes of the shell elements.
(b) Physical model with the beam
offset by half-shell thickness
+ half-beam height.
x
of f set
n i +
x
0
voff
n i +
+ =
n i x
of f set
n i +
x
0
voff
n i +
voff
0
voff
n i +
R
n i +
( )voff
0
=
R
n i +
( )
n i +

basic formulation for this tying relationship is similar to the RBE2 link described in the previous section.The
relation between the incremental displacement of the node in the offset position to the corresponding
displacement at the original position is expressed as follows:
(9-64)
where the tying matrix is based on a linearized rigid body relation and its terms are derived from offset
vector components. For linear analysis, is zero. For large displacement analysis, may not match
the nonlinear constraint exactly. Therefore, an error vector, , is required to meet the constraint. It is
defined in a manner similar to Equation (9-57).
In order to reduce linearization errors, an optional recycling feature is offered through the CONTROL model
and history definition options. Recycling occurs if the maximum iterative rotation change over the offsets
nodes exceeds a rotation tolerance (default is 0.0001 radians). This check can be by passed by setting the rigid
link rotation tolerance to zero (in conjunction with the FEATURE,5701 parameter) or a negative number.
The element stiffness matrix is formed on the offset element geometry. It is then converted to the original
element geometry through a suitable tying relationship.
(9-65)
All distributed loads on the offset element are also treated in the same way. Note that point load vectors are not
offset and are always calculated at the original nodal positions. The nodal displacements at the original
positions are calculated during the solution phase. Finally, during the recovery phase, the displacements at the
offset position are calculated from Equation (9-64) and element strains, stresses are calculated on the offset
element geometry.
Additional details for beam/shell offsets in Marc are provided below:
Offset Vector Specification
For beams, offset nodal vectors can be specified in four different ways:
1. Offset nodal vector specified in the global coordinate system.
2. Offset nodal vector specified in the local beam coordinate system:
For 2-D beams (element types 5, 45), the x-axis is along the beam axis, the z-axis is (0,0,1) and the y axis is
perpendicular to both.
For 3-D beams, the z-axis is along the beam, the x-axis is user-specified and the y-axis is perpendicular to both.
This option is not available for heat transfer links.
3. Offset nodal vector along the associated shell normal:
This option is particularly useful when beams are used as stiffeners for shells. In the case, only the offset
magnitude is provided by the user. The associated shell normal vector at the node is automatically calculated
and is used for defining the offset vector. This option is not available for 2-D beams.
4. Offset nodal vector specified in the local coordinate system at the node:
The local coordinates at the node can be specified using the TRANSFORMATION, CYLINDRICAL, or similar
options.
u
of f set
i
Su
or i gi nal
i
C
non
+ =
S
C
non
Su
r
i
C
non
K
or i gi nal
S
T
K
of f set
S =
659
CHAPTER 9
Boundary Conditions
For shells, the offset magnitude is provided by the user and the element normal calculated at the centroid of the shell
element is used to define the nodal offset vector.
Supported Elements
Beam/Shell offset support is provided for all standard beam and shell elements with global translational and rotational
degrees of freedom at the corner nodes.
Supported mechanical beam elements include element types 5, 14, 25, 45, 52, 76, 77, 78, 79, 98. Heat transfer links,
element type 36, 65, 99, and 100 are also supported (see section on field analysis below). Non-standard beam elements
that are not supported include element types 13, 16, 51. Supported mechanical shell elements include element types
1, 22, 75, 89, 138, 139, 140. Heat transfer shell elements, element types 50, 85, 86, 87, 88 are also supported (see
section on field analysis below). Non-standard shell elements that are not supported include elements types 4, 8, 15,
24, 49, 72.
For higher order beam/shell elements (element types 45, 22, 89), there is a choice to use interpolated values of the
corner node offset vectors at the midside nodes. If this is not chosen, the offset is taken as 0 at the mid-side nodes. For
beam elements 76, 77, where the midside node degree of freedom is not compatible with the corner node degree of
freedom, only the nodal position is updated for the offset and no other tying is undertaken at the midside node.
Post File
All nodal vectors on the post file are calculated and depicted at the original nodal positions. This allows visual
compatibility when elements with offsets and elements without offsets are in the model. Element quantities like strains
and stresses on the post file are calculated based on the offset element geometries.
Adaptive Meshing
There is no support for global remeshing with beam/shell offsets. Local adaptive meshing is supported. Nodal offsets
are automatically interpolated and transferred over to the new nodes created during the adaptive meshing process.
Field Analysis
Beam/shell offsets are predominantly supported for mechanical analysis. Since offset elements can also be used in
thermo-mechanical analysis and other field analyses, offset support is extended for field analyses other than purely
mechanical analysis. In this case, only the coordinates of the element are adjusted for the offsets, as per Equation
(9-62). No other tying of the field variables or element matrices is undertaken.
Contact
The average nodal offset vector is calculated at every node of all beam/shell contact bodies. All touching nodes and
touched patches are then updated using the average nodal offset vectors.
Dynamics
The stiffness matrix, mass matrix and damping matrix for the offset element are all calculated in a manner similar to
Equation (9-65). Implicit in this calculation is the assumption that the displacements as well as velocities and
accelerations between the original and offset positions are tied in a manner similar to Equation (9-64).

Pin Code for Beam Elements
Pin code is used to remove connections between the node and selected degrees of freedom of the beam. It can only be
applied on the nodes at the ends of the beam. It is not allowed to use it for the middle node of 3-noded beams. The
degrees of freedom are defined in the elements coordinate system (see Marc Volume B: Element Library for more
detail description how the local coordinate system of beam element is set up), and the pin code is applied at the offset
ends of the beam.
Marc manipulates the pin code in two different ways. The first one, by default, is done by internally generating a new
node which is tied to the flagged node. The tied degrees of freedom are the negation of the pinned ones. The tying
process is carried out at element level. The element stiffness matrix and internal forces of the element are calculated
using a standard procedure. During the assembly process they are expanded to include the degrees of freedom of the
new internally generated node. This procedure works fine both for quasi static and dynamics analysis.
The second one, using FEATURE, 6901, is done by using the so-called static condensation technique. In this way,
no internal node is generated. The pinned degrees of freedoms are condensed out from the element stiffness and
internal forces before they are assembled into the global system. This feature is added to allow Marc pin code fully
compatible with Nastran implementation. For quasi-static analysis, both ways give exactly the same behavior. For
dynamics analysis, the coefficients used to condensed out the element stiffness matrix is reused for mass matrix. This
is an approximation to the correct procedure where the static condensation has to be done on the element operator
matrix before being assembled into global system.
The pin code is activated using PIN CODE or the GEOMETRY option. For postprocessing purposes, the element
connectivity is always presented with the original nodes. For example, in a stress-free large displacement analysis for
a beam element with pin code that simulate slider, the beam length will become shorter or longer although the element
is stress free (see Figure 9-53; node 4 is generated internally. If FEATURE, 6901 is activated, node 4 is not
generated). The pin code is defined on element 10 at node 2. The axial-displacement and out-of-plane rotation degrees
of freedom are flagged.
Figure 9-53 Slider Connection using Pin Code
Mesh Independent Connection Methods
In Marc, there are various ways to model connections between structural components in a finite element model. These
connections may represent spot welds, seam welds, bolts, screws, rivets, and so on. Depending on the modeling goals,
they can be modeled using line elements (like trusses and beams), linear or nonlinear springs, rigid body connections
(RBE2, RBE3), and tyings or servo links (MPC constraints). In this way, general connections can be made between
parts, but they often require a lot of modeling effort in defining the connections or in assuring that meshes of the parts
involved are congruent so that they match in those points where connections are desired. Data and input preparation
1
10
2
20
3
1
10
3
20
2
4
661
CHAPTER 9
Boundary Conditions
for these connections can be a formidable task in real-world applications. Often, it is desired to model certain stiffness
characteristics in the connections. An easy way to establish this is by making point-to-point connections between the
parts and using beam elements or springs as the connector elements. Apart from the high meshing effort, this modeling
technique may also produce poor results. Increasing mesh refinement can introduce stiffness errors; point-to-point
connections in which the effective cross-sectional areas are larger than 20% of the characteristic element face areas
can often lead to significant underestimation of connector stiffness. Point-to-point connections can also cause
numerical problems in finite element shell models. The stiffness of the rotations normal to the shell plane may not exist
or may be represented by a penalty stiffness to suppress the associated singularity, but in the point-to-point
connections, such rotations may be excited. If the surface of one of the parts is defined by the faces of continuum
elements, there are no rotational degrees of freedom for the nodes of these elements and point-to-point connections
may introduce undesired singularities.
The CWELD and CFAST connection methods that are available in Nastran have also been implemented in Marc to
address and solve the aforementioned modeling issues. Connections can now be established with ease between points,
elements, patches, or any of their combinations and load transfer between the parts can be established over multiple
nodes on each side of a connection providing a more accurate stiffness representation of the connection. The following
connections between two parts can be made: patch-to-patch, point-to-patch and point-to-point as illustrated in
Figure 9-54.
Figure 9-54 Supported Connections Between Different Parts
Each connection involves the two end nodes of a connector element and two surfaces that are to be connected at a
certain location. The connector element can be any Marc beam element, a truss element or a cbush element. The
surfaces can be defined by shell elements, the faces of continuum elements or by nodal points. These surface element
faces are referred to as patches, can be triangular or quadrilateral, and may be linear or quadratic. On each side of the
connection, either a patch connection or a point connection is made thus giving the three connections shown in
Figure 9-54. A patch-to-patch connection can be shell-to-shell, shell-to-solid, or solid-to-solid. A point-to-patch
connection can be point-to-shell or point-to-solid. A point connection is the simplest connection since the end node of
the connector element is directly connected to a node of the surface; the connector element and the surface share a
common node and no additional constraints are required to establish the connection. A patch connection connects the
end node of a connector element to one or more element faces and is more general as it eliminates the need of
congruent meshes. The location where the end node pierces the surface is automatically determined. The element face
being pierced is called the master patch. Internal constraints are generated to assure a rigid (i.e., a force and moment
carrying) connection of the connector end node to the surface. Two types of patch connections can be made: Direct
Patch Connection which connects the connector end node directly to the nodes of the master patch with tying type 44;
the constraints involved in this tying are also known as Kirchhoff constraints, (see Rigid Tying to a Surface Patch in
Patch-to-patch Connection
For noncongruent meshes,
area wise connection
Point-to-patch Connection
For noncongruent meshes,
point to area wise connection
n
Point-to-point Connection
For nearly congruent meshes,
point wise connection

this chapter) or the Indirect Patch Connection which connects the connector end node indirectly to the nodes of the
surface by constructing a four-node auxiliary patch at the end node normal to the connector. The size of this auxiliary
patch is determined from the diameter or area of the connector. The end node of the connector makes a rigid (tying
type 44) connection to this auxiliary patch. The four nodes of this auxiliary patch (the auxiliary nodes) are connected
to the surface by additional constraints. The location where each auxiliary node pierces the surface is automatically
determined. The patch being pierced is called the secondary patch. By default, RBE3 constraints are generated between
each auxiliary node and the nodes of the secondary patch it pierces, but other constraints may be chosen as well. Note
that, depending on the mesh density, two or more auxiliary nodes may connect to the same secondary patch. All
piercing points are determined by normal projections to the surface to which the connection is made. The two patch
connection types are illustrated in Figure 9-55 which displays a top view of one side of the connection involving
regular four-node quadrilateral element faces. Other element face types and element faces with collapsed edges may
be present as well. All constraints are applied internally as tying constraints and they are automatically generated when
processing the input.
Figure 9-55 Patch Connection Types
The direct patch connection is the simpler type, but may lead to asymmetrical load transfer between the surfaces when
the projections are near the element face boundaries and the element faces are not nicely facing each other. It may also
lead to a very localized load transfer in case of fine meshes, resulting again in a poor stiffness representation of the
connection. The indirect patch connection improves the load transfer by incorporating more element faces, and thus
more nodes, in the connection, but this will also lead to more constraint equations.
In general, a connection between two surfaces can be defined by entering a point that identifies the approximate
location of the connection and the two surfaces that are to be connected. This point can be entered by referencing a
node or by entering its coordinates when defining the connection and it is referred to as the GS-node or GS-point.
These GS reference points determine the spacing of the connections and they can be defined beforehand,
independently of the surface meshes. The surfaces can be identified by entering the element faces to which the
connection will be made or by entering element or face sets from which the element faces to connect to will be sought.
For indirect patch connections, only the master patches need to be entered. The secondary patches are automatically
found in the vicinity of the master patches. If the surfaces are entered as sets, the master patches and the secondary
patches are searched from these sets. The connection between the two surfaces is established as follows. The GS-node
is projected to the first (side A) and second (side B) surface by a normal projection. These two projections determine
the end nodes of the connector and therefore its final position and are referred to as the GA-node for side A and the
GB-node for side B. The nodes on side A and B of the master patches are referred to as GAi and GBi respectively,
master patch
master patch
auxiliary patch
secondary patch
Direct Patch Connection Indirect Patch Connection
663
CHAPTER 9
Boundary Conditions
where i ranges from one to the number of nodes in the patch. The auxiliary nodes are referred to as GAHi and GBHi
respectively, where i always ranges from one to four. The nodes of a secondary patch are referred to as GAij and GBij
respectively, where j ranges from one to the number of nodes in the patch and i refers to the i
th
auxiliary node. The
definitions of the connection are made on the CWELD or CFAST model definition options. The properties of the
connector elements are defined on the PWELD or PFAST model definition options. These properties are material and
geometric properties in case of CWELD connections and stiffness properties in case of CFAST connections. A number
of control parameters can be entered on the SWLDPRM option to enhance the search and projection process and to
control its output.
CWELD patch-to-patch connections
There are several methods to define CWELD patch-to-patch connections. The methods available are ELEMID,
GRIDID, ELPAT, and PARTPAT. In each of these methods, the surfaces on both sides have to be defined together with
the GS reference point of the connection. The ELEMID and GRIDID methods make the simpler direct patch
connections and the ELPAT and PARTPAT methods make the more advanced indirect patch connections. All methods
can be used to connect noncongruent meshes of any type and they are recommended when the cross-sectional area of
the connector is larger than 20% of the characteristic element face area. The connector element can be any Marc beam
element. The default type used is 98 which is a straight beam in space that includes transverse shear effects. By default,
it assumes elastic behavior and a solid circular cross-section. It is always required to include the ELEMENTS parameter
for each element type used as a connector, even when it does not appear in the CONNECTIVITY input. The data for
each connection are defined on the CWELD model definition input. Some data inputs are always required, because no
defaults exist for them. Other data inputs are optional and if not defined, Marc assumes appropriate default values for
them. These optional data inputs concern the connector element type, the way the projections are carried out, the type
of constraints and the orientations of specific CWELD connections.
Direct Patch-to-patch Connections
The ELEMID and GRIDID methods define the connection directly between two master patches. For ELEMID the
patches are entered as shell elements or the faces of continuum elements (SHIDA and SHIDB). For GRIDID, the
patches are entered as nodes (GAi and GBi) following the same connectivity conventions as for corresponding shell
patches. The two methods are illustrated in Figure 9-56. The GS-node is projected to both patches and, in case it is
already positioned between the two patches, these projections determine the connector end node locations. But it is
not a requirement to position the GS-node between the two patches, and if it is not, the projection is done in two steps
as illustrated in Figure 9-57. First, a normal projection to patch A and patch B is made resulting in the points GA and
GB. Then the midpoint GC between these two points is projected to both patches and this determines the connector
end point locations GA and GB. The nodal coordinates of the connector element are updated after this projection. The
GS-node must have a proper projection to the patch within its boundaries. The PROJTOL parameter can be defined to
accept projections outside but near patch boundaries.

Figure 9-56 ELEMID and GRIDID Connection Methods
Figure 9-57 Determination of the Connector End Node Locations GA and GB
The data needed to define a connection with ELEMID are:
1. The reference node GS or its coordinates (XS,YS,ZS).
2. The patches SHIDA and SHIDB on both sides, where for shell elements these are the element numbers and for
continuum elements these are element and face number pairs.
3. A PWID, which is the identification number of a corresponding PWELD property entry.
The data needed to define a connection with GRIDID are:
2. The patch nodes GAi and GBi on both sides. The number of nodes determines the patch type, which can be
quad4, quad8, tria3 or tria6. No mixed patches with partly linear and partly quadratic edges are supported.
SHIDB
GS
SHIDA
ELEMID
GA2
GS
SHIDA
GB2
GA3
GA4
GB3
GB1
GA1
GRIDID
S
h
e
ll A
GA
GA
GC GC
S
h
e
ll B
GB
GB
SHIDB
PIDA
SHIDA
D
L
GA
GB
PIDB
GS GS
665
CHAPTER 9
Boundary Conditions
Indirect Patch-to-patch Connections
The ELPAT and PARTPAT methods make the connection of the connector end nodes to the surfaces involved in two
phases, as illustrated in Figure 9-58 and Figure 9-59. First the reference GS-node is projected to a master patch. For
the ELPAT method, the master patches are specified in the input. For the PARTPAT method, the master patches are
searched from sets that have been specified in the input. Around each projection point, a square four-node auxiliary
patch is constructed. This auxiliary patch is normal to the line connecting the two projection points. The dimensions
of this auxiliary patch follow from the diameter or the area of the connector as displayed in the right part of
Figure 9-58. For the four auxiliary nodes, new projections are computed on the master patches or on patches in their
vicinity. The patches to which these new projections are found are called secondary patches and each auxiliary node
is connected by default with RBE3 constraints to the nodes of its respective secondary patch.
Figure 9-58 ELPAT or PARTPAT Connections
The master patches and the secondary patches do not have to be distinct as can be seen from Figure 9-58 and
Figure 9-59. For a coarse mesh, they may even be one and the same patch but for fine meshes they may all be distinct.
The requirement is that the secondary patches do not span a larger area of the mesh than a 3 x 3 area as shown in the
right part of Figure 9-59 otherwise a number of nodes not involved in the constraints defining the connection is present
under the area covered by the auxiliary patch. If the NREDIA parameter is set and the secondary patches violate the 3
x 3 requirement, the auxiliary patch size is reduced a number of times in an attempt to fulfill this requirement.
Shell A
Shell B
GAH2
GAH3
GAH4
GA
GAH1
GB
GBH1
GBH2
GBH3
GBH4
GBi
GAi
GAH2 GAH1
GAH4
GAH3
GA
a
a
a a
D/2 D/2
a
D
4
---- =

Figure 9-59 Connections in Coarse and Fine Meshes for the ELPAT and PARTPAT Methods
The main purpose of the ELPAT and PARTPAT methods is to improve the load transfer between the two surfaces.
These methods allow for a more symmetrical load transfer than the simpler ELEMID or GRIDID methods as
illustrated in Figure 9-60. On side A in the left part of the figure, the shaded patch is the master patch and the
constraints are only applied between the connector node and the nodes of this patch. On side A in the right part of the
figure the neighboring patches also become involved in the connection and the load at the connector node is transferred
through the four auxiliary nodes to all the nodes of the secondary patches in the surface, leading to a better load
transfer. It also avoids a too localized load transfer as may be clear from the right part in Figure 9-59. If this connection
were of the ELEMID type, the only patch involved in the load transfer would be the one entirely covered by the shaded
area leading to a much more localized load transfer.
The PARTPAT method defines a connection between two sets. In case the part is meshed with shell elements, the set
can be an element set identifying the elements that make up the surface. In case the part is meshed with continuum
elements, the set must be a face set identifying the faces that make up the surface. A shell surface can also be identified
by a face set, but the face information is not essential and therefore element sets can be used as well.
The ELPAT method is similar to the PARTPAT method. The only difference is that the ELPAT method defines the
master patches, whereas the PARTPAT method searches them from a set of elements. A further difference is that for
shell meshes the ELPAT method searches its secondary patches from the whole mesh, whereas the PARTPAT method
searches them from the same sets from which the master patches were found, but sets may be used to limit the search
regions. Sets are only required for the ELPAT method if the surfaces contain faces of continuum elements.
GAH1
GA
GAH2
GAH3 GAH4
GAij GAij
GAH2
GAH1
GAH3
GAH4
Coarse Mesh Fine Mesh
667
CHAPTER 9
Boundary Conditions
Figure 9-60 Comparison of Load Transfer in the ELEMID and the ELPAT Methods
The data needed to define a connection with PARTPAT are:
2. The sets SetA and SetB for both sides, where, for shell elements, these sets may be element sets, and for
continuum elements, they must be face sets containing element and face number pairs. The two sets SetA and
SetB do not have to be disjoint, but the search procedure is facilitated if they are.
The data needed to define a connection with ELPAT are:
2. The patches SHIDA and SHIDB on both sides, where for shell elements these are the element numbers and for
continuum elements these are element and face number pairs.
3. The sets SetA and SetB for both sides, where, for shell elements, these sets may be element sets and for
continuum elements they must be face sets containing element and face number pairs. The two sets SetA and
SetB do not have to be disjoint. For surfaces defined by shell elements, these sets are optional, but for surfaces
defined by faces of continuum elements, they must be specified.
Parameters for the Projection Process
When searching master or secondary patches, it may happen that no projection is found at all to any of the candidates
because the surface is not flat and the GS-node or an auxiliary node only has a projection outside two neighboring
patches as illustrated in the left part of Figure 9-61 or because one or more of the auxiliary nodes lie outside a free
edge as illustrated in the right part of Figure 9-61.
SHIDA
SHIDB
SHIDB
SHIDA
Connected Elements for ELEMID Format Connected Elements for ELPAT Format

Figure 9-61 Exceptional Cases needing Special Consideration
In the first case, a projection is accepted to the common edge of the two patches and the nearest patch is selected as
the master or secondary patch. In the second case, two parameters may be used to improve the situation. The GSMOVE
parameter allows the GS-node to be moved in the direction of the auxiliary nodes that have found successful
projections. The NREDIA parameter allows the area of the auxiliary patch to be reduced without moving the GS-node.
Both parameters may be used simultaneously. In that case, it is first tried to move the GS-node a number of times and
if that is not successful the area is reduced a number of times. If that still fails, it is once more tried to move the GS-
node a number of times, now using the reduced area for the auxiliary patch. Successful projections of the auxiliary
nodes are counted in pairs. A successful pair means that the two corresponding auxiliary nodes on both sides of the
connection have found a successful projection. So in the situation in the right part of Figure 9-61, two pairs of auxiliary
nodes are successful and two pairs have failed, because their projections fall over a free edge of one or both of the
surfaces. In addition, the PROJTOL parameter may be used to accept projections outside but near patch boundaries. If
GSMOVE is nonzero, the new GS location is determined from the successfully projected GAHi and GBHi pairs as
follows:
If one pair of GAHi and GBHi finds a successful projection, GS is moved to the middle position between that
pair and the original GS.
If two pairs of GAHi and GBHi find a successful projection, GS is moved to the middle position between those
two pairs and the original GS.
If three pairs of GAHi and GBHi find a successful projection, GS is moved closer to the second pair found.
The move is one quarter of the distance from GS to the second pair.
If no pairs are found at all, moving GS is abandoned.
If the number of moves has been exhausted and NREDIA is nonzero, the characteristic diameter of the CWELD
connector is halved a number of times reducing the auxiliary patch size, thus bringing its nodes closer to each other.
If one or more pairs still fail after this diameter reduction, a last attempt is made to move the GS node while
maintaining the reduced sized of the auxiliary patch. The size of this movement is twice that of the movement before
the size reduction which, in general, is still a small movement, because of the reduced auxiliary patch size. This second
attempt to move the GS-node is only carried out when NREDIA is nonzero.
In some occasions, it may happen that a proper projection is found far away from the GS point (e.g., when the surface
is curved and the situation displayed in the left part of Figure 9-61 occurs for patches near to the GS point. In that case,
the GSTOL parameter may be used to reject projections that are further away than specified by the parameter, so
projections will only be accepted that are relatively near to the GS point.
GS Free Edge
669
CHAPTER 9
Boundary Conditions
It is assumed that the two surfaces that are being connected are reasonably parallel. The GSPROJ parameter may be
used to check this explicitly as illustrated in Figure 9-62. If the master patch normal vectors on both sides make an
angle larger than specified by the parameter, the projection is rejected.
It is also assumed that master patches and associated secondary patches are reasonably parallel. It is not desirable that
elements almost normal to the master patch are selected as secondary patches. The GSCURV parameter may be used to
reject secondary patches that make a large angle with the master patches as illustrated in Figure 9-62. This may occur
in sharp corners of the surface which in practice may represent edges of the structure.
The parameters discussed in this section are defined on the SWLDPRM option.
Figure 9-62 Angle Parameter GSPROJ and GSCURV to Control the Patch Selection Process
Internal Constraints to Connect Two Surfaces
Each connector end node, that is not making a point connection, makes a rigid connection to its master or auxiliary
patch by means of a Kirchhoff constraint (tying type 44). This constraint rigidly constrains the displacements and
rotations of the end node to the motion of the patch it connects to. The possible patches to which such a connection
can be made are linear and quadratic quadrilateral and linear and quadratic triangular patches. With the ELPAT and
PARTPAT methods, the auxiliary patch nodes are connected to their partnering secondary patches by additional
constraints. By default, these will be RBE3 constraints, but alternatively Kirchhoff constraints may be used or a
combination of RBE2 and RBE3 constraints. The RBE2/RBE3 combination makes a RBE2 constraint between the
auxiliary patch node and its projection point on the secondary patch and RBE3 constraints between this projection
point and the nodes of this patch.
When a surface is curved, the auxiliary patch nodes will in general not lie on the surface. A special parameter may be
set in the CWELD input to force a repositioning of these nodes onto the surface. This slightly alters the orientation of
the auxiliary patch and in some cases this might allow the connector to follow the curvature of the surface somewhat
better. By default, this repositioning is not done.
For the RBE3 constraints involved in the CWELD connection, a weighting other than uniform weighting can be used
by specifying a distance weighting exponent. The weight factor for each retained node in the RBE3 is:
GSPROJ
n
A
n
B
GAH1
GA
n
A
GAH2
GSCURV
GSCURV
n
AH1
f
i
1
d
i
n
------ =

where is the weighting factor for retained node i, is the distance from the tied node to retained node i and n is
the weighting exponent. Negative values for n are not recommended, since they will result in heavier weighting for
nodes further away. The default results in uniform weighting .
The RBE2/RBE3 combination is only useful when the auxiliary nodes are not lying on the secondary patches and when
a nonuniform distance weighting is used. It also comes with a cost penalty because it involves more constraint
equations in each connection.
If the CONNECTIVITY input contains no elements with nodes having displacements and rotations as degrees of
freedom, it is necessary to set the RBE parameter.
Multiple Surface Connections
The patch-to-patch connection can be used to connect more than two layers of shell elements by using the same GS-
node in multiple CWELD definitions. For example, if three layers need to be connected, a second CWELD definition
is made, that refers to the same GS-node as the first CWELD definition. Surface B on the first CWELD definition is
repeated as surface A in the second CWELD definition. If the surfaces are not curved, the locations of the connector
end nodes connecting from the top and bottom side to the surface in the middle automatically coincide, but for curved
surfaces, this is not guaranteed. For all connection types except PARTPAT, there is no real difference between
repeating the GS-node or defining it a multiple number of times.
When the PARTPAT method is used and a set pair is repeated in a number of connections that use the same GS-node
(e.g., the same set is used for side A and side B for each of these connections), this repeated usage of the GS-node
results in searching through a stack of patches that can be found from the two sets, in order to connect multiple plate
surfaces. However, when the set pairs are different, i.e., at least one side in a new connection refers to a set with a name
different from the name referred to by this side in a previous connection using the same GS-node, the effect is as in all
other connection types.
The definition for a search through a stack of surfaces can only be made by repeated usage of a GS-node (i.e., a
repetition of its node ID, not just a repetition of its coordinates XS, YS, and ZS) in combination with the same sets for
side A and side B for all the connections that must be present in the stack. It only applies to the PARTPAT connection
type and in that case the sets necessarily have an overlap, i.e., they are not disjoint. The stack is limited to 10
connections, i.e., 11 plate surfaces.
Coupled Thermo-mechanical Analysis
Cweld connections are be used in a coupled thermo-mechanical analysis. In the heat transfer pass of an increment, the
Kirchhoff constraints will tie the temperature of the tied node to the temperature of the projection point on its
corresponding patch. For RBE3 constraints, the temperature of the tied node will be the least squared weighted average
of the temperatures of its retained nodes. The same weighting factors apply as in the mechanical pass. If the retained
nodes are the nodes of a shell element, all temperature degrees of freedom of the shell node are used in the weighting
process. For RBE2 constraints, the temperature of the tied node is set equal to the temperature of its corresponding
retained node, so the constraint in the heat pass is identical to typing type 1.
f
i
d
i
f
i
1 = ( )
671
CHAPTER 9
Boundary Conditions
Rules for usage of the GS-node or the GA-, GB-node Pair in the
Projection Process
In most cases, the GS-node will be used to find the projections determining the end nodes of the connector element.
The GS-node and its XS,YS,ZS-coordinates are ignored on a CWELD definition if both the GA and the GB end nodes
of the connector are specified. In that case, GA and GB should, at least, be approximately in the right locations and
GA is projected to side A and GB is projected to side B of the connection. If GS and GA are not specified, but GB is,
then GB is used as GS and if GS and GB are not specified, but GA is, then GA is used as GS. If GS, GA, and GB are
not specified, then the XS,YS,ZS coordinates must be specified. If GS is specified and maximally one of the GA or
GB, then GS is used for the projections on both sides and the GA or GB is ignored for this purpose. If one side of the
connector makes a point connection and the connector node for that side has been left blank and the GS-node has been
specified, the point connection will be made to the GS-node and the GS-node will be projected onto the opposite side
to determine the location of the other connector node. If the opposite side also makes a point connection, the node for
that side must have been specified.
CWELD Point-to-patch Connections
It is possible to connect one end of the connector element to the surface by a point connection as is shown in
Figure 9-63. This may even be a free node to which certain boundary conditions may be applied. In such cases the
surface information that identifies the master patch for the particular side is omitted. This means that for the ELEMID
and ELPAT method, the input for SHIDA or SHIDB is omitted, for the GRIDID method, the nodes GAi or GBi are
omitted, and for the PARTPAT method, the set SetA or SetB is omitted. The GS-node is projected to the surface on
the other side according to the specifications made in the input.
Figure 9-63 Point-to-Patch Connection on Side A using GRIDID
GA2
GA1
GA
GA4
GA3
GS
n

CWELD Point-to-point Connections
All methods may be used to define point-to-point connections by omitting the surface information that identifies the
master patches and by specifying the connector end nodes on the CWELD input. These must be existing nodes in the
structure, because they can no longer be generated internally. Using any of the patch connection methods would lead
to cumbersome definitions of such point-to-point connections and the ALIGN method was designed to simplify their
definition. It is shown in Figure 9-64. The GS-node is no longer required, because the connector end nodes in general
will be nodes belonging to the two surfaces. Point-to-point connections are only recommended for nearly congruent
mesh, so the connector element will be approximately normal to the surfaces. However, they are not recommended
when the connector cross sectional area exceeds 20% of the characteristic element face area.
Figure 9-64 Point-to-point Connection using ALIGN
CWELD Properties
PWELD Property Definition
The properties for the connector element can be entered on the PWELD model definition option. The material
properties are defined by entering the material identification number of a valid material for the element. The default
cross section is a solid circular section for which the diameter can be entered on the PWELD input. If other geometric
properties than those of a circular section are required, a complete set of geometric input data can be entered here as
well. In that case, the diameter is only used to determine the size of the auxiliary patches. If no diameter is entered but
a set of geometric data is, an attempt is made to estimate the auxiliary patch size from the geometric data. The
geometric data input also specifies the local directions of the cross section. If their corresponding data values are all
zero, the local directions are determined by the procedure outlined in the Connector Orientation section in this chapter,
otherwise the local directions are determined from the geometric data input according to the usual Marc convention.
The default length of the connector is given by the distance between the two end points GA and GB. Two control
values on this length can be entered to prevent overly stiff or overly flexible behavior of the connector element.
Additionally, a spot weld connection type can be selected as an alternative to the default general weld connection type.
In both cases, the effective stiffness of the element is changed by making certain length modifications. Their details
are discussed later in the Connector Length Modifications and Spot Connections section.
GA
GB
n
n
Upper Shell Midsurface
Lower Shell Midsurface
673
CHAPTER 9
Boundary Conditions
Connector Orientation
The connector element is positioned between the two projections GA and GB and these points determine the end nodes
of the element and thus its direction. The section forces for the default beam type are displayed in Figure 9-65 and the
local section directions are determined as follows. The element z-axis points from GA to GB.
The first procedure outlined in this section is used when the 4th, 5th, and 6th fields of the appropriate data blocks in
the PWELD or GEOMETRY options are all zero and no coordinate system (MCID) is specified in the CWELD option.
This procedure is as follows:
is the local z-axis
In case of zero length, the average normal of the two surfaces is taken. Find the index j of smallest component of .
In case of two equal components, the smaller index i is taken. Define the corresponding base vector as follows:
, where when and 0 otherwise, e.g. for ,
Figure 9-65 Connector Orientation, Local Cross-section Directions and Forces
e
3
x
B
x
A
x
B
x
A
------------------------- =
e
3
e
3
j
mi n e
3
i
{ } = i 1 2 3 , , =
b
j
1j
2j
3j
=
kj
1 = k j = j 3 = b
3
0
0
1
=
m
xB
f
x
y
e
plane 1
plane 2
x
e
f
z
GA
f
y
m
z
m
yA
f
x
m
xA
GB
z
e
f
y
m
yB
f
z
m
z
f
z
f
x
f
y
m
z
m
xA
m
xB
m
yB
m
yA
axial force
shear force plane 1
shear force plane 2
torque
bending moment end A, plane 2
bending moment end B, plane 2
bending moment end A, plane 1
bending moment end B, plane 1

It provides a good directional choice for . In addition, the vector must be orthogonal to .
and is the local x-axis
The cross product of local z and local x provides the local y-axis
The second procedure outlined in this section is used when the 4th, 5th, and 6th fields of the appropriate data blocks
in the PWELD or GEOMETRY options are all zero, but a coordinate system (MCID) is specified in the CWELD option.
In that case the local directions are determined by a procedure that uses a coordinate system defined in the COORD
SYSTEM option. This procedure is as follows:
is again the local z-axis.
In case of zero length, the average normal of the two surfaces is taken. The 2nd direction (T2) defined by MCID is used
to define the orientation vector of the connector element.
The local y-axis is then defined as
The cross product of local y and local z provides the local x-axis
If any of the 4th, 5th, or 6th fields of the appropriate data blocks in the PWELD or GEOMETRY options is nonzero, the
conventional Marc procedure is used to determine the local element directions, irrespective of the presence of a MCID
input in the CWELD option. A coordinate system may be associated with the first beam node (GA), but offsets are not
possible and are ignored.
An angle may be specified for each connector element, which rotates the local axes about the line connecting the
two end points. This rotation takes effect after the orientation has been found by either one of the procedures outlined
above or by the geometric specifications made in the PWELD or GEOMETRY options.
Connector Length Modifications and Spot Connections
In the default general connection, the distance between the points GA and GB determines the length of the connector
element. To prevent the connector element from overly stiff or overly flexible behavior two tolerance values, LDMIN
and LDMAX can be entered on the PWELD input. LDMIN and LDMAX are the minimum and maximum length to
diameter ratio of the connection. If the length to diameter ratio is less than LDMIN, the behavior of the connection may
become overly stiff and if it is more than LDMAX, it may become overly flexible. In either case, a length modification
of the CWELD is made such that its behavior remains within the two limits. There are three methods available to
account for this length modification.
e
1
e
1
e
3
e
1
b
j
e
3
b
j
( )e
3
= e
1
e
1
e
1
----------- =
e
2
e
3
e
1
=
e
3
x
B
x
A
x
B
x
A
------------------------- =
v
e
2
e
3
v
e
3
v
--------------------- =
e
1
e
2
e
3
=
675
CHAPTER 9
Boundary Conditions
The first method is stiffness scaling. For linear elastic material behavior, the length modification could be accounted
for by scaling the Youngs modulus such that the axial, shear, and torsional stiffness of the beam with its actual length
(i.e., the length determined by the distance between its end nodes) is identical to the axial, shear, and torsional stiffness
of a beam with the modified length using the original Youngs modulus. This scaling can only be done when the beam
material behavior is linear elastic and the cross-section properties are not entered through the BEAM SECT parameter.
Internally, the cross-section geometric properties of the beam are scaled and not the Youngs modulus.
The second method is repositioning of the end nodes of the beam. Each beam node is moved in a direction normal to
its patch such that the beam obtains the desired length. The distance to the patch is treated as an offset in the Kirchhoff
constraints. If there is a point connection on one side, then that node is not repositioned.
The third method is repositioning of the end nodes and the auxiliary nodes. For methods ELPAT and PARTPAT, the
auxiliary nodes can be repositioned together with the beam nodes such that the beam obtains the desired length. This
requires updating the projections of the auxiliary nodes which may fail if these nodes become located far away from
their respective surfaces and these surfaces are curved. In that case, they may no longer have a projection inside any
of the element face boundaries.
If the element and its material behavior allow it, the default method is the first method, stiffness scaling, otherwise, no
length modification is made. The length modification method can be activated by defining the CWSPOT parameter on
the SWLDPRM option. If a CWELD has a point connection on both sides (like in ALIGN) or the distance between the
two end nodes of the beam is zero, no length modification will be made.
Alternatively, a spot weld connection can be defined. In this case, the length of the connecting beam is modified to be
where is the patch shell thickness on side A and is the patch shell thickness on side B. If the
surface on one side is defined by the faces of a solid element, its thickness contribution is zero. If both sides are faces
of solid elements, the connection type is automatically reset to a general connection.
The type of connection, general connection or spot connection, can be selected on the PWELD input. Spot connections
are only possible when one of the methods ELEMID, ELPAT, or PARTPAT is used.
If the distance between the connector end nodes GA and GB is zero, the two nodes will be connected rigidly using
tying type 100 and no beam element is activate in the connection.
If stiffness scaling is requested as the length modification method and the material is nonlinear or the cross section
properties are entered through the BEAM SECT parameter, the scaling is not done and a warning message is written to
the output.
CWELD Input Styles
For CWELD connections, two different input styles can be used to define the connections and these styles may be
mixed in one model.
The first style is the most natural style, where only the GS-node and the two surfaces are defined in the CWELD input.
This style was used in the foregoing discussion. The beam connector elements and its nodes are generated internally
and are not present in the CONNECTIVITY and COORDINATES inputs. All necessary geometric properties of the
connection are defined in the PWELD input and its material properties are defined by entering a valid material
identification number in the PWELD input.
With the second style, the connector elements are entered explicitly in the CONNECTIVITY option and their nodes in
the COORDINATES option. The connector nodes do not have to be in the right location as long as the GS-node is in
0.5
t
A
t
B
+ ( ) t
A
t
B

the approximate location. The material properties of the connectors may then be entered through the usual material
options and their cross-section properties through the GEOMETRY option. All these properties are assigned to the
elements directly. The CWELD option still defines the connection, but only refers to the connector elements and a
PWELD input is generally no longer required. The second style requires that the CWELD input is made before the
CONNECTIVITY input of the element.
However, it is not required to rigidly adhere to the input styles mentioned above. A connector element with its ID,
type, and nodes may be defined entirely on the CWELD option. If no element ID is specified, Marc generates it
internally. If no node IDs are specified, Marc generates them internally. It no type is specified, type 98 is chosen for a
3-D analysis and type 5 for a 2-D analysis. If the CONNECTIVITY for an element is entered after its CWELD input, the
CONNECTIVITY input determines the element type and the nodes of the element. If the CWELD input for an element
is entered after its CONNECTIVITY input, the CWELD input determines the element type and nodes of the element. If
the CWELD input specifies the element ID, a CONNECTIVITY input for the element is not required. It is optional and
can be used to redefine the nodes if these were left undefined in the CWELD input.
If the connector element is only defined through the CWELD input, it is necessary to enter an ELEMENTS parameter
for the type of the connector element as it is necessary to enter an ELEMENTS parameter for each element type that is
used in the CONNECTIVITY input.
If a side of a connection makes a point connection, the node to which this connection is made cannot be an
automatically generated node. It must be a node defined by the user in the input. The user has no control over
automatically generated nodes; i.e., how their IDs are assigned to them. Therefore, it is not possible to load or
constrain such nodes or to connect them to other parts of the model.
CWELD Error and Warning Messages
A number of error and warning messages can be written to the output file for any of the connector end nodes or the
auxiliary nodes. Error messages obtain a positive code and terminate the job with exit number 13. Warning messages
obtain a negative code and allow the job to continue. The possible messages are listed in Table 9-27.
Table 9-27 Codes for CWELD Error and Warning Messages
Value Description
1 Projection algorithm did not converge within desired tolerance.
2 Projection converged, but not inside the patch boundary.
-2 Projection converged, but near the patch boundary (inside: 0.01, outside: PROJTOL).
3 The patch is severely distorted or inside out (normal vectors in different locations make larger than
90 angles).
-3 The patch is OK, but still has much distortion (normals in different locations make larger than 10
angles).
4 The angle of an element/face normal to its auxiliary or master patch normal is too large; the patches
are almost normal to each other.
-4 The angle of an element/face normal to its auxiliary or master patch normal is rather large.
5 GS projects OK, but is located too far away from the projection.
677
CHAPTER 9
Boundary Conditions
-5 GS or an auxiliary node projects OK, but is located far away from the projection.
For GS nodes we use GSTOL as the tolerance.
For auxiliary nodes, we use XDIM (characteristic cweld dimension) as the tolerance.
, where D is characteristic cweld diameter.
6 The element we want to project to is unknown (i.e., useless, because element type = 0).
-6 A projecting is being made to a deactivated element, but accepted anyway.
7 The beam is almost parallel to the patch. (its angle with the patch is less than GSPROJ)
-7 The beam is not very normal to the patch (it makes an angle larger than 2*GSPROJ to the patch
normal).
8 The patches to which the auxiliary nodes tie are more than 3 x 3.
9 The patch nodes have invalid degrees of freedom.
10 The patch has an invalid number of retained nodes.
11 A search over a group of patches resulted in no success at all; no projection has been even attempted
to any of the patches.
12 A search over a group of patches resulted in no success at all; although projections were tried, they
were all unsuccessful.
13 The projection is outside the patch boundary and there are no neighbors; so, there is no positioning to
a common edge or a common corner.
-14/14 Warning/Error: Position the cweld node to corner 1 of patch.
-18/18 Warning/Error: Position the cweld node to edge 1-2 of patch.
-20/20 Warning/Error: Position the cweld node to edge 3-4 (quad) or 3-1 (tria) of patch.
22 No face id was specified for the continuum element.
24 No viable patches in the set (PARTPAT, ELPAT).
26 No viable elements in the whole model (PARTPAT, ELPAT).
-27 To account for a length modification, the beam node on this side of the patch has been offset with
regard to the patch it is connecting with.
28 The beam node is shared with other elements, while not making a point connection.
Or the GS-node is shared with elements on the side that makes a point connection.
Table 9-27 Codes for CWELD Error and Warning Messages (continued)
Value Description
XDIM 1 4D =

CWELD Output
A CWELD connection defines a beam element as the connector element and generates internal tyings to define the
connector constraints. Output to the .t16 or .t19 post file can be requested for the beam elements by using the standard
element post codes available for them. Tying forces can be requested by using the standard nodal post codes available
for them. The displacements of all nodes involved in the connection will by default be on the post file.
Four sets are automatically generated and available on the post file when the CWSETS parameter on the SWLDPRM
option is set to one. The first set contains all the beam elements used in the connections. The second and third sets
contain all patches defining side A and side B of all connections, respectively. The fourth set contains all elements that
have caused warnings. An error set is not available because, upon encountering an error, no attempt is made to
complete a connection. This means that, in case of an error, the information at best is incomplete, but because of its
intermediate character it may often be misleading and often it is not yet available at all.
The sets are called:
Set 1: fastener_all_beams_inc0000
Set 2: fastener_all_faces_sidea_inc0000
Set 3: fastener_all_faces_sideb_inc0000
Set 4: fastener_all_warnings_inc0000
Set definitions defining sets with any of these names should not be made in the input.
The PRTSW parameter in the SWLDPRM input can be used to control the amount of output of each CWELD connection.
If PRTSW = 0 or is left undefined, no diagnostic messages are printed to the output file. If PRTSW = 1, only errors
are printed. If PRTSW = 2, errors and warnings are printed. if PRTSW = 3, all projection diagnostics are printed for
each CWELD connection. If PRTSW = 4, all projection diagnostics and all specifics of the tying constraints involved
in each connection are printed.
Two Dimensional Models
The previous sections described the CWELD connections for three dimensional models. However, CWELD
connections can also be made in two dimensional models. In that case, the patches consist of beams, trusses, or edges
-28 The GS-node is shared with other elements.
29 The face ID identifying a patch is missing for a continuum element.
30 The connector node cannot connect to this element because it is not a valid connection type.
32 In 3-D, a connection to a line type of element was attempted; this is not possible.
-33/33 The connector node made a point connection but is not connected to any other element than the
connector element. If this node is an automatically generated node, the result is an error; otherwise, a
warning.
34 The projection is to an element/face sharing the point connection node.
35 An element/face of another connection lies between the current element/face pair.
Table 9-27 Codes for CWELD Error and Warning Messages (continued)
Value Description
679
CHAPTER 9
Boundary Conditions
of continuum elements. The same connection methods that are available for three dimensional models can be used in
two dimensional models. The connector element must be a two dimensional beam or an axisymmetric shell depending
on the type of two dimensional analysis. The number of auxiliary nodes is two instead of four; one to the left and one
to right of the projection point on the master patch. Two patch types are available: linear (with two nodes) and
quadratic (with three nodes) patches. The 3 x 3 requirement only applies in one direction now, so a connection on one
side can not span more than three patches of the surface.
CFAST Connections
CFAST connections can only be defined using indirect patch-to-patch connections. Two methods, ELEM or PROP, are
available to create them. They are equivalent with the CWELD ELPAT and PARTPAT methods, respectively (see Figures
9-57 and 9-58). In each of these methods, the surfaces on both sides have to be defined together with the GS reference
point of the connection. Both methods can be used to connect noncongruent meshes of any type. The constraints
between the connector nodes and the auxiliary patches are applied through Kirchhoff constraints (typing type 44) and
the constraints between the auxiliary nodes and the secondary patches are applied through RBE3 constraints.
The connector element is a cbush element. In a two dimensional model, it is element type 194 and in a three
dimensional model, it is element type 195. For more details about the cbush elements, refer to Marc Volume B:
Element Library. It is always required to include the ELEMENTS parameter for each element type used as a connector
even when it does not appear in the CONNECTIVITY input. The data for each connection is defined on the CFAST
model definition input. Some data inputs are always required because no defaults exist for them. Other data inputs are
optional and if not defined, Marc assumes appropriate defaults values for them. These optional data inputs concern the
connector element type, the way the projections are carried out, the type of constraints, and the orientations of the
specific CFAST connections. For a more detailed description of the CFAST option, refer to Marc Volume C: Program
Input.
CFAST Properties
PFAST Property Definition
The properties of the CFAST connection are defined using the PFAST model definition option. It contains the diameter
of the fastener, the stiffness matrix, damping coefficient, and mass. The diameter data is only used to compute the
locations of the auxiliary nodes involved in the CFAST connection. It will not affect the element stiffness as the latter
has to be explicitly entered. For a more detailed description of the PFAST option, refer to Marc Volume C: Program
Input.
CFAST Element Coordinate System
The stiffness and damping values are given is a specified coordinate system . This coordinate system is
used as the element coordinate system. The output of CFAST is given in this system. There are two options to define it:
1. Using the connector element orientation. The x-axis points from GA to GB.
e
1
e
2
e
3
, , ( )
e
1
x
B
x
A
x
B
x
A
------------------------- =

In case of zero length, the normal of the first patch is taken. Find the index of j of smallest component of
In case of two equal components, the smaller index i is taken. Define an auxiliary vector:
, where when and 0 otherwise (for example, for , ) is
created by the normalized cross product of and . And, finally, is the normalized cross product of
and . For geometrically nonlinear analysis, this coordinate system co-rotates with the element.
2. Using a global or user-defined coordinate system . For this option, you may choose between a
relative and an absolute orientation.
a. the absolute orientation:
For a geometrically non-linear analysis, this coordinate system is fixed in time.
b. The relative orientation: the x-axis points from GA to GB.
In case of zero length, the normal of the first patch is taken. is defined by the normalized cross product
of and . is defined by the normalized cross product of and . For a geometrically nonlinear
analysis, this coordinate system co-rotates with the element.
SWLDPRM, Special CWELD/CFAST Parameters
A number of global parameters that control the behavior of CWELD and CFAST connections and their output to the
jobid.out file can be entered through the SWLDPRM model definition option. These parameters and their descriptions
are summarized in Table 7-28.
Table 7-28 SWLDPRM Parameter Names and Descriptions
Name Type Default Description
CHKRUN Integer > 0
(0 or 1)
0 This parameter is available in MD Nastran but has no meaning in
Marc and is ignored.
GSMOVE Integer > 0 0 Maximum number of times GS is moved in case a complete projection
of all points has not been found.
e
1
e
1
j
mi n e
1
i
{ } =
i 1 2 3 , , =
b
1j
2j
3j
=
kj
1 = k j = j 3 = b
0
0
1
= e
3
e
1
b e
2
e
3
e
1
e
1
u
e
2
u
e
3
u
, , ( )
e
1
e
2
e
3
, , ( ) e
1
u
e
2
u
e
3
u
, , ( ) =
e
1
x
B
x
A
x
B
x
A
------------------------- =
e
3
e
1
e
2
u
e
2
e
3
e
1
681
CHAPTER 9
Boundary Conditions
NREDIA 0 < Integer < 4 0 Maximum number of times the characteristic diameter D is reduced in
half in case a complete projection of all points has not been found.
PRTSW 0 < Integer < 4 0 Parameter to control the CWELD/CFAST diagnostic output to the
Marc output file (jobid.out).
0 = no diagnostic output
1 = print errors only
2 = print errors and warnings only
3 = print projection diagnostics with no tying details
4 = print all diagnostics
GSPROJ -90 < Real < 90 20.0 Maximum angle allowed between the normal vectors of master patch
A and master patch B. The connection will not be generated if the
angle between these two normal vectors is greater than the value of
GSPROJ. If GSPROJ is negative, the program will always accept the
connection and will only issue a warning if the angle is larger than
|GSPROJ| (see Figure 9-62).
GSCURV -90 < Real < 90 20.0 Maximum angle allowed between the normal vectors of a patch to
which an auxiliary node projects and its corresponding auxiliary and
master patches. It provides a measure to monitor the curvature of a
surface and to recognize patches that belong to, for example,
stiffeners. A connection is not generated if the angle between the
normal vectors is greater than 90-GSCURV meaning that the patches
are almost normal to each other. In that case, the patch is rejected and
the search proceeds to the next patch in the list. If the angle is between
zero and GSCURV, no message is displayed. If the angle is between
GSCURV and 90-GSCURV, a large angle warning is displayed. The
following three tests are performed in the order given below when
GSCURV is positive:
If 0 < angle < GSCURV => OK
If GSCURVE < angle < 90-GSCURV => trigger a warning.
If angle > 90-GSCURV => reject.
Note that the warning condition is never triggered when GSCURV > 45
as it is overruled by the reject condition.
If GSCURV is negative, the projection is always accepted and a warning
is issued when the angle is larger than |GSCURV| (see Figure 9-62)
Table 7-28 SWLDPRM Parameter Names and Descriptions (continued)

GSTOL Real 0.0 Maximum allowable distance of the node GS to its projection on a
patch. IF GSTOL is positive, the distance is relative to the characteristic
CWELD/CFAST diameter D, (the tolerance is GSTOL*D). If GSTOL is
negative, the distance is absolute (i.e., the tolerance is -GSTOL). If GS
is used for the projection together with one of the methods
PARTPAT/PROP or ELPAT/ELEM, an error is issued if the distance is too
large. If GA and GB are used for the projection or if one of the ELEMID
or GRIDID methods is used, the test only issues a warning if the
distance is too large. If GSTOL is zero, any distance is accepted.
PROJTOL 0.0 < Real < 0.2 0.0 Tolerance to accept the projected point GA or GB if the computed
coordinates of the projection point lie outside the patch boundary, but
are located within PROJTOL*(dimension of the patch).
ACTVTOL Integer > 0
Integer < 2211
1111 Parameter controlling the behavior of PROJTOL for the different
CWELD/CFAST connection methods. This parameter is entered as an
integer and is converted to a four-character string. If its value is less
than 1000, the string is prepended with zeros. The first character (from
the left) controls the behavior when the PARTPAT/PROP method is
used. The second controls the behavior when the ELPAT/ELEM method
is used. The third controls the behavior when the ELEMID method is
used and the fourth controls the behavior when the GRIDID method is
used. For ALIGN, the PROJTOL tolerance has no significance. Each
digit in the string can have the value 0 or 1 or 2, where the value
2 only has significance for the ELPAT/ELEM or PARTPAT/PROP
methods. The values have the following meaning:
0 = PROJTOL is completely deactivated
1 = PROJTOL is activated for ELEMID and GRIDID, PROJTOL is
activated in initial projections for ELPAT/ELEM, PROJTOL is
only activated over free edges of the patch in auxiliary
projections for ELPAT/ELEM, and in initial and auxiliary
projections for PARTPAT/PROP. Free edges have no neighbors
within the set that defines the complete surface.
2 = PROJTOL is always activated
d
i
( )
683
CHAPTER 9
Boundary Conditions
CWSETS Integer > 0
(0 or 1)
0 Parameter to control the automatic creation of four element sets with
the elements involved in the CWELD/CFAST connections.
0 = the sets are not created
1 = four sets are created automatically:
fastener_all_beams_inc000, the set containing all
connector beam elements.
fastener_all_faces_sidea_inc0000, the set
containing all elements with patches on side A of the
connection.
fastener_all_faces_sideb_inc0000, the set
containing all elements with patches on side B of the
connection.
fastener_all_warnings_inc0000, the set
containing all elements involved in CWELD/CFAST warning
messages.
Defining sets with any of these names must be avoided and are
considered an error.
MAXEXP Integer > 0 2 Parameter to control the maximum number of expansions in the
search for projections of the auxiliary nodes. First, the master patch is
tried. If no projection is found on the master patch, a first expansion
is made including all neighboring patches of the master patch. If no
projection is found on any of the new patches, a second expansion is
made including all neighbors of the patches tried so far. This process
continues until the number of expansions exceeds MAXEXP. Two
patches are neighbors if they share at least one node in their
connectivities.
DLDMIN Real > 0.0 0.2 Default value for LDMIN; the smallest ratio of length to characteristic
diameter.
DLDMAX Real > 0.0 5.0 Default value for LDMAX; the largest ratio of length to characteristic
diameter.
MAXITR Integer > 0 25 The maximum number of iterations allowed in the iteration process
for finding the projection on a patch.
EPSITR Real > 0.0 1.0E-5 Tolerance to terminate the iteration process for finding the projection
on a patch. If the parametric coordinate change in an iteration is less
than EPSITR, the projection is accepted as converged.

DELMAX Real > 0.0 0.1 Maximum allowable parametric coordinate change during the
iteration process for finding the projection on a patch. At first DELMAX
is not activated (i.e., the parametric coordinate change is not limited
during the iteration process). The parameter is only activated when the
full Newton Raphson iteration process for a projection did not
converge. In that case, the iteration process is restarted with DELMAX
activated.
CWSPOT 0 < Integer < 3 1 Parameter to choose the method for modifying the beam length.
1 = scale the stiffness of the beam
2 = reposition the end nodes of the beam
3 = reposition the auxiliary cpath nodes and the end nodes of the
beam.
RBE3WT Real 0.0 Default RBE3 distance weighting exponent.
The weight factor for each retained node in a RBE3 involved in a
CWELD/CFAST connection is:
where
is the weighting factor for retained node i.
is the distance from the tied node to retained node i
is the weighting exponent RBE3WT.
Negative values for RBE3WT are not recommended, since they result
in heavier weighting for nodes further away. The default results in
uniform weighting .
BOXING -1 < Integer < 1 0 Parameter to control the boxing algorithm used to speed up the search
for master patches when connection method PARTPAT/PROP is used.
-1 = The boxing algorithm is always deactivated
0 = The boxing algorithm may or may not be activated depending
on the number of elements in the sets.
1 = The boxing algorithm is always activated
f
i
1
d
i
n
----- =
f
i
d
i
n
f
i
( ) 1 =
Chapter 10 Element Library
10
Element Library
Truss Elements 687
Membrane Elements 688
Continuum Elements 688
Beam Elements 689
Plate Elements 690
Shell Elements 690
Heat Transfer Elements 690
Magnetodynamic Analysis 692
Soil Analysis 692
Fluid Analysis 692
Piezoelectric Analysis 692
Special Elements 692
Incompressible Elements 694


Marc includes an extensive element library. The element library allows you to model various types of one-, two-, and
three-dimensional structures, such as plane stress and plane strain structures, axisymmetric structures, full
three-dimensional solid structures, and shell-type structures. (See Marc Volume B: Element Library for a detailed
description of each element, a reference to the use of each element, and recommendations concerning the selection of
element types for analysis).
After you select an element type or a combination of several element types for your analysis with the ELEMENTS or
SIZING parameter, you must prepare the necessary input data for the element(s). In general, the data consist of element
connectivity, thickness for two-dimensional beam, plate and shell elements, cross section for three-dimensional beam
elements, coordinates of nodal points, and face identifications for distributed loadings.
You can choose different element types to represent various parts of the structure in an analysis. If there is an
incompatibility between the nodal degrees of freedom of the elements, you have to provide appropriate tying
constraints to ensure the compatibility of the displacement field in the structure. Marc assists you by providing many
standard tying constraint options, but you are responsible for the consistency of your analysis.
You can use almost all of the Marc elements for both linear and nonlinear analyses, with the following exceptions.
No plasticity or cracking is allowed in the elastic beams, the shear-panel, and the Fourier elements.
The updated Lagrange and finite strain plasticity features are not available for all elements.
Plasticity is available for the Herrmann elements when using the F
e
F
p
formulation only.
Only heat links and two-dimensional heat transfer elements can be used for hydrodynamic bearing analysis.
Fourier analysis can be carried out only for a limited number of axisymmetric elements.
Marc defines all continuum elements in the global coordinate system. Truss, beam, plate, and shell elements are
defined in a local coordinate system and the resulting output must be interpreted accordingly. You should give special
attention to the use of these elements if the material properties have preferred orientations. The ORIENTATION option
is available to define the preferred directions.
All Marc elements are numerically integrated. Element quantities, such as stresses, strains, and temperatures, are
calculated at each integration point of the element if you use the ALL POINTS parameter. This is the default in Marc.
These quantities are computed only at the centroid of the element, if the CENTROID parameter is used.
Distributed loads can be applied along element edges, over element surfaces, or in the volume of the element. Marc
automatically evaluates consistent nodal forces through numerical integration. (See Marc Volume B: Element Library
for details on this process). Concentrated forces must be applied at the nodal points.
All plate and shell elements can be used in a composite analysis. You can have a variable thickness shell, and control
the thickness and material property and orientation for each layer. For the thick shell elements (types 22, 45, 75, or
140), the interlaminar shear can also be calculated.
Five concepts differentiate the various elements types. These concepts are listed invalid for pure heat transfer. below.
1. The type of geometric domain that the element is modeling. These geometric domains are:
Truss
Membrane
Beam
Plate
Shell
687
CHAPTER 10
Element Library
Plane stress
Plane strain
Generalized plane strain
Axisymmetric
Three-dimensional solid
Special
2. The type of interpolation (shape) functions used in the element. These functions are:
Linear
Quadratic
Cubic
Hermitian
Special
The interpolation function is used to describe the displacement at an arbitrary point in the body.
(10-1)
where is the displacement at x, are the interpolation (shape) functions, and are the generalized nodal
displacements.
Engineering strain is
(10-2)
Therefore, the computational evaluation is
(10-3)
Hence,
(10-4)
3. The number of nodes in a particular element.
4. The number of degrees of freedom associated with each node, and the type of degrees of freedom.
5. The integration method used to evaluate the stiffness matrix. Marc contains elements which use full integration
and reduced integration.
Truss Elements
Marc contains 2- and 3-node isoparametric truss elements that can be used in three dimensions. These elements have
only displacement degrees of freedom.
Since truss elements have no shear resistance, you must ensure that there are no rigid body modes in the system.
u
i
x ( ) N
i
x ( )u
i
=
u x ( ) N u
i j
x ( ) 1 2
u
i
x ( )
x
j
----------------
u
j
x ( )
x
i
---------------- +
\ .
| |
=
i j
x ( ) 1 2
N
i
u
i
x
j
--------------
N
j
u
j
x
i
-------------- +
\ .
| |

i j
u
i
= =
i j
N
i
x
j
--------- =

Membrane Elements
Marc contains 3-, 4-, 6-, and 8-node isoparametric membrane elements that can be used in three dimensions. These
elements have only displacement degrees of freedom.
Since membrane elements have no bending resistance, you must ensure that there are no rigid body modes in the
system.
Membrane elements are often used in conjunction with beam or truss elements.
Continuum Elements
Marc contains continuum elements that can be used to model plane stress, plane strain, generalized plane strain,
axisymmetric and three-dimensional solids. These elements have only displacement degrees of freedom. As a result,
solid elements are not efficient for modeling thin structures dominated by bending. Either beam or shell elements
should be used in these cases.
The solid elements that are available in Marc have either linear or quadratic interpolation functions.
They include
3-, 4-, 6-, and 8-node plane stress elements
3-, 4-, 6-, and 8-node plane strain elements
6 (4 plus 2)- and 10 (8 plus 2)-node generalized plane strain elements
3-, 4-, 6-, and 8-node axisymmetric ring elements
8-, 10-, and 20-node brick elements
4- and 10-node tetrahedron
6-node and 15-node pentahedral
12- and 27-node semi-infinite brick elements
In general, the elements in Marc use a full-integration procedure. Some elements use reduced integration. The lower-
order reduced integration elements include an hourglass stabilization procedure to eliminate the singular modes.
Continuum elements are widely used for thermal stress analysis. For each of these elements, there is a corresponding
element available for heat transfer analysis in Marc. As a result, you can use the same mesh for the heat transfer and
thermal stress analyses.
Marc has no singular element for fracture mechanics analysis. However, the simulation of stress singularities can be
accomplished by moving the midside nodes of 8-node quadrilateral and 20-node brick elements to quarter-point
locations near the crack tip. Many fracture mechanics analyses have used this quarter-point technique successfully.
The 4- and 8-node quadrilateral elements can be degenerated into triangles, and the 8-and 20-node solid brick elements
can be degenerated into wedges and tetrahedra by collapsing the appropriate corner and midside nodes. The number
of nodes per element is not reduced for degenerated elements. The same node number is used repeatedly for collapsed
sides or faces. When degenerating incompressible elements, exercise caution to ensure that a proper number of
Lagrange multipliers remain. You are advised to use the higher-order triangular or tetrahedron elements wherever
possible, as opposed to using collapsed quadrilaterals and hexahedra.
689
CHAPTER 10
Element Library
Beam Elements
Marcs beam elements are 2- and 3-node, two- and three-dimensional elements that can model straight and curved-
beam structures and framed structures and can serve as stiffeners in plate and shell structures. A straight beam of a
circular cross section can be used for modeling the straight portion of piping systems. Translational and rotational
degrees of freedom are included in beam elements. The cross section of the beam can be solid or thin-walled. The solid
sections can be standard elliptical, rectangular, trapezoidal, or hexagonal sections or arbitrary solid sections. The thin-
walled sections can be standard closed circular sections or arbitrary closed or open sections. The BEAM SECT
parameter is used to define all solid sections and to define arbitrary thin-walled open or closed sections.
The beam elements are numerically integrated along their axial direction using Gaussian integration. The stress strain
law is integrated through thin-walled sections using a Simpson rule and through solid sections using a Simpson,
Newton-Cotes, or Gauss rule depending on the input specifications. Stresses and strains are evaluated at each
integration point through the thickness. This allows an accurate calculation if nonlinear material behavior is present.
In elastic beam elements, only the total axial force and moments are computed at the integration points.
If element 52 or 98 is used with numerical section integration, six stiffness factors may be used to modify its stiffness
behavior in axial tension, bending, torsion and shear. The modified internal virtual work of beam element type 98 when
using any of the stiffness factors is given as:

For element type 52 the transverse shear terms are absent. In this expression, is the normal stiffness factor, and
are the bending stiffness factors, and are the transverse shear stiffness factors and is the torsional stiffness
factor. All stiffness factors have a default value of 1 and must be positive. If stiffness factors with other values than 1
are used, the stress output must be interpreted as follows:
Generalized stresses will be scaled with their respective stiffness factors (i.e., , , etc. will be written
to the output or post file)
Layer stresses are not scaled with their respective stiffness factors, meaning that their section integration will
result in the unscaled generalized stresses (i.e., , , etc.)
If the section is used as a pre-integrated section all stiffness factors can have arbitrary positive values, but if the section
is used with numerical integration throughout the analysis the stiffness contributions resulting from the direct stresses
must use identical positive stiffness factors (i.e., ) and the stiffness contributions resulting from the
shear stresses must use identical positive stiffness factors (i.e. ) in order to guarantee symmetry of the
stiffness matrix. If these conditions are not met, and will silently be set to and and will silently be set
to .
W
i nt
f
1
N
z
z
f
2
M
z
z
f
3
M
y
y
f
4
D
x
x
f
5
D
y
y
f
6
T
z
z
+ + + + + ( ) z d
0
L
}
=
f
1
f
2
f
3
f
4
f
5
f
6
f
1
N
z
f
2
M
x
N
z
M
x
f
1
f
2
f
3
= =
f
4
f
5
f
6
= =
f
2
f
3
f
1
f
4
f
5
f
6

Plate Elements
The linear shell elements (Types 49, 72, 75, 138, 139, or 140) or the quadratic element (Type 22) can be used
effectively to model plates and have the advantage that tying is unnecessary. For element type 49, you can indicate on
the GEOMETRY option that the flat plate formulation is to be used. This reduces computational costs. To further
reduce computational costs for linear elastic plate analysis, the number of points through the thickness can be reduced
to one by use of the SHELL SECT parameter.
Shell Elements
Marc contains three isoparametric, doubly curved, thin shell elements: 3-, 4-, and 8-node elements (Types 4, 8, and
24, respectively) based on Koiter-Sanders shell theory. These elements are C
1
continuous and exactly represent rigid-
body modes. The program defines a mesh of these elements with respect to a surface curvilinear coordinate system.
You can use the FXORD model definition option to generate the nodal coordinates. Tying constraints must be used at
shell intersections.
Thin shell analysis can be performed using either the 3-node discrete Kirchhoff theory (DKT) based element (Type
138), the 4-node bilinear DKT element (Type 139), the 6-node bilinear semi-loof element (Type 49), or the 8-node
bilinear semi-loof element (Type 72). Thick shell analysis can be performed using the 4-node bilinear Mindlin element
(Type 75), the 4-node reduced integration with hourglass Mindlin element (Type 140), or the 8-node quadratic Mindlin
element (Type 22). The thick shell elements have been developed so that there is no locking when used for thin
shell applications.
The global coordinate system defines the nodal degrees of freedom of these elements. These elements are convenient
for modeling intersecting shell structures since tying constraints at the shell intersections are not needed.
Marc contains three axisymmetric shell elements: 2-node straight, 2-node curved, and 3-node curved. You can use
these elements to model axisymmetric shells; combined with axisymmetric ring elements, they can be used to simulate
the thin and thick portions of the structure. The program provides standard tying constraints for the transition between
shell and axisymmetric ring elements.
Heat Transfer Elements
The heat transfer elements in Marc consist of the following:
2- and 3-node three-dimensional links
3-, 4-, 6-, and 8-node planar and axisymmetric elements
6- and 9-node planar and axisymmetric semi-infinite elements
8-, 10-, and 20-node solid elements
4- and 10-node tetrahedral
6-, 15-node pentahedral
12- and 27-node semi-infinite brick elements
2-, 3-, 4-, 6-, and 8-node shell elements.
3-, 4-, 5-, and 8-node 3-D membrane elements
4-, 8-node planar interface elements
4-, 8-node axisymmetric interface elements
6-, 8-, 15-, 20-node solid interface elements.
691
CHAPTER 10
Element Library
For each heat transfer element, there is at least one corresponding stress element, enabling you to use the same mesh
for both the heat transfer and thermal stress analyses. Heat transfer elements are also employed to analyze coupled
thermo-electrical (Joule heating) problems.
In heat transfer continuum elements or link elements, temperature is the only nodal degree of freedom.
The heat transfer shell elements can be used in two modes. In the first mode, the number of degrees of freedom is two
or three for linear or quadratic temperature distribution across the complete thickness. In the second mode, the number
of degrees of freedom are or for linear or quadratic temperature distribution for each layer. This
is defined through the HEAT parameter.
The heat transfer membrane elements have only one degree of freedom per node, so there is no thermal gradient
through the thickness.
In Joule heating, the voltage and temperature are the nodal degrees of freedom. Shell elements are not available for
Joule heating.
Acoustic Analysis
Heat transfer elements are also used to model the compressible media in acoustic analysis. In this case, the pressure is
the single nodal degree of freedom.
Heat transfer elements are also used for electrostatic analysis. The scalar potential is the degree of freedom.
Coupled Electrostatic-Structural
Stress elements are used for a coupled electrostatic-structural analysis. For each of these elements, there is a
corresponding heat transfer element available for the electrostatic pass. When a medium has no stiffness, like air, heat
transfer elements can be chosen for this region. Then this region will be inactive in the stress pass.
Fluid/Solid Interaction
Heat transfer elements are used to model the inviscid, incompressible fluid/solid interaction problems. The hydrostatic
pressure is the degree of freedom.
The three-dimensional heat transfer links and planar elements are used to model the lubricant film. As no variation
occurs through the thickness of the film, two-dimensional problems are reduced to one-dimensional, and three-
dimensional problems are reduced to two-dimensional. The pressure is the degree of freedom.
For two-dimensional problems, a scalar potential can be used; hence, the heat transfer planar and axisymmetric
elements are employed. The single degree of freedom is the potential. For three-dimensional analyses, magnetostatic
elements are available. In such cases, there are three degrees of freedom to represent the vector potential.
M 1 + 2
*
M 1 + M

In magnetodynamic problems, a vector potential, augmented with a scalar potential, is used. There is a full family of
lower-and higher-order, planar, axisymmetric, and 3-D elements available for these analyses.
Soil Analysis
There are three types of soil/pore pressure analysis. If a pore pressure analysis only is performed, heat transfer
elements (41, 42, or 44) are used. If an uncoupled soil analysis is performed, the standard elements (21, 27, or 28) are
used. If a coupled analysis is performed, the Herrmann elements (32, 33, or 35) are used. In this case, the last degree
of freedom is the pore pressure.
Fluid Analysis
When performing fluid analysis, any planar, axisymmetric or solid continuum element can be used. These elements
either used a mixed formulation or a penalty formulation based upon the value of the FLUID parameter. For the mixed
formulation, each node of lower- or higher-order continuum element has velocities as well as pressure degrees of
freedom. The penalty method yields elements with only the velocities as the nodal variables.
In a piezoelectric analysis, a strong coupling exist between stress and electric field. There are lower-order elements
available for this analysis. The first two or three degrees of freedom (in resp. 2-D or 3-D) are available for
displacement components, and the last (3rd or 4th) degree of freedom is available for the electric potential.
Special Elements
Marc contains unique features in the program.
Gap-and-Friction Elements
The gap-and-friction elements (12, 97) are based on the imposition of gap closure constraint and frictional-stick or
frictional-slip through Lagrange multipliers. These elements provide frictional and gapping connection between any
two nodes of a structure; they can be used in several variations, depending on the application. In the default
formulation, the elements simulate a gap in a fixed direction, such that a body does not penetrate a given flat surface.
Using the optional formulation, you can constrain the true distance between two end-points of the gap to be greater
than some arbitrary distance. This ability is useful for an analysis in which a body does not penetrate a given two-
dimensional circular or three-dimensional spherical surface. Finally, you can update the gap direction and closure
distance during analysis for the modeling of sliding along a curved surface.
693
CHAPTER 10
Element Library
Pipe-bend Element
The pipe-bend (3-node elbow) element 17 is designed for linear and nonlinear analysis of piping systems. It is a
modified axisymmetric shell element for modeling the bends in a piping system. The element has a beam mode
superposed on the axisymmetric shell modes so that ovalization of the cross-section is admitted. The twisting of a pipe-
bend section is ignored because the beam has no flexibility in torsion. Built-in tying constraints are used extensively
for coupling the pipe-bend sections to each other and to straight beam elements.
Curved-pipe Element
The curved-pipe (2-node) element 31 is designed for linear analysis only. The stiffness matrix is based upon the
analytic elastic solution of a curved pipe.
Shear Panel Element
The shear panel (4-node) element 68 is an elastic element of arbitrary shape. It is an idealized model of an elastic sheet.
This element only provides shearing resistance. It must be used with beam stiffeners to ensure any normal or bending
resistance. The shear-panel element is restricted to linear material and small displacement analysis.
Cable Element
The cable element (51) is an element which exactly represents the catenary behavior of a cable. It is an elastic element
only.
Rebar Elements
The rebar elements (23, 46, 47, 48, 142, 143, 144, 145, 146, 147, 148, 165, 166, 167, 168, 169, and 170) are hollow
elements in which you can place single-direction strain members (reinforcing cords or rods). The rebar elements are
used in conjunction with other solid elements (filler) to represent a reinforced material such as reinforced concrete.
The reinforcing members and the filler are accurately represented by embedding rebar elements into solid elements.
The rebar elements can be used for small as well as large strain behavior of the reinforcing cords. Also, any kind of
material behavior can be simulated by the rebar elements in Marc.
Interface Elements
Elements 186 through 193 are special elements which are used to model the delamination in composite structures
where each ply is modeled with solid elements. The material behavior of these elements is defined with the COHESIVE
model definition option.
Elements 220 through 227 are analogous elements for thermal analysis of interfaces. The material of these elements
is defined with the COHESIVE thermal option.
Note that these interface elements do not have a mass matrix or a specific heat matrix and distributed loads, fluxes, or
films are not applied to these elements. One can use either a Guass or Newton-Cotes intergration scheme. The latter
may be advantageous in cases where the interface has significantly different stiffness compared to the surrounding
material.

Incompressible Elements
Incompressible and nearly incompressible materials can be modeled by using a special group of elements in the
program. These elements, based on modified Herrmann variational principles, are capable of handling large
deformation effects as well as creep and thermal strains. The incompressibility constraint is imposed by using
Lagrange multipliers. Generally, the low (linear) order elements have a single additional node which contains the
Lagrange multiplier, while the high-order (quadratic) elements have Lagrange multipliers at each corner node.
Elements 155 (plane strain triangle), 156 (axisymmetric triangle), and 157 (3-D tetrahedron) are low-order elements
and are exceptions within the incompressible element group. They have an additional node located at the center of the
elements and have Lagrange multipliers at each corner node. The shape function of the center node is a bubble
function. See Figure 10-1 for an element 155 case. The degrees of freedom of the center node is condensed out on the
element level before the assembly of the global matrix.
Figure 10-1 Element 155
Large Strain Elasticity
The incompressible elements based on Herrmann formulations can be used for large strain analysis of rubber-like
materials, using either total Lagrangian formulation or updated Lagrangian formulation.
Large Strain Plasticity
Element Library
The incompressible elements can be used for large strain analysis of elasto-plastic materials using updated Lagrangian
formulation, based on multiplicative decomposition of deformation gradient.
Incompressible Elements
Rigid-Plastic Flow
In rigid-plastic flow analysis, with effectively no elasticity, incompressible elements must be used. For such analyses,
you have a choice of using either the elements with Lagrange multipliers discussed above, or standard displacement
continuum elements. In the latter case, a penalty function is used to satisfy the incompressibility constraint.
x
1
2
3
4
Nodes with both displacements and pressure as
degrees of freedom
Center node with only displacement degrees of
freedom. The shape function for this node is a
bubble function.
x
695
CHAPTER 10
Element Library
Constant Dilatation Elements
You can choose an integration scheme option, which makes the dilatational strain constant throughout the element.
This can be accomplished by setting the second field of the GEOMETRY option to one. Constant dilatational elements
are recommended for use in approximately incompressible, inelastic analysis, such as large strain plasticity, because
conventional elements can produce volumetric locking due to overconstraints for nearly incompressible behavior. This
option is only available for elements of lower order (Types 7, 10, 11, 19, and 20). For the lower-order reduced
integration elements (114 to 117) with hourglass control, as only one integration point is used, these elements do not
lock, and effectively, a constant dilatation formulation is used. (The constant dilatation approach has also been referred
to mean dilatation and B-bar approach in the literature.)
Reduced Integration Elements
Marc uses a reduced integration scheme to evaluate the stiffness matrix or the thermal conductivity matrix for a
number of isoparametric elements. The mass matrix and the specific heat matrix of the element are always fully
integrated.
For lower-order, 4-node quadrilateral elements, the number of numerical integration points is reduced from 4 to 1; in
8-node solid elements, the number of numerical integration points is reduced from 8 to 1. For 8-node quadrilateral
elements, the number of numerical integration points is reduced from 9 to 4; for 20-node solid elements, the number
of numerical integration points is reduced from 27 to 8.
The energy due to the higher-order deformation mode(s) associated with high-order elements is not included in the
analysis. Reduced integration elements and fully integrated elements can be used together in an analysis. Reduced
integration elements are more economical than fully integrated elements and they can improve analysis accuracy.
However, with near singularities and in regions of high-strain gradients, the use of reduced integration elements can
lead to oscillations in the displacements and produce inaccurate results. Using reduced integration elements results in
zero-energy modes, or breathing nodes. In the lower-order elements, an additional stabilization stiffness is added
which eliminates these so-called hourglass modes.
Continuum Composite Elements
There is a group of isoparametric elements in Marc which can be used to model composite materials. Different material
properties can be used for different layers within these elements. The number of continuum layers within an element,
the thickness of each layer, and the material identification number for each layer are input via the COMPOSITE option.
A maximum number of 1020 layers can be used in one 2-D continuum composite element. A 3-D continuum
composite element only allows a maximum of 510 layers within the element.
These structural elements are available as both lower- and higher-order and can be used for plane strain (element types
151 and 153), axisymmetric (element types 152 and 154) or 3-D solid analysis (element types 149 and 150).
Corresponding heat transfer elements are available for lower- and higher-order and can be used for planar (element
types 177 and 179), axisymmetric (element types 178 and 180) or 3-D solid analysis (element types 175 and 176).

Fourier Elements
A special class of elements exists which allows the analysis of axisymmetric structures with nonaxisymmetric loads.
The geometry and material properties of these structures do not change in the circumferential direction, and the
displacement can be represented by a Fourier series. This representation allows a three-dimensional problem to be
decoupled into a series of two-dimensional problems. Both solid and shell Fourier elements exist in Marc. These
elements can only be used for linear elastic analyses.
Semi-infinite Elements
This group of elements can be used to model unbounded domains. In the semi-infinite direction, the interpolation
functions are exponential, such that the function (displacement) is zero at the far domain. The rate of decay of the
function is dependent upon the location of the midside nodes. The interpolation function in the non semi-infinite
directions are either linear or quadratic.
These elements can be used for static plane strain, axisymmetric, or 3-D solid analysis. They can also be used for heat
transfer, electrostatic, and magnetostatic analyses.
Cavity Surface Elements
This group of elements can be used to define the boundaries of cavities where standard finite elements are not used;
for example, along rigid boundaries. These elements are for volume calculation purposes only and do not contribute
to the stiffness equations of the model. They can be used for plane strain, plane stress, axisymmetric, and 3-D analyses.
No material properties are needed for these elements. They do not undergo any deformation except if they are attached
or glued to other elements or surfaces.
Assumed Strain Formulation
Conventional isoparametric four-node plane stress and plane strain, and eight-node brick elements behave poorly in
bending. The reason is that these elements do not capture a linear variation in shear strain which is present in bending
when a single element is used in the bending direction. For the six elements (3, 7, 11, 160, 161, and 163), the
interpolation functions have been modified such that shear strain variation can be better represented. For the lower-
order reduced integration elements (114 to 123), an assumed strain formulation written with respect to the natural
coordinates is used. For elastic isotropic bending problems, this allows the exact displacements to be obtained with
only a single element through the thickness. This is invoked by using the ASSUMED STRAIN parameter or by setting
the third field of the GEOMETRY option to one.
Follow Force Stiffness Contribution
When activating the FOLLOW FOR parameter, the distributed loads are calculated based upon the current deformed
configuration. It is possible to activate an additional contribution which goes into the stiffness matrix. This improves
the convergence. This capability is available for element types 3, 7, 10, 11, 18, 72, 75, 80, 81, 82, 83, 84, 114, 115,
116, 117, 118, 119, 120, 139, 140, 149, 151, 152, 160, 161, 162, or 163. Inclusion of the follower force stiffness can
result in nonsymmetric stiffness for nonenclosed volumes, thereby, resulting in increased computational times. You
can flag the nonsymmetric solver in the SOLVER option.
697
CHAPTER 10
Element Library
Explicit Dynamics
The explicit dynamics formulation IDYN=5 model is restricted to the following elements:
1, 2, 3, 5, 6, 7, 9, 10, 11, 18, 19, 20, 52, 64, 75, 89, 98, 114, 115, 116, 117, 118, 119, 120, 130, 139, and 140.
When using this formulation, the mass matrix is defined semi-analytically; for example, no numerical integration is
performed. In addition, a quick method is used to calculate the Courant stability limit associated with each element.
For these reasons, this capability is limited to the elements mentioned above.
Adaptive Mesh Refinement
Marc has a capability to perform local adaptive mesh refinement to improve the accuracy of the solution. This
capability is invoked by using the ADAPTIVE parameter and model definition option. The adaptive meshing is
available for the following 2-D and 3-D elements:
2, 3, 6, 7, 10, 11, 18, 19, 20, 37, 38, 39, 40, 43, 75, 80, 81, 82, 83, 84, 111, 112, 113, 114, 115, 116, 117, 118,
119, 120, 121, 122, 123, 138, 139, 140, 155, 156, 157, 160, 161, 162, 163, 164, 196, 198, and 201.
Chapter 11 Solution Procedures for Nonlinear Systems
11
Solution Procedures for
Nonlinear Systems
Considerations for Nonlinear Analysis 699
Newton-Raphson Method 712
Modified Newton-Raphson Method 713
Direct Substitution 715
Arc-length Methods 715
Convergence Controls 722
Singularity Ratio 724
Solution of Linear Equations 726
Flow Diagram 731
Remarks 732
References 733
699
CHAPTER 11
Solution Procedures for Nonlinear Systems
This chapter discusses the solution schemes in Marc for nonlinear problems. Issues of convergence controls,
singularity ratio, and available solvers for linearized system of equations are also discussed. In a nonlinear problem, a
set of equations must be solved incrementally. The governing equation of the linearized system can be expressed, in
an incremental form, as
(11-1)
where and are the correction to the incremental displacements and residual force vectors, respectively.
There are several solution procedures available in Marc for the solution of nonlinear equations:
Newton-Raphson Method
Modified Newton-Raphson Method
Direct Substitution
Arc-length Methods
Considerations for Nonlinear Analysis
Nonlinear analysis is usually more complex and expensive than linear analysis. Also, a nonlinear problem can never
be formulated as a set of linear equations. In general, the solutions of nonlinear problems always require incremental
solution schemes and sometimes require iterations (or recycles) within each load/time increment to ensure that
equilibrium is satisfied at the end of each step. Superposition cannot be applied in nonlinear problems.
The iterative procedures supported in Marc are: Newton-Raphson, Modified Newton-Raphson, Newton-Raphson with
strain correction, and direct substitution. If the R-P flow contribution model is chosen, a direction substitution is used.
A nonlinear problem does not always have a unique solution. Sometimes a nonlinear problem does not have any
solution, although the problem can seem to be defined correctly.
Nonlinear analysis requires good judgment and uses considerable computing time. Several runs are often required.
The first run should extract the maximum information with the minimum amount of computing time. Some design
considerations for a preliminary analysis are:
Minimize degrees of freedom whenever possible.
Halve the number of load increments by doubling the size of each load increment.
Impose a coarse tolerance on convergence to reduce the number of iterations. A coarse run determines the area
of most rapid change where additional load increments might be required. Plan the increment size in the final
run by the following rule of thumb: there should be as many load increments as required to fit the nonlinear
results by the same number of straight lines.
Marc solves nonlinear static problems according to one of the following two methods: tangent modulus or initial
strain. Examples of the tangent modulus method are elastic-plastic analysis, nonlinear springs, nonlinear foundations,
large displacement analysis and gaps. This method requires at least the following three controls:
A tolerance on convergence
A limit to the maximum allowable number of recycles
Specification of a minimum number of recycles
Ku r =
u r

An example of the initial strain method is creep or viscoelastic analysis. Creep analysis requires the following
tolerance controls:
Maximum relative creep strain increment control
Maximum relative stress change control
A limit to the maximum allowable number of recycles
To input control tolerances, use the model definition option CONTROL for all simulations. These values can be reset
upon restart or through the CONTROL history definition option. For creep simulations, addition controls are provided
in the AUTO CREEP or AUTO STEP options. See Convergence Controls in this chapter for further discussion on
tolerance controls.
Behavior of Nonlinear Materials
Nonlinear behavior can be time- (rate-) independent, or time- (rate-) dependent. For example, plasticity is time-
independent and creep is time-dependent. Both viscoelastic and viscoplastic materials are also time-dependent.
Nonlinear constitutive relations must be modeled correctly to analyze nonlinear material problems. A comprehensive
discussion of constitutive relations is given in Chapter 7.
Scaling the Elastic Solution
The SCALE parameter causes scaling of the linear-elastic solution to reach the yield stress in the highest stressed
element. Scaling takes place for small displacement elastic-plastic analysis, where element properties do not depend
on temperature. The SCALE parameter causes all aspects of the initial solution to be scaled, including displacements,
strains, stresses, temperature changes, and loads. Subsequent incrementation is then based on the scaled solution.
Load Incrementation
Several history definition options are available in Marc to input mechanical and thermal load increments (see
Table 11-1). The choice is between a fixed and an automatic load stepping scheme.
For a fixed scheme, the load step size remains constant during a load case. The fixed schemes are AUTO LOAD for
static mechanical, CREEP INCREMENT for creep, DYNAMIC CHANGE for dynamic mechanical, and TRANSIENT
NON AUTO for thermal/thermo-mechanically coupled.
Table 11-1 History Definition Options for Load Incrementation
Load Type Fixed Adaptive
Mechanical AUTO LOAD
DYNAMIC CHANGE
CREEP INCREMENT
AUTO STEP
AUTO INCREMENT
AUTO THERM
AUTO CREEP
AUTO THERM CREEP*
Thermal TRANSIENT NON AUTO AUTO STEP
TRANSIENT*
*The option is obsolete; use AUTO STEP instead.
701
CHAPTER 11
In many nonlinear analyses, it is useful to have Marc automatically determine the appropriate load step size. For an
adaptive scheme, the load step size changes from one increment to the other and also within an increment depending
on convergence criteria and/or user-defined physical criteria. The adaptive schemes are AUTO STEP and AUTO
INCREMENT for static mechanical, AUTO STEP and AUTO CREEP for creep, AUTO STEP, AUTO THERM, and
AUTO THERM CREEP for thermally driven mechanical problems, and AUTO STEP for dynamic mechanical, and
AUTO STEP and TRANSIENT for thermal/thermo-mechanically coupled.
The adaptive stepping scheme of choice is AUTO STEP. AUTO STEP has been designed as a unified load stepping
scheme and many of the capabilities of the other stepping schemes can now be handled by AUTO STEP. AUTO STEP
can be used for mechanical (static, creep, dynamic), thermal, and thermo-mechanically coupled analysis problems.
In the AUTO STEP scheme, either a recycle based convergence criterion is used to automatically determine the
time step based upon a comparison of the actual number of recycles needed in an increment to a user-specified
desired number of recycles or a damping energy procedure is used. In addition to this, user-defined or
program-determined physical criteria based upon strain, stress, displacement, or temperature increments can be
used to control the time step. Reductions in the time step through cut-backs are used to satisfy both
convergence criteria and physical criteria.
More details on the AUTO STEP option are provided in the next section.
Post buckling or snap-through analyses require the so-called arclength method which is available through the
AUTO INCREMENT option. This option can only be used in static mechanical analyses and the applied load is
automatically increased or decreased in order to maintain a certain arclength.
For creep analysis, the available adaptive options are AUTO CREEP and AUTO STEP. The AUTO CREEP
option determines the time step in an explicit creep analysis based on the creep strain change and the stress
change (see Volume C: Program Input, Chapter 3). These checks are not available by default in AUTO STEP.
These may however be incorporated by using an absolute or relative creep strain increment criterion and an
absolute or relative stress increment criterion as additional user-defined physical criteria in combination with
the default convergence criterion. Addition of these user-defined criteria for AUTO STEP is quite simple. If an
appropriate input flag is set, two physical criteria are automatically added by the program at run-time for
explicit creep problems: creep strain increment/elastic strain = 0.5, and stress increment/stress at beginning of
increment = 0.5. It should also be noted that AUTO STEP is usually more reliable in cases involving creep and
contact.
For the case of automatic load stepping for a thermally loaded elastic-creep/elastic-plastic-creep stress
analysis, the available adaptive schemes are AUTO THERM CREEP and AUTO STEP. Allowable increments
for normalized creep strain, normalized stress and state variables can be optionally prescribed for AUTO STEP
either through user-defined criteria or program determined automatic physical criteria.
For thermally driven mechanical problems, the available options are AUTO THERM and AUTO STEP. The
thermal loads derived from a thermal analysis are applied using the CHANGE STATE option in a mechanical
analysis. In the AUTO THERM scheme, the load step of the mechanical analysis is automatically adjusted based
upon user-specified (allowed) changes in temperature from the thermal analysis per increment. For example, if
Note:
AUTO INCREMENT can also be used for general situations without instabilities, but, in general, the
AUTO STEP option is preferred for these situations.

there is a change of 50 in the thermal analysis in one increment but only a change of 10 per increment is
allowed in the mechanical analysis, AUTO THERM splits up the thermal increment into five mechanical
increments. Allowable state variable increments can be optionally prescribed for AUTO STEP either through a
user-defined criterion or program determined automatic physical criterion. If these criteria are violated in an
increment, AUTO STEP cuts the time-step back and repeats the increment with a smaller time step.
Selecting Load Increment Size
Selecting a proper load step increment is an important aspect of a nonlinear solution scheme. Large steps often lead to
many recycles per increment and, if the step is too large, it can lead to inaccuracies and nonconvergence. On the other
hand, using too small steps is inefficient.
Fixed Load Incrementation
When a fixed load stepping scheme is used, it is important to select an appropriate load step size that captures the
loading history and allows for convergence within a reasonable number of recycles. For complex load histories, it is
necessary to prescribe the loading through time tables while setting up the run.
For fixed stepping, there is an option to have the load step automatically cut back in case of failure to obtain
convergence. When an increment diverges, the intermediate deformations after each recycle can show large
fluctuations and the final cause of program exit can be any of the following: maximum number of recycles reached
(exit 3002), elements going inside out (exit 1005 or 1009) or, in a contact analysis, nodes sliding off a rigid contact
body (exit 2400), and nodes not being projected properly onto 3-D NURBS (exit 2401). These deformations are
normally not visible as post results (there is a feature to allow for the intermediate results to be available on the post
file, see the POST option). If the cutback feature is activated and one of these problems occur, the state of the analysis
at the end of the previous increment is restored and the increment is subdivided into a number of subincrements. The
step size is halved until convergence is obtained or the user-specified number of cutbacks has been performed. Once
a subincrement is converged, the analysis continues to complete the remainder of the original increment. No results
are written to the post file during subincrementation. When the original increment is finished, the calculation continues
to the next increment with the original increment count and time step maintained. If the global remeshing option is
activated in conjunction with the cutback feature, then, for exit 1005 or 1009, the chosen contact body is remeshed and
the analysis is repeated with the original time step before the first cutback.
Arc Length Method
The arc-length procedures assume that the control of the nonlinear behavior and possible instabilities is due to
mechanical loads, and that the objective is to obtain an equilibrium position at the end of the loadcase. Hence, while
the program may increase or decrease the load, the load can always be considered to be , where
and are the loads at the beginning and end of the loadcase. The scale factor does not necessarily vary linearly
from 0 to 1 over the increments, and may, in fact, become negative. This would result in negative time steps as well;
hence, the AUTO INCREMENT history definition option cannot be used with dynamics and should not be used with
table driven input where the load is a function of time.
Mechanical loads, as shown above, are applied in a proportional manner and thermal loads are applied instantaneously.
F F
b
F
e
F
b
( ) + =
F
b
F
e
703
CHAPTER 11
This means that any automatic load incrementation method is limited to mechanical input histories that only have
linear variations in load or displacement and thermal input histories that have immediate change in temperature. For
example, one may not use a rigid body with a linearly changing velocity, since the resulting displacement of the rigid
body would give parabolically changing displacements. In this case, one would need to use a constant velocity for the
arc length method to work properly.
For the arc length method, care must be taken to appropriately define the loading history in each loadcase. The load
case should be defined between appropriate break points in the load history curve. For example, in Figure 11-1, correct
results would be obtained upon defining three distinct loadcases between times , , and during the
model preparation. However, if only one load case is defined for the entire load history between , the total
applied load for the loadcase is zero. This is also true when the AUTO STEP history definition option is used when
the table driven input is not used.
Figure 11-1 Defining Loadcases for Automatic Load Incrementation
When the table driven input method is used with the AUTO INCREMENT history definition option and behavior as
shown in Figure 11-1 is desired, the independent variable should be the load case number and not time. A quasi-static
total mechanical load can, hence, be given which is both a function of position and the load case number.
AUTO STEP
The scheme appropriate for most applications is AUTO STEP. The primary control of the load step is based upon the
number of recycles needed to obtain convergence. There are a number of optional user-specified physical criteria that
can be used to additionally control the load step. The user inputs needed to define the AUTO STEP scheme are
described in Marc Volume C: Program Input.
Recycling Criterion
The default recycle based criterion works as follows: The user specifies a desired number of recycles. For most
problems, it is sufficient to provide a value in the range of three to five. For problems with severe nonlinearities or for
problems with very small convergence tolerances, it may be necessary to increase this number. This number is used
as a target value for the load stepping scheme. If the number of recycles required in the current increment is less than
the desired number, the load step for the next increment is increased. The time step increase is based on a factor, ,
that can also be specified by the user. Typical values for are in the range of 1.2 to 1.5. While the time step increase
0 t
1
t
1
t
2
t
2
t
3
0 t
3
P

(
L
o
a
d
)
P
2
P
1
t
1
t
2
t
3
t (Time)
P

(
L
o
a
d
)
P
2
P
1
t
3
t (Time)
0 0
a. Three Defined Loadcases b. One Defined Loadcase
S
u
S
u

is obviously more aggressive with larger scale factors, it should be noted that there may be excessive recycling and
cutbacks if sudden nonlinearities are encountered. In order to avoid this, Marc uses the following logic for higher scale
factors. If the actual number of recycles in an increment is greater than 60 % of the desired number of recycles (i.e.,
the current increment did not converge easily), the increase scale factor for the next increment is limited to 1.25 for
scale factor values between 1.25 and 1.5625, and to 80 % of the value for scale factors above 1.5625.
See Damping Energy for alternative approach.
Time Step Cutback Scheme
The load step is never increased during an increment. If the number of recycles needed to obtain convergence exceeds
the desired number, the load step size is scaled back, the recycling cutback number is incremented by 1 and the
increment is performed again with the new load step. The scaleback factor for the
th
cutback is taken as , where
the factor is calculated from the expression ; where is the maximum number
of recycling related cutbacks for the increment and is calculated from , is the time
increment before any recycling related cutbacks occur for the increment and is the minimum possible time step
for the increment. is equal to the value set by the user ( by default) if there is no quasi-static artificial damping
and is equal to times the value set by the user ( by default) if there is quasi-static artificial damping. The
scaleback factor for any cutback is the smaller of ( , ). This scheme guarantees that no matter what the
starting time step for an increment, the minimum time step is reached in a reasonable number of cutbacks if the
increment consistently fails to converge. For special cutbacks such as maximum number of recycles reached (exit
3002), elements going inside out (exit 1005 or 1009) or, in a contact analysis, nodes sliding off a rigid contact body
(exit 2400) nodes not being projected properly onto a 3-D NURB (exit 2401), the scaleback factor is the smaller of
( , 0.5). If the minimum time step is reached and the analysis still fails to converge, it is terminated with exit 3015.
If the proceed when not converged option is used, then the analysis proceeds to the next increment if and when the
maximum number of recycles are reached.
Exceptions
There are some exceptions to the basic scheme outlined above. If an increment is consistently converging with the
current load step and the number of recycles exceeds the desired number, the number of recycles is allowed to go
beyond the desired number until convergence or up to the user specified maximum number. The time step is then
decreased for the next increment by . An increment is determined to be converging if the convergence ratio was
decreasing in three previous recycles.
Special rules also apply in a contact analysis. For quasi-static problems, the AUTO STEP option is designed to only
use the automated penetration check option (see CONTACT option, 7th field of 2nd data block; option 3 is always
used). During the recycles, the contact status can keep changing (new nodes come in contact, nodes slide to new
segments, separate etc.). Whenever the contact status changes during an increment, a new set of contact constraints
are incorporated into the equilibrium equations and more recycles are necessary in order to find equilibrium. These
extra recycles, due to contact changes, are not counted when the recycle number is checked against the desired number
N
r
N
r
s
Nr
s s T
s
T
m
[ ]
2 N
r m
N
r m
1 + ( ) ( )
= N
r m
N
r m
log
10
10
5
*
T
s
T
m
( ) = T
s
T
m
T
m
10
5
10
3
10
8
s
Nr
1 S
u
s
Nr
1 S
u
705
CHAPTER 11
for determining if the load step needs to be decreased within the increment. Thus, only true Newton-Raphson iterations
are taken into account. For the load step of the next increment, the accumulated number of recycles during the previous
increment is used. This ensures that the time step is not increased when there are many changes in contact during the
previous increment.
Thermal Analysis
For the most part, the recycling criterion works in a similar fashion for thermal analysis (heat transfer analysis, or
thermal part of a coupled thermo-mechanical analysis). The recycling criterion is used to satisfy the tolerance value
provided for the temperature error in estimate on the CONTROL option; that is, the analysis cuts back and starts the
increment over if the temperature error in estimate is not consistently converging within the desired number of
recycles. If the temperature estimate consistently converges in three previous recycles, the analysis continues
recycling. Once the temperature estimate tolerance is satisfied, the actual incremental temperature change is
calculated and checked against the corresponding tolerance provided on the CONTROL option. It should be
noted that if a user criterion on temperature is also provided, the latter parameters over-ride the one provided on the
CONTROL option. More details on how user criteria are handled are described in the next section. If the incremental
temperature change is not satisfied, the time step is scaled-back using a factor = . In this case, the
thermal pass is simply continued with the smaller time step without a cut-back because if there was a cutback, the
analysis would have redo the temperature error in estimate convergence from scratch. However, if the maximum
number of allowed recycles is reached before thermal convergence is achieved, a cutback with a scale-back factor of
min( , 0.5) is made and the increment is repeated.
User-defined Physical Criteria
In addition to allowing Marc to use the number of recycles for automatically controlling the step size for AUTO STEP,
user-specified physical criteria can optionally be used for controlling the step size. The user-specified physical criteria
work as follows. The user can specify the maximum allowed incremental change within certain ranges for specific
quantities during an increment. The quantities available are displacements, rotations, stresses, strains, strain energy,
temperature (in thermal or thermomechanically coupled analyses), and state variables (in mechanical or
thermomechanically coupled analysis). These criteria can be utilized in one of two ways.
By default, they are used as limits, which means that the load step is immediately decreased if a criterion is violated
during any iteration of the current increment, but they do not influence the decision to change the load step for the next
increment; that is, only the actual number of recycles versus desired number of recycles controls the load step for the
next increment.
The criteria can also be used as targets; in which case, they are used to control the time step for the current and next
increments. If the calculated values of the criteria are higher than the user-specified values in any iteration, the time
step is scaled down and the current increment is repeated. If the calculated values of the criteria for the current
increment are consistently smaller than the user-specified values prior to convergence, the time step for the next
increment is scaled up. The scale factor used for reduction or increase is the ratio between the actual value and the
target value and this factor is limited by user-specified minimum and maximum factors (defaults to 0.1 and 10
respectively). If this type of load step control is used together with the recycle based control, the time step can be
reduced in the current increment due to whichever criterion that is violated. The decision to increase the step size for
the next increment is solely based upon the physical criteria.
T
a
T
m
0.8 T
m
T
a
( )
s
Nr

Specification of user-defined physical criteria can be further simplified by setting a special flag in the AUTO STEP
option that allows for physical criteria to be automatically added by the solver at run-time. These automatic criteria
serve as upper-bound controls to prevent run-away Newton-Raphson iterations that ultimately cause the program to
abort. Currently, four mechanical criteria are automatically added depending on the kind of analysis that is being run:
a total strain criterion is added for any large displacement analysis and the maximum allowable equivalent total strain
increment at any point in the model is set to 50%; a plastic strain criterion is added for any large displacement, finite
strain analysis and the maximum allowable equivalent plastic strain increment at any point in the model is set to10%;
a relative creep strain criterion and a relative stress change criterion are added for any explicit creep analysis wherein
the maximum allowable creep strain change/elastic strain and the maximum allowable equivalent stress
change/equivalent stress are each set to 0.5; a state variable criterion is added for any large displacement analysis
wherein the maximum allowable temperature increment is such that the equivalent stress increment associated with
the change in thermal properties of the materials does not exceed 50% of the total equivalent stress. These criteria are
only added in the analysis if there are no competing explicitly defined user-criteria found. It should also be noted that
these automatic criteria are only used as limits; i.e. they are used to control the time step within an increment but not
for the next increment.
Failure to satisfy user-defined physical criteria can occur due to two reasons the maximum number of cutbacks
allowed by the user can be exceeded, or the minimum time step can be reached. In this case, the analysis terminates
with exit 3002 and exit 3015, respectively. These premature terminations can be avoided by using the option to
continue the analysis even if physical criteria are not satisfied. If this flag is set on the AUTO STEP option, and either
the maximum number of user-allowed cutbacks or the minimum time step is reached, a mechanical analysis moves on
to the next increment if it is otherwise converged (see Convergence Controls in this chapter) or continues to recycle
and scales back based solely on the recycling criterion. Setting this flag for a thermal analysis simply allows it to move
on to the next increment.
User Programmed Time Steps Selection
In addition to controlling the time step through the recycle based criterion and the physical criteria, direct control of
the time step is possible through the use of the UTIMESTEP user subroutine. In this case, the new time step that is
determined by the auto step algorithm enters the program as input and the modified time step by the user is returned
as output. More details are provided in Marc Volume D: User Subroutines and Special Routines, Chapter 2.
Post Files Output
In many analyses it is convenient to obtain post file results at specified time intervals. This is naturally obtained with
a fixed load stepping scheme but not with an automatic scheme. Traditionally, the post output frequency is given as
every n
th
increment. With the AUTO STEP history definition option, you can request post output to be obtained at
equally spaced time intervals. In this case, the time step is temporarily modified to exactly reach the time for output.
The time step is then restored in the following increment.
Table Considerations
When tables are used in the new input format to specify loads with complex time variations, in most cases, it is
important that the exact peaks and valleys of the loading history are not missed due to the adaptive time stepping. By
default, AUTO STEP adjusts the time step temporarily so that the peaks and valleys of the loading history are reached
exactly. The time step is then restored in the following increment. In analyses where the boundary conditions are
obtained from experimental data containing many points, it is often useful to turn this behavior off, in which case large
time steps will be achieved and the load will be smoothed out.
707
CHAPTER 11
Quasi-Static Damping Scheme
The AUTO STEP option also has an optional artificial damping feature available for mechanical statics analyses.
Damping Energy
A comparison between the incremental strain energy and the estimated damping energy is used as a criteria for time
step control when either a 4 or 5 is given on the 10th field of the 2nd data block (idampit flag).The criteria based
upon the number of cycles is bypassed. Decisions to decrease and increase the time step are based on the damping
energy rate of the system. Furthermore, if a 4 is entered, then artificial damping is added to the system. The details of
the time stepping procedure and the damping procedure are as follows:
A damping factor, , is introduced, which at the start of the loadcase, is set to 0. The time step for the first increment
is set equal to the user given initial time step. During the first cycle of the loadcase, a small strain energy value
is calculated as
where is the first component of the stress-strain relation at the integration point and is 10
-6
. This small
energy value is used later to distinguish between real deformations of the structure and zero-energy modes.
During the assembly of the stiffness matrix and the right-hand side vector , the following contributions are added
to and , respectively.
where is the lumped mass matrix, is the incremental displacement and is the time increment. Note that
when , there is no change to the equations. During the recovery phase, the incremental damping energy for
the nth iteration is calculated as
when and as
when is not .
The check to determine if the time-step should be reduced through a cut-back is made as follows: If , perform
a cutback if
F
d
E
smal l
E
smal l
0.5D 1 1 , ( ) 1 ( )
2
V d
V
}
=
D 1 1 , ( ) E 1 ( )
K F
K F
K
damp
F
d
M ti mi nc =
F
damp
F
d
Mdu t imi nc =
M du t imi nc
F
d
0 =
E
damp
n
du
T
Mdu t imic =
F
d
0 =
E
damp
n
F
d
Mdu t iminc =
F
d
0
F
d
0 =
E E
smal l
and E
dmap
n
4E
damp
1
> >

where is the incremental strain energy. The time step reduction factor in this case is given by
Such a cutback is triggered in the first increment of the loadcase in the case of severe non-linearities, as expressed by
the fact that the damping energy for the nth iteration deviates considerably from that of the first iteration.
If is not , perform a cutback if
where is the predicted damping energy for the current increment, is the maximum damping
energy in any previously converged increment and is the global user-defined convergence tolerance. The reduction
factor in this case is set to 0.5. This cut-back is triggered in the case where the solution is clearly diverging and avoids
unnecessary recycles. The same reduction factor (0.5) is also used for cases when the maximum number of iterations
are reached or when user criteria are violated.
For the first increment of the loadcase, the calculation of and predicted energy is as follows:
Estimated strain energy for the loadcase
, where .
Estimated damping energy for the loadcase , where
.
where is a user-defined factor. Recommended default value is 10
-4
.
Predicted damping energy for next increment
where is the time step for the next increment and is the time increment for the
previous increment.
Set and
For subsequent increments of the loadcase, the modification of and the time step is as follows:
If the total strain energy is larger than twice the estimated strain energy , then
.
If , then increase the time step by a factor of 1.5.
E
f
r educ
E
damp
1
E
damp
n ( )
=
F
d
0
E
damp
n
4E
pr edi ct ed
and E
damp
n
E
max
and tol 2 > > >
E
pr edi ct ed
E
max
t ol
F
d
E
est i mat ed
E ti mi nc
2
=
E
max
max E
est i mat ed
E , ( ) = E total strain energy in system =
E
damp_est i mat ed
E
damp
n
toti nc ti minc =
t oti nc total time period =
F
d
E
max
E
damp_est i mat ed
=
E
pr edi ct ed
F
d
E
damp
n
t imincnew t imi ncol d =
t imi ncnew t imi ncol d
E
pr ev
F
d
E
damp
n
t imi nc = E
max
F
d
E
damp
n
=
F
d
E E
est i mat ed
F
d
F
d
E E
est i mat ed
=
E
damp
n
ti mi nc 1.1E
pr ev
<
709
CHAPTER 11
Avoiding Exit 3015
When the minimum time step is reached, Marc normally exits with 3015. If the quasi-static damping schemes
(10th field, 2nd data block) is a 0 or 1, a quasi-static damping option is triggered in an attempt to avoid this premature
exit: When exit 3015 is encountered for the first time, the increment is repeated with a new time step given by
, where fclrg is the maximum ratio between steps (4th field of the 2nd data block of the AUTO
STEP option = 10 by default). The increment is further stabilized by added the factored lumped mass matrix to the
stiffness matrix and modifying the force vector consistently. The damping factor associated with the lumped mass
matrix is again based on the ratio of the estimated strain energy to the estimated damping energy in a manner similar
to the idampit = 4 scheme. If exit 3015 is again encountered, a second attempt is made with and
again stabilizing the incremental solution by adding the quasi-static damping in the solution. It should be noted that
this procedure is not available for the first increment of the auto step loadcase since the damping factor is only
calculated after the first increment is completed. It should also be noted that once exit 3015 is encountered and
damping is turned on, it remains on for the rest of the loadcase.
Auto Step for Transient Dynamics
The AUTO STEP algorithm is further modified for transient dynamics problems:
When penetration is detected in dynamic contact problems, instead of using the default iterative penetration
procedure, a time cutback is made and the increment is repeated with a smaller time step that avoids the
penetration. This scheme allows for momentum conservation without spurious penetration induced oscillations
in the response. If no other cutbacks are made, the time step is restored in the following increment.
Additional, optimal checks are available in transient dynamic problems to control the time step since larger
time steps that may have been assessed based on the recycle based criterion or physical criteria can give rise to
unacceptable time integration errors. The time integration error check is only turned on if the 13th field of the
3rd data block of the AUTO STEP option is set to 1. If this field is not present or set to 0, the time integration
error check is skipped and the time stepping algorithm for dynamic problems is then similar to that of static
problems. If the check is turned on, an additional check is made at the end of each increment to see if the time
step needs to be reduced for the following increment. This check is only made for the Newmark-Beta (NB) and
the Single Step Houbold (SSH) methods. The scheme that is followed is a modified version of the scheme
outlined by Bergan and Mollestad [Ref. 1]. The time step for increment should satisfy the inequality
where is the incremental displacement vector at increment ; M is the mass matrix; is the tangent
stiffness matrix, and is a user-customizable scaling factor. The default value of is as follows:
for the SSH operator . and
for the NB operator ,
where is the smallest ratio between time steps (3rd field of the 2nd data block of the AUTO STEP
option = 0.1 by default). The smaller default value of for SSH problems compared to
t
mi n
t
new
fclrg t
mi n
=
t
new
fclrg2 t
mi n
=
n 1 +
t ( )
n 1 +
2
u
n
T
Mu
n
u
n
T
K
T
u
n
---------------------------------
u
n
n K
T

0.375 * fcsml =
0.75 * fcsml =
fcsml
0.0375 =

for NB problems is to better control the artificial damping inherent in the SSH method. The user
can control the accuracy of the solution by changing . The time step is only reduced if the value
predicted by the above equation is less than 67% of the current time step. The check is bypassed if is
already at , if the strain energy is negligible (for example, rigid body motion). When there are multiple
loadcases in a transient dynamic run and the time integration error check is flagged for all of them, the initial
time step of a loadcase that follows a transient dynamic loadcase is also adjusted if it is too high compared to
the time step predicted by the error check. This avoids accuracy problems associated with using an initial user-
prescribed time step that is too large.
The defaults of the AUTO STEP option are carefully chosen to be adequate in a wide variety of applications. There are
cases, however, when the settings may need to be modified. Assume that the default settings are used, which means
that the recycle based control is active with an initial load of one per cent of the total. If the structure is weakly
nonlinear, convergence is obtained in just a few recycles and the time steps for successive increments get progressively
larger. This can lead to problems if the initially weakly nonlinear structure suddenly exhibits stronger nonlinearities;
for instance, occurrence of plasticity or parts coming into contact. Possible remedies to this problem include:
a. decrease the time step scale factor to a smaller number so the step size does not grow so rapidly;
b. use a physical criterion like maximum increment of displacements to limit the load step;
c. use the maximum time step to limit large steps;
d. decrease the desired and maximum number of recycles to decrease the load step if more recycles are needed.
Another situation is if the structure is highly nonlinear and convergence is slow. In this case, it may be necessary to
increase the desired number and maximum number of recycles. In general, there is a close connection between the
convergence tolerances used and the desired number and maximum number of recycles. In many cases, it may be
beneficial to use one or more physical criteria; for example, the increment of plastic strain as targets for controlling
the load step. This can easily be achieved by allowing the program to add automatic physical criteria where
appropriate. This is especially a good idea if the proceed if not converged option is used or if the non-positive
definite flag is set since the added physical criteria then serve as controls to limit the time step and produce a realistic
numerical solution in each increment rather than letting the solution proceed unchecked with unrealistic results.
Residual Load Correction
The residual load is applied as a correcting force to ensure that equilibrium is maintained and, hence, that an accurate
solution is obtained for nonlinear problems.
The residual load correction enforces global equilibrium at the start of each new increment. This prevents the
accumulation of out-of-equilibrium forces from increment to increment and makes the solution less sensitive to the
step size. Figure 11-2 shows how stiffness is based on the state at the start of a step. The variables are defined below
for increments i = 1,2,3:
applied forces for i = 1,2,3
calculated displacements for i = 1,2,3
residual loads for i = 1,2,3
0.075 =
fcsml
t
n 1 +
t
mi n
F
i
u
i
R
i
711
CHAPTER 11
Figure 11-2 Stiffness Based on State at Start of Step
The residual load correction is the difference between the internal forces and the externally applied loads. The residual
load correction is expressed as
(11-2)
where is the differential operator which transforms displacements to strains, is the current generalized stresses,
is the total applied load vector, and is the residual load correction.
In order to evaluate the residual load correction accurately, evaluate the integral by summing the contributions from
all integration points. The residual load correction feature requires that stresses be stored at all the integration points.
Data storage at all integration points is the default in Marc, but can be overridden in linear analysis by use of the
CENTROID parameter.
Residual load correction should always be used unless the ELASTIC parameter is invoked; this is the default.
Restarting the Analysis
The RESTART model definition option creates a restart file for the current analysis which can be used in subsequent
analyses. It can also be used to read in a previously generated file to continue the analysis. The RESTART option is
very important for any multi-increment analysis because it allows you to continue the analysis at a later time. The
default situation writes the restart information to unit 8 and reads a previously generated file from unit 9. For post
processing, the RESTART option can be used to plot or combine load cases (see CASE COMBIN). Upon restart, you
can use the model definition REAUTO option to redefine parameters associated with an automatic load sequence.
To save storage space, it is not necessary to store each increment of analysis. The frequency can be set using the
RESTART option, and subsequently modified using the RESTART INCREMENT option. It is also possible to store only
the last converged solution by using the RESTART LAST option. The input data describing the problem is not saved,
and therefore must be read in with each restart. This should include the model geometry including CONNECTIVITY
and the COORDINATES material properties contact data and boundary conditions. Additional tables may be added, or
they can have additional data added. The RESTART option specifies restart parameters; for example, input/output files,
restart increment, and intervals at which restarts are to be written.
F
1
F
2
F
3
U
1
U
2
U
3
R
1
R
2
R
3
R P
T
dV
}
=

P R

The Newton-Raphson method can be used to solve the nonlinear equilibrium equations in structural analysis by
considering the following set of equations:
(11-3)
where is the nodal-displacement vector, is the external nodal-load vector, is the internal nodal-load vector
(following from the internal stresses), and is the tangent-stiffness matrix. The internal nodal-load vector is obtained
from the internal stresses as
(11-4)
In this set of equations, both and are functions of . In many cases, is also a function of (for example, if
follows from pressure loads, the nodal load vector is a function of the orientation of the structure). The equations
suggest that use of the Newton-Raphson method is appropriate. Suppose that the last obtained approximate solution
is termed , where indicates the iteration number. Equation (11-3) can then be written as
(11-5)
This equation is solved for and the next appropriate solution is obtained by
and (11-6)
Solution of this equation completes one iteration, and the process can be repeated. The subscript denotes the
increment number representing the state . Unless stated otherwise, the subscript is dropped with all
quantities referring to the current state.
The Newton-Raphson method is the default in Marc (see Figure 11-3). The Newton-Raphson method provides good
results for most nonlinear problems, but is expensive for large, three-dimensional problems, when the direct solver is
used. The computational problem is less significant when the iterative solvers are used. Figure 11-3 illustrates the
graphical interpretation of the Newton-Raphson iteration technique in one dimension to find the roots of the function
, starting from increment 1 where to increment 2 where.
.The iteration process stops when the convergence criteria are satisfied. Similarly, Figure 11-4
graphically demonstrates the modified Newton-Raphson iteration technique for the same one dimensional problem.
K u ( )u F R u ( ) =
u F R
K
R
T
v d
V
}
elem
=
R K u F u
F
u
i i
K u
n 1 +
i 1
( )u F R u
n 1 +
i 1
( ) =
u
i
u
i
u
i 1
u
i
+ = u
n 1 +
i
u
n
u
i
+ =
n
t n = n 1 +
F u ( ) 1 u 1 0 = = F u
0
( ) 0.2 =
F u
l ast
( ) 1.0 =
713
CHAPTER 11
Figure 11-3 Newton-Raphson
The modified Newton-Raphson method is similar to the Newton-Raphson method, but does not reassemble the
stiffness matrix in each iteration.
(11-7)
Figure 11-4 Modified Newton-Raphson
The process is computationally inexpensive because the tangent stiffness matrix is formed and decomposed once.
From then on, each iteration requires only forming the right-hand side and a backward substitution in the solution
process. However, the convergence is only linear, and the potential for a very large number of iterations, or even
nonconvergence, is quite high.
If contact or sudden material nonlinearities occur, reassembly cannot be avoided. The modified Newton-Raphson
method is effective for large-scale, only mildly nonlinear problems. When the iterative solver is employed, simple
back substitution is not possible, making this process ineffective. In such cases, the Newton-Raphson method should
be used instead.
0 0.5 1 1.5
0
0.2
0.4
0.6
0.8
1
1.2
Displacement (u)
F
o
r
c
e
F = u
1/2
Residual Checking
Displacement Checking
F
r esidual
8
F
r eaction
---------------------------- TOL
1
<
8
8
u
u
--------------- TOL
1
<
8
|| ||
|| ||
|| ||
|| ||
Convergence Criteria:
Inc 1
Inc 2
Residual
1
u
1
u
2
u
3
Recycle 1 2 3
3
4
0.992 0.993 0.994 0.995 0.996 0.997 0.998 0.999 1 1.001
0.995
0.996
0.997
0.998
0.999
1
1.001
K u
0
( )u
i
F R u
i 1
( ) =
0 0.5 1 1.5
0
0.2
0.4
0.6
0.8
1
1.2
Displacement (u)
F
o
r
c
e
Inc 1
Inc 2
0.97 0.98 0.99 1 1.01 1.02 1.03
0.97
0.98
0.99
1
1.01
1.02
Residual
1
u
1
u
2
u
3
Recycle 1 2 3
u
0
F = u
1/2
Residual Checking
Displacement Checking
F
r esidual
8
F
r eaction
---------------------------- TOL
1
<
8
8
u
u
--------------- TOL
1
<
8
|| ||
|| ||
|| ||
|| ||
Convergence Criteria:

If the load is applied incrementally, Marc recalculates the stiffness matrix at the start of each increment or at selected
increments, as specified.
Strain Correction Method
The strain correction method is a variant of the Newton method. This method uses a linearized strain calculation, with
the nonlinear portion of the strain increment applied as an initial strain increment in subsequent iterations and recycles.
This method is appropriate for shell and beam problems in which rotations are large, but membrane stresses are small.
In such cases, rotation increments are usually much larger than the strain increments, and, hence, the nonlinear terms
can dominate the linear terms. After each displacement update, the new strains are calculated from and
which yield
(11-8)
This expression is linear except for the last term. Since the iteration procedures start with a fully linearized calculation
of the displacement increments, the nonlinear contributions yield strain increments inconsistent with the calculated
displacement increments in the first iteration. These errors give rise to either incorrect plasticity calculations (when
using small strain plasticity method), or, in the case of elastic material behavior, yields erroneous stresses. These
stresses, in their turn, have a dominant effect on the stiffness matrix for subsequent iterations or increments, which
then causes the relatively poor performance.
The remedy to this problem is simple and effective. The linear and nonlinear part of the strain increments are
calculated separately and only the linear part of
(11-9)
is used for calculation of the stresses. The nonlinear part
(11-10)
is used as an initial strain in the next iteration or increment, which contributes to the residual load vector defined by
(11-11)
This strain correction term is defined by
(11-12)
Since the displacement and strain increments are now calculated in a consistent way, the plasticity and/or equilibrium
errors are greatly reduced. The performance of the strain correction method is not as good if the displacement
increments are (almost) completely prescribed, which is not usually the case. Finally, note that the strain correction
method can be considered as a Newton method in which a different stiffness matrix is used.
E
i 1 +
u
i
u = u
i
( )
E
i 1 +
E
i
=
1
2
--- ( u
,
u
,
) u
,
i
u
,
u
,
u
i
u
,
u
,
+ + + + +
E
l
( )

E
i
=
1
2
--- u
,
u
,
+ ( ) u
,
i
u
,
u
,
u
i
+ + +
E
nl
( )
i 1 + 1
2
--- u
,
u
,
=
R
C
,
X
,
L
nl
V d
V
0
}
=
K u
n 1 +
i
( )u F R u
n 1 +
i
( ) R
C
=
715
CHAPTER 11
Direct Substitution
In the Eulerian formulation (R-P FLOW parameter), the governing equation of the system can be expressed as
(11-13)
where is a velocity vector, and is a force vector. This equation is very nonlinear because is a nonlinear function
of . By default, a direct substitution method is used to solve the problem. If is the velocity at iteration , the result
of iteration is
(11-14)
If this method does not converge in 10 iterations, it is possible to switch into a Newton-Raphson method.
Arc-length Methods
The solution methods described above involve an iterative process to achieve equilibrium for a fixed increment of
load. Besides, none of them have the ability to deal with problems involving snap-through and snap-back behavior.
An equilibrium path as shown in Figure 11-5 displays the features possibly involved.
Figure 11-5 Snap-through Behavior
The issue at hand is the existence of multiple displacement vectors, , for a given applied force vector, . The arc-
length methods provide the means to ensure that the correct displacement vector is found by Marc. If you have a load
controlled problem, the solution tends to jump from point 2 to 6 whenever the load increment after 2 is applied. If you
have a displacement controlled problem, the solution tends to jump from 3 to 5 whenever the displacement increment
after 3 is applied. Note that these problems appear essentially in quasi-static analyses. In dynamic analyses, the inertia
forces help determine equilibrium in a snap-through problem.
Kv F =
v F K
v v
i
i
i 1 +
K v
i
( ) v
i 1 +
F =
Displacement (u)
Load (F)
Snap-through Behavior
r
: E
q
u
ilib
rium Path
3
4
5
6
2
u F

Thus, in a quasi-static analysis sometimes it is impossible to find a converged solution for a particular load (or
displacement increment):
This is illustrated in Figure 11-5 where both the phenomenon of snap-through (going from point 2 to 3) and snap-back
(going from point 3 to 4) require a solution procedure which can handle these problems without going back along the
same equilibrium curve.
As shown in Figure 11-6, assume that the solution is known at point A for load level . For arriving at point B on
the equilibrium curve, you either reduce the step size or adapt the load level in the iteration process. To achieve this
end, the equilibrium equations are augmented with a constraint equation expressed typically as the norm of
incremental displacements. Hence, this allows the load level to change from iteration to iteration until equilibrium is
found.
Figure 11-6 Intersection of Equilibrium Path with Constraining Spherical Surface
The augmented equation, , describes the intersection of the equilibrium curve with an auxiliary surface for
a particular size of the path parameter :
(11-15)
Variations of the parameter
moves the surface whose intersection with the equilibrium curve generates a sequence of points along the curve. The
distance between two intersection points, denoted with and , denoted by l is the so-called arc-length.
Linearization of Equation (11-15) around point A in Figure 11-6 yields:
(11-16)
n 1 +
F
n
F F =
n
F
Displacement (u)
Load (F)
Arc- Length Procedure
r : Equilib
r
iu
m

P
a
t
h
g
: C
o
n
s
t
r
a
i
n
t

S
u
r
f
a
c
e
l : a
r
c
le
n
g
t
h

n
F
n+1
F
n+2
F
n+3
F
A
B
c u , ( ) g
r u , ( ) F R u ( ) 0 = = c u , ( ) g u , ( ) 0 = =
0

K P
n
T
n
0
u
)
`

r
r
0
)
`

=
717
CHAPTER 11
where:
(11-17)
(11-18)
(11-19)
(11-20)
It can be noted that a standard Newton-Raphson solution procedure is obtained if the constraint condition is not
imposed. The use of the constraint equation causes a loss of the banded system of equations which would have been
obtained if only the matrix was used. Instead of solving the set of equations iteratively, the block elimination
process is applied.
Consider the residual at iteration to which the fraction of load level corresponds
(11-21)
The residual for some variation of load level, , becomes
(11-22)
which can be written as:
(11-23)
where (11-24)
and (11-25)
Notice that does not depend on the load level. The equation above essentially establishes the influence of a
change in the load level during one iteration on the change in displacement increment for that iteration. After one
iteration is solved, this equation is used to determine the change in the load level such that the constraint is followed.
There are several arc-length methods corresponding to different constraints.
Among them, the most well-known arc-length method is one proposed by Crisfield, in which the iterative solution in
displacement space follows a spherical path centered around the beginning of the increment. This requirement is
translated in the formula:
(11-26)
K
r
u
------ : P
r
------ = =
n
T c
u
------ : n
0
c
------ = =
r F R =
r
0
g u , ( ) =
K N 1 +
i
i 1
r
i
i 1
( )
i 1
F R
i
u
i 1
( ) =
i
r
i
i 1
i
+ ( )
i
F r
i
i 1
( ) + =
u
i
i 1
i
+ ( ) u
i
i 1
( )
i
u
*
i
+ =
u
i
i 1
( ) K
i
( )
1
r =
u
*
i
K
i
( )
1
F =
u
*
i
i
c l
2
u
i
u
i
= =

where l is the arc length. The above equation with the help of Equations (11-25) and (11-26) is applied as:
(11-27)
The equation above is interpreted with and in the prediction phase while retaining the full form of
Equation (11-27) in the correction phase. Two solutions for are available. There are several methods to determine
which root to select. By default, we choose the one that maintains a positive angle of the displacement increment from
one iteration to the next.
The two roots of this scalar equation are and . To avoid going back on the original load-deflection
curve, the angle between the incremental displacement vectors, and (before and after the current iteration,
respectively) should be positive. Two alternative values of (namely, and corresponding to
and are obtained and the cosine of two corresponding angles ( and ) are given by
(11-28)
and (11-29)
Once again, the prediction phase is interpreted with and , while Equations (11-28) and
(11-29) retain their full form in the correction phase.
As mentioned earlier, the appropriate root, or is that which gives a positive . In case both the
angles are positive, the appropriate root is the one closest to the linear solution given as:
(11-30)
Crisfields solution procedure, generalized to an automatic load incrementation process, has been implemented in
Marc as one of the options under the AUTO INCREMENT history definition option. Various components of this process
are shown in Figure 11-7.
u
*
i
( )
T
u
*
i
[ ]
i
( )
2
2 u
i 1
u
i
i 1
( ) + ( )
T
u
*
i
[ ]
i
( ) + +
u
i 1
u
i
i 1
( ) + ( )
T
u
i 1
u
i
i 1
( ) + ( ) l
2
[ ] 0 =
i 1 = u
1
0 =
i
( )
1

i
( )
2
u
i 1
u
i
u
i
u
i
( )
1
u
i
( )
2

i
( )
1
i
( )
2

1

2
1
cos
u
n 1 +
i
( )
1
[ ]
T
u
n 1 +
i 1
l
----------------------------------------------------- =
2
cos
u
n 1 +
i
( )
2
[ ]
T
u
n 1 +
i 1
l
----------------------------------------------------- =
i 1 = u
n 1 +
0
u
n
=
i
( )
1

i
( )
2
cos
i u
i 1
u
i
+ ( ) u
i 1
u
i
+ ( ) l
2
2 u
i 1
u
i
+ ( )u
*
i
-------------------------------------------------------------------------------------- =
719
CHAPTER 11
Figure 11-7 Crisfields Constant Arc Length
The constraints in Equations (11-26) and (11-27) are imposed at every iteration. Disadvantage of the quadratic
equation suggested by Crisfield is the introduction of an equation with two roots and thus the need for an extra equation
to solve the system for the calculated roots if two real roots exists. This situation arises when the contribution (or
) is very large in comparison to the arc-length. This can be avoided in most cases by setting sufficiently small
values of the error tolerance on the residual force. In case the above situation still persists despite the reduction of error
tolerance, Marc has two options to proceed:
a. To attempt to continue the analysis with the load increment used in the initial step of auto increment process.
b. Use the increment resulting from the linear constraint for the load.
There are two alternate methods using the Crisfield method for selecting the roots:
Singularity ratio method; where if the system of equations is positive definite, the largest root is used; while if
the system was negative, the smallest root is used.
Falzons method described in Reference 7 is applicable for problems which have multiple nearly equal
buckling modes.
This is circumvented in Ramms procedure due to the linearization.
Another approach to impose the constraint is due to Ramm, who also makes use of a quadratic equation to impose the
constraint giving rise to the Riks-Ramm method. The difference is that while Crisfield imposes the constraint as a
quadratic equation, Ramm linearized the constraint.
Geometrically, the difference between the two methods is that the Crisfield method enforces the correction on the
curve of the augmented equation introducing no residual for the augmented equation. Ramm takes the intersection
between the linearizations of the curves which gives a residual of the augmented equation for the next step. Both
methods converge to the same solution, the intersection of the two curves, unless approximations are made.
The Riks-Ramm constraint is linear, in that:
Displacement (u)
Load (F)
1
F
F
2
=
2
F
r
1
u
2
(
1
) = K
2
-1
(
1
F - r
1
)
B
u
1
u
1
c l
2
u
n
u
n 1 +
= =

which results in a linear equation for :
Thus, the load parameter predictor is calculated as:
(11-31)
while during the corrector phase it is:
(11-32)
It is noted that in the definition of the constraint, the normalized displacement of the previous step is used for the
normal to the auxiliary surface . Thus, problems can arise if the step size is too big. In situations with sharp
curvatures in the solution path, the normal to the prediction may not find intersections with the equilibrium curve. Note
that the norm of the displacement increment during the iterations is not constant in Riks-Ramm method.
In contact problems, sudden changes of the stiffness can be present (due to two bodies which are initially not in contact
suddenly make contact). Hence, a potential problem exists in the Riks-Ramm method if the inner-product of the
displacement due to the load vector and the displacement increment is small. This could result in a very
large value of the load increment for which convergence in the subsequent iterations is difficult to achieve. Therefore,
a modified predictor can be used resulting in a modified Riks-Ramm procedure as:
(11-33)
where
(11-34)
This method effectively scales the load increment to be applied in the prediction and is found to be effective for contact
problems.
Refinements and Controls
The success of the methods depend on the suitable choice of the arc-length:
u
n
T
u
i
u
*
i
+ ( ) l
2
=
n 1 +
1 u
n
l u
n
( )
T
K
i
( )
1
r
i
[ ]
u
n
T
u
*
1
( )
---------------------------------------------------------------------- =
n 1 +
i

u
n 1 +
i
( )
T
K
i
( )
1
r
i
[ ]
u
n 1 +
i
( )
T
u
*
i
( )
---------------------------------------------------- =
c
u
------ n =
u
*
i
u
n
1
l
n 1
u
*
1
u
*
1
[ ]
T
u
1
u
*
1
[ ]u
*
1
------------------------------------------------------------------ =
u
n
T
u
*
i
u
n
T
u
*
i
-------------------------- =
C l
2
=
721
CHAPTER 11
The initial value of the arc-length is calculated from the initial fraction of the load specified by you in the following
fashion:
(11-35)
(11-36)
In subsequent steps the arc-length can be reduced or increased at the start of a new load step depending on the number
of iterations in the previous step. This number of iterations in compared with the desired number of iterations
which is typically set to 3 or 5. The new arc-length is then given by:
(11-37)
Often, this results in too large of an arc-length growth, especially initially when the system is nearly linear and is
small (often one). In this case, one can gradually grow the value as:
Two control parameters exist to limit the maximum enlargement or the minimum reduction in the arc-length.
(11-38)
The min and max defined here are the 7th and 5th field of the 2nd data block. It is also possible that the maximum
fraction of the load allowed can vary during the loadcase. This is often useful if one knows that one is approaching the
limit load.
In addition, the maximum fraction of the total load may be defined. In general, control on the limiting values with
respect to the arc-length multiplier is preferred in comparison with the maximum fraction of the load to be applied in
the iteration since a solution is sought for a particular value of the arc-length.
Also, attention must be paid to the following:
1. In order to tract snap-through problems, the method of allowing solution if the stiffness matrix becomes
nonpositive needs to be set.
2. The maximum number of iterations must be set larger than the desired number of iterations.
if
if
if
if
Ku F R =
l
i ni
2
u =
I
0
I
d
rt I
d
I
o
=
l
new
2
rt l
pr ev
2
=
I
o
l
new
sf l
prev
=
sf .1 = rt .1
sf 1 = .1 rt 1 < <
sf 1.25 = 1 rt 3 <
sf 1.5 = 3 rt
min
l
2
l
ini
2
------- max < <

The default procedure for convergence criterion in Marc is based on the magnitude of the maximum residual load
compared to the maximum reaction force. This method is appropriate since the residuals measure the out-of-
equilibrium force, which should be minimized. This technique is also appropriate for Newton methods, where zero-
load iterations reduce the residual load. The method has the additional benefit that convergence can be satisfied
without iteration.
The basic procedures are outlined below.
1. RESIDUAL CHECKING
(11-39)
(11-40)
(11-41)
(11-42)
Where is the force vector, and is the moment vector. and are control tolerances.
indicates the component of with the highest absolute value. Residual checking has two drawbacks. First, if
the CENTROID parameter is used, the residuals and reactions are not calculated accurately. Second, in some
special problems, such as free thermal expansion, there are no reaction forces. The program uses displacement
checking in either of these cases.
2. DISPLACEMENT CHECKING
(11-43)
(11-44)
(11-45)
(11-46)
where is the displacement increment vector, is the correction to incremental displacement vector,
is the correction to incremental rotation vector, and is the rotation iteration vector. With this method,
convergence is satisfied if the maximum displacement of the last iteration is small compared to the actual
displacement change of the increment. A disadvantage of this approach is that it results in at least one iteration,
regardless of the accuracy of the solution.
F
r esi dual

F
react i on

------------------------------- TOL
1
<
F
r esi dual

F
react i on

------------------------------- TOL
1
and <
M
r esi dual

M
react i on

--------------------------------- TOL
2
<
F
r esi dual

TOL
1
<
F
r esi dual

TOL
1
and M
r esi dual

TOL
2
< <
F M TOL
1
TOL
2
F
F
u
---------------- TOL
1
<
u
---------------- TOL
1
and <

---------------- TOL
2
<
u
TOL
1
<
u
TOL
1
and
TOL
2
< <
u u
723
CHAPTER 11
11
Figure 11-8 Displacement Control
3. STRAIN ENERGY CHECKING
This is similar to displacement testing where a comparison is made between the strain energy of the latest
iteration and the strain energy of the increment. With this method, the entire model is checked.
(11-47)
where is the strain energy of the increment and is the correction to incremental strain energy of the
iteration. These energies are the total energies, integrated over the whole volume. A disadvantage of this
approach is that it results in at least one iteration, regardless of the accuracy of the solution. The advantage of
this method is that it evaluates the global accuracy as opposed to the local accuracy associated with a
single node.
Different problems require different schemes to detect the convergence efficiently and accurately. To do this,
the following combinations of residual checking and displacement checking are also available.
4. RESIDUAL OR DISPLACEMENT CHECKING
This procedure does convergence checking on both residuals (Procedure 1) and displacements (Procedure 2).
Convergence is obtained if one converges.
5. RESIDUAL AND DISPLACEMENT CHECKING
This procedure does a convergence check on both residuals and displacements (Procedure 4). Convergence is
achieved if both criteria converge simultaneously.
In several types of analyses, maximum reactions or displacements are extremely small (even close to the round-off
errors of computers). In such circumstances, not all types of relative convergence criteria may work properly. For
example, in a problem with stress-free motion, the convergence check based on relative displacement increments
works correctly but not the convergence check based on relative residual or strain energy. In this situation, it is
necessary to check the convergence with absolute values of reactions or strain energy; otherwise, the analysis may
terminate prematurely. Similarly, this kind of situation may happen for problems with springback and free thermal
expansion or constraint thermal expansion. The details for the cases where convergence checking with relative values
may encounter difficulties are listed in Table 11-2. The AUTO SWITCH option is designed for Marc to switch to the
proper convergence check scheme automatically if any of the situations mentioned above occur during the analysis.
The AUTO SWITCH option allows Marc to automatically switch the convergence check scheme to check as required
on either residuals or displacements if small reactions or displacements are detected, or to use the absolute strain
energy checking if necessary.
F
u
0
Correction to incremental displacements of i
th
iteration
i
Displacements at increment n u
n
j
j 0 =
i
----------------
Tolerance
u
n 1 +
0
u
n 1 +
k 1 +
E
E
------- TOL
1
<
E E

If AUTO SWITCH is turned on, it:
1. Switches on the relative residual checking if the relative displacement criterion is used (which fails when the
maximum incremental displacement becomes very small
Max._incremental_displacement/Smallest_element_size<1.0e-6)
2. Switches on the relative displacement checking if the relative residual force (moment) criterion is used (which
fails when the maximum reaction force becomes very small <1.0e-8)
3. To switch on the absolute energy checking if the structure is free of stress and deformation (strain energy
density < 1.0e-15).
Singularity Ratio
The singularity ratio, , is a measure of the conditioning of the system of linear equations. is related to the
conditioning number, , which is defined as the ratio between the highest and lowest eigenvalues in the system. The
singularity ratio is an upper bound for the inverse of the matrix conditioning number.
(11-48)
and establish the growth of errors in the solution process. If the errors on the right-hand side of the equation are
less than prior to the solution, the errors in the solution will be less than , with
(11-49)
Table 11-2 Effectiveness of various Relative Tolerance Convergence Testing Criterion
Analysis Type
Convergence Variable
Displacement/
Rotation
Residual
Force/Torque
Strain Energy
Stress-free motion Yes No No
Springback No Yes No
Free Thermal Expansion Yes No No
Constraint Thermal Expansion No Yes Yes
Yes relative tolerance testing works.
No relative tolerance testing does not work.
Notes:
1. If any criterion with absolute value testing is being used as convergence checking, this feature is
deactivated.
2. If the combined convergence check scheme with relative displacements or residuals is chosen as
convergence criterion, it is not necessary to turn the AUTO SWITCH on.
3. Once the extreme cases disappear in the analysis, Marc recovers to use the convergence criteria
given by user.
R R
C
1 R C
C R
E
CE
725
CHAPTER 11
The singularity ratio is a measure that is computed during the Crout elimination process of Marc using the direct solver.
In this process, a recursive algorithm redefines the diagonal terms
(11-50)
where is a function of the matrix profile. is a diagonal of the k
th
degree of freedom. The singularity ratio is
defined as
(11-51)
If all and are positive, the singularity ratio indicates loss of accuracy during the Crout elimination
process. This loss of accuracy occurs for all positive definite matrices. The number of digits lost during the elimination
process is approximately equal to
(11-52)
The singularity ratio also indicates the presence of rigid body modes in the structure. In that case, the elimination
process produces zeros on the diagonal . Exact zeros never appear because of numerical error; therefore, the
singularity ratio is of the order
(11-53)
where is the accuracy of floating-point numbers used in the calculation. For most versions of Marc, .
If rigid body modes are present, is very small or negative. If either a zero or a negative diagonal is encountered,
execution of Marc is terminated because the matrix is diagnosed as being singular.
You can force the solution of a nonpositive definite or singular matrix. In this case, Marc does not stop when it
encounters a negative or small term on the diagonal. If you use Lagrangian multiplier elements, the matrix
becomes nonpositive definite and Marc automatically disables the test on the sign of . However, it still tests on
singular behavior.
Note:
The correctness of a solution obtained for a linearized set of equations in a nonpositive definite system is
not guaranteed.
K
kk
k ( )
K
kk
k 1 ( )
K
mk
m i =
k 1
= K
mk
1 i k 1
i K
kk
R min K
kk
k ( )
K
kk
k 1 ( )
=
K
kk
k ( )
K
kk
k 1 ( )
n
l ost
log
10
R =
K
kk
k ( )
0
R O 10
n
di gi t
\ .
| |
=
n
digit
n
digit
12 >
K
kk
k ( )
K
kk
k ( )
K
kk
k ( )

AutoSPC
The autospc capability, which can be activated either on the AUTOSPC parameter or on the SOLVER option, attempts
to eliminate rigid body motion by constraining degrees of freedom. This procedure occurs if the singularity ratio is
smaller than 1.e-11. This option is only available with the direct solution procedure. This is different than setting the
non-positive definite flag, which puts a small number into the stiffness matrix.
Solution of Linear Equations
The finite element formulation leads to a set of linear equations. The solution is obtained through numerically inverting
the system. Because of the wide range of problems encountered with Marc, there are several solution
procedures available.
Most analyses result in a system which is real, symmetric, and positive definite. While this is true for linear structural
problems, assuming adequate boundary conditions, it is not true for all analyses. Marc has two main modes of solvers
direct and iterative. Each of these modes has two families of solvers, based upon the storage procedure. While all of
these solvers can be used if there is adequate memory, only a subset uses spill logic for an out-of-core solution. Finally,
there are classifications based upon nonsymmetric and complex systems. This is summarized in the following table:
Table 11-3 Solver Behavior
Direct
Profile
Iterative
Sparse
Direct
Sparse
Vendor
Provided
Sparse*
Multifrontal
Sparse
Solver Option 0 2 4 6 8
Real Symmetric Yes Yes Yes Yes Yes
Real Nonsymmetric Yes No No No Yes
Complex Symmetric Yes No No No Yes
Complex nonsymmetric No Yes No No Yes
Out-of-core Yes Yes Yes SGI only Yes
Possible problem with poorly conditioned systems No Yes No No No
Solver Parallelization No No No Yes Yes**
Memory Architecture N/A N/A N/A Shared Shared
Available with DDM Yes Yes Yes Yes Yes
CASI Iterative
Mixed
Direct
Iterative
Pardiso
Direct
Sparse** MUMPS Direct**
Solver Option 9 10 11 12
Real Symmetric Yes Yes Yes Yes
Real Nonsymmetric No No Yes Yes
*Available for SGI platforms only.
**Available for Win32, Win64, and Linux 64 platforms only.
727
CHAPTER 11
The choice of the solution procedure is made through the SOLVER option.
Direct Methods
Traditionally, the solution of a system of linear equations was accomplished using direct solution procedures, such as
Cholesky decomposition and the Crout reduction method. These methods are usually reliable, in that they give
accurate results for virtually all problems at a predictable cost. For positive definite systems, there are no
computational difficulties. For poorly conditioned systems, however, the results can degenerate but the cost remains
the same. The problem with these direct methods is that a large amount of memory (or disk space) is required, and the
computational costs become very large.
This version of MUMPS is based on public domain software developed during the Esprit IVEuropean project
PARASOL (1996-1999) by CERFACS, ENSEEIHT-IRIT and RAL. Since this first public domain version in 1999,
the developments are supported by the following institutions: CERFACS, ENSEEIHT-IRIT, and INRIA. [Refs. 8., 9.,
and 10.]
Solver 11 is the preferred direct solver on Windows and Linux platforms; otherwise, solver 8 is recommended. If a
multiprocessor machine is available, then Solver 11 or Solver 12 is recommended.
Iterative Methods
Marc offers iterative solvers as a viable alternative for the solution of large systems. These iterative methods are based
on preconditioned conjugate gradient methods. The single biggest advantage of these iterative methods is that they
allow the solution of very large systems at a reduced computational cost. This is true regardless of the hardware
configuration. The disadvantage of these methods is that the solution time is dependent not only upon the size of the
problem, but also the numerical conditioning of the system. A poorly conditioned system leads to slow convergence
hence increased computation costs.
Complex Symmetric No No Yes No
Complex nonsymmetric No No Yes Yes
Out-of-core Yes No Yes No
Possible problem with poorly conditioned systems No No Yes Yes
Solver Parallelization Pre-release No Yes Yes
Memory Architecture Shared or
Distributed
N/A Shared Shared or
Distributed
Available with DDM Pre-release No Yes Yes
CASI Iterative
Mixed
Direct
Iterative
Pardiso
Direct
Sparse** MUMPS Direct**
*Available for SGI platforms only.
**Available for Win32, Win64, and Linux 64 platforms only.

When discussing iterative solvers, two related concepts are introduced: fractal dimension, and conditioning number.
Both are mathematical concepts, although the fractal dimension is a simpler physical concept. The fractal dimension,
the range of which is between 1 and 3, is a measure of the chunkiness of the system. For instance, a beam has a
fractal dimension of 1, while a cube has a fractal dimension of 3.
The conditioning number is related to the ratio of the lowest to the highest eigenvalues of the system. This number is
also related to the singularity ratio, which has been traditionally reported in the Marc output when using a direct
solution procedure. In problems involving beams or shells, the conditioning number is typically small, because of the
large differences between the membrane and bending stiffnesses.
The CASI iterative solver is preferred.
Mixed Direct-Iterative Solver
The mixed direct iterative solver attempts to combine the advantages of both solution methods. In the first iteration of
an increment, the Multifrontal Sparse Solver (type 8) is used to obtain a solution while simultaneously creating the
Cholesky pre-conditioner. In subsequent iterations, the sparse iterative solver is used with the previously calculated
preconditioner. If solver convergence is not being obtained at a fast enough rate, the program switches back to the
direct solver. This procedure has been shown to be effective for contact problems where small deformation or inelastic
problems occur. This solver is not supported with DDM.
Preconditioners
The choice of preconditioner can substantially improve the conditioning of the system, which in turn reduces the
number of iterations required. While all positive definite systems with degrees of freedom converges in
iterations, a well conditioned system typically converges in less than the square root of iterations.
The available preconditioners in the sparse iterative solver are diagonal, scaled diagonal, and incomplete Cholesky.
The available preconditioners in the CASI sparse iterative solver are: CASI Primal and CASI Standard.
The iterative solvers require an error criteria to determine when convergence occurs. The default is to use an error
criteria based upon the ratio between the residuals in the solution and the reaction force. After obtaining the solution
of the linear equations evaluate:
(11-54)
The residual from the solution procedure is:
(11-55)
If the system is linear ( does not change) and exact numerics are preformed, then .
Because this is an iterative method the residual is nonzero, but reduces in size with further iterations. Convergence is
obtained when
(11-56)
The tolerance is specified through the SOLVER option.
N N
N
u
c
Ku
C
F
C
=
Res F
A
F
C
F
A
Ku
C
= =
K Res 0 =
Res Reac TOL <
729
CHAPTER 11
Storage Methods
In general, a system of linear equations with N unknowns is represented by a matrix of size N by , or variables.
Fortunately, in finite element or finite difference analyses, the system is banded and not all of the entries need to be
stored. This substantially reduces the memory (storage) requirements as well as the computational costs.
In the finite element method, additional zeroes often exist in the system, which results in a partially full bandwidth.
This profile storage method is used in Marc to store the stiffness matrix when solver type 0 is chosen. When many
zeroes exist within the bandwidth, the sparse storage methods can be quite advantageous. Such techniques do not store
the zeroes, but require additional memory to store the locations of the nonzero values. You can determine the sparsity
of the system (before decomposition) by examining the statements:
Number of nodal entries excluding fill in x
Number of nodal entries including fill in y
If the ratio ( ) is large, then the sparse matrix storage procedure is advantageous.
Nonsymmetric Systems
The following analyses types result in nonsymmetric systems of equations:
Inclusion of convective terms in heat transfer analysis
Coriolis effects in transient dynamic analysis
Fluid mechanics
Steady state rolling
Soil analysis
Follower force stiffness
Frictional contact
The first four always result in a nonsymmetric system. The last three can be solved either fully using the nonsymmetric
solver, or (approximately) using a symmetric solver. The nonsymmetric problem uses approximately twice as much
memory for storing the stiffness matrix.
Complex Systems
Marc utilizes a complex operator matrix for dynamic harmonic analyses when the damping matrix is present or for
harmonic magnetodynamic, acoustic, or piezoelectric analyses. The matrix is always symmetric.
Iterative Solvers
In Marc, an iterative sparse solver is available using a sparse matrix technique. This method is advantageous for
different classes of problems.
There exist certain types of analyses for which the sparse iterative solver, CASI iterative or Mixed direct-iterative
solvers are not appropriate. These types include:
elastic analysis
explicit creep analysis
complex harmonic analysis
superelements/substructures
N N
2
x y

central difference techniques
eigenvalue analysis
use of gap elements
auto increment loadcases
If the CASI solver is used with an option that is not supported, the program will switch to the direct solver type 8.
Elastic or explicit creep analysis involves repeated solutions using different load vectors. When a direct solver is used,
this is performed very efficiently using back substitution. However, when an iterative solver is used, the stiffness
matrix is never inverted, and the solution associated with a new load vector requires a complete re-solution.
The sparse iterative solver can exhibit poor convergence when shell elements or Herrmann incompressible elements
are present.
The CASI iterative solver should not be used with higher order Herrmann elements which includes element types 155,
156, and 157.
If the mixed direct-iterative solver is used and convergence is poor, Marc switches to the multifrontal sparse direct
solver.
Parallel Direct Solvers
The solution of the linear equations may be solved using multi-processors using the hardware provided solver, the
multifrontal solver, the Pardiso solver, or the MUMPS solver depending on the hardware environment as shown in
Table 11-3. For the case of the hardware, multifrontal and Pardiso solver, the number of processors used is controlled
by the -nthread ntx command when submitting the job. Here, ntx is the number of threads. For a shared memory
machine with a single chip but with four cores, one would use -nthread 4 to reduce the amount of wall time
consumed. When using a distributed memory machine or cluster, one would use the MUMPS solver and the -
nsolver nsx control line argument. Where nsx is the number of processors/threads to be used. If network computing
is used, then a host file is required as well.
Basic Theory
A linear finite element system is expressed as:
(11-57)
And a nonlinear system is expressed as:
(11-58)
where is the elastic stiffness matrix, is the tangent stiffness matrix in a nonlinear system, is the
displacement vector, is the applied load vector, and is the residual.
The linearized system is converted to a minimization problem expressed as:
(11-59)
Ku F =
K
T
u F R r = =
K K
T
u
F r
u ( ) 1 2u
T
Ku u
T
F =
731
CHAPTER 11
For linear structural problems, this process can be considered as the minimization of the potential energy. The
minimum is achieved when
(11-60)
The function decreases most rapidly in the direction of the negative gradient.
(11-61)
The objective of the iterative techniques is to minimize function, , without inverting the stiffness matrix. In the
simplest methods,
(11-62)
where
(11-63)
The problem is that the gradient directions are too close, which results in poor convergence.
An improved method led to the conjugate gradient method, in which
(11-64)
(11-65)
The trick is to choose to be conjugate to . Hence, the name conjugate gradient methods. Note
the elegance of these methods is that the solution may be obtained through a series of matrix multiplications and the
stiffness matrix never needs to be inverted.
Certain problems which are ill-conditioned can lead to poor convergence. The introduction of a preconditioner has
been shown to improve convergence. The next key step is to choose an appropriate preconditioner which is both
effective as well as computationally efficient. The easiest is to use the diagonal of the stiffness matrix. The incomplete
Cholesky method has been shown to be very effective in reducing the number of required iterations.
Flow Diagram
Figure 11-9 is a diagram showing the flow sequence of Marc. This diagram shows the input phase, equivalent nodal
load vector calculation, matrix assembly, matrix solution, stress recovery, and output phase. It also indicates load
incrementation and iteration within a load increment.
u K
1
F =
u ( ) F Ku r = =
u
k 1 +
u
k

k
r
k
+ =
k
r
k
T
r
k
r
k
T
Kr
k
=
u
k 1 +
u
k

k
P
k
+ =
k
P
k
T
r
k 1
P
k
T
KP
k
=
P
k
K P
1
P
2
P
k 1
, , ,

Figure 11-9 Marc Flow Diagram
Remarks
Which solution method to use depends very much on the problem. In some cases, one method can be advantageous
over another; in other cases, the converse might be true.
Whether a solution is obtainable or not with a given method, usually depends on the character of the system of
equations being solved, especially on the kind on nonlinearities that are involved.
As an example in problems which are linear until buckling occurs, due to a sudden development of nonlinearity, it is
necessary for you to guide the arc-length algorithm by making sure that the arc length remains sufficiently small prior
to the occurrence of buckling.
Even if a solution is obtainable, there is always the issue of efficiency. The pros and cons of each solution procedure,
in terms of matrix operations and storage requirements have been discussed in the previous sections. A very important
variable regarding overall efficiency is the size of the problem. The time required to assemble a stiffness matrix, as
No
No
Yes
Yes
Equivalent Nodal
Load Vector
Matrix Assembly
Input Phase:
Read Input Data
Space Allocation
Data Check
Matrix Solution
Stress Recovery
Convergence
Output Phase
Adapt Mesh
Next
Increment
Stop
Incremental
Loads
T
i
m
e

S
t
e
p

L
o
o
p
I
t
e
r
a
t
i
o
n

L
o
o
p
733
CHAPTER 11
well as the time required to recover stresses after a solution, vary roughly linearly with the number of degrees of
freedom of the problem. On the other hand, the time required to go through the direct solver varies roughly
quadratically with the bandwidth, as well as linearly with the number of degrees of freedom.
In small problems, where the time spent in the solver is negligible, you can easily wipe out any solver gains, or even
of assembly gains, with solution procedures such as a line search which requires a double stress recovery. Also, for
problems with strong material or contact nonlinearities, gains obtained in assembly in modified Newton-Raphson can
be nullified by increased number of iterations or nonconvergence.
The development of new solution procedures is still an active field of research in the academic community.
References
1. Bergan, P. G. and E. Mollestad, An Automatic Time-Stepping Algorithm for Dynamic Problems, Computer
Methods in Applied Mechanics and Engineering, 49, (1985), pp. 299 - 318
2. Bathe, K. J. Finite Element Procedures, Prentice Hall, Englewood Cliffs, NJ, 1996.
3. Riks, E. An incremental approach to the solution of solution and buckling problems, Int. J. of Solids and
Structures, V. 15, 1979.
4. Riks, E. Some Computational Aspects of the Stability Analysis of Nonlinear Structures, Comp. Methods in
Appl. Mech. and Eng., 47, 1984.
5. Crisfield, M. A. A fast incremental iterative procedure that handles snapthrough, Comput. & Structures, V.
13, 1981.
6. Ramm, E. Strategies for tracing the nonlinear response near limit points, in K. J. Bathe et al (eds), Europe-
US Workshop on Nonlinear Finite Element Analysis in Structural Mechanics, Ruhr University Bochum,
Germany, Springer-Verlag, Berlin, pp/ 63-89. Berlin, 1985.
7. Cerini, M. and Falzon, B. G., Use of Arc-Length Method for Capturing Mode Jumping in Postbuckling
Aerostructures, AIAA Journal, V43,3,(2005), pp 681-689.
8. P. R. Amestoy, I. S. Duff and J.-Y. L'Excellent, Multifrontal parallel distributed symmetric and unsymmetric
solvers, in Comput. Methods in Appl. Mech. Eng., 184, 501-520 (2000).
9. P. R. Amestoy, I. S. Duff, J. Koster and J.-Y. L'Excellent, A fully asynchronous multifrontal solver using
distributed dynamic scheduling, SIAM Journal of Matrix Analysis and Applications, ol 23, No 1, pp 15-41
(2001).
10. P. R. Amestoy and A. Guermouche and J.-Y. L'Excellent and S. Pralet, Hybrid scheduling for the parallel
solution of linear systems. Parallel Computing Vol 32 (2), pp 136-156 (2006).
Chapter 12 Output Results
12
Output Results
Introduction 735
Workspace Information 735
Increment Information 738
Selective Printout 740
Restart 742
Element Information 742
Nodal Information 747
Post File 749
Forming Limit Parameter (FLP) 750
Program Messages 753
Marc HyperMesh Results Interface 754
Marc SDRC I-DEAS Results Interface 754
Marc - ADAMS Results Interface 755
Status File 757
Element Group Information 757
References 758
735
CHAPTER 12
Output Results
Introduction
This chapter summarizes the information that Marc provides in the output. In addition to reviewing your input, Marc
provides information about the procedures the program uses and the workspace allocation. All calculated results are
automatically written to the output file unless the user specifically requests otherwise.
Workspace Information
Marc reports several aspects of workspace information, specifically, the allocation of memory workspace and the size
of the work files.
For the general memory, Marc first gives the workspace needed for the input and stiffness assembly. This number tells
how much memory is needed to store the user-supplied data, the program-calculated data, and two-element stiffness
matrices. Each set of data comprises three parts: overhead, element information, and nodal information. Element
information consists of properties, geometries, strains, and stresses, and nodal information consists of coordinates,
displacements, and applied forces.
Next, Marc specifies the internal core allocation parameters. These values are useful in user subroutines and are often
provided to the user subroutines. The values provided here are:
number of degrees of freedom per node (ndeg)
maximum number of coordinates per node
maximum number of invariants per integration point (neqst)
maximum number of nodes per element (nnodmx)
maximum number of stress components per integration point (nstrmx)
number of strain components per integration point (ngens)
Note the following:
The number of degrees of freedom per node indicates the number of boundary conditions necessary to
eliminate rigid body modes.
The number of stresses per integration point is the number of stresses per layer multiplied by the number of
layers.
The number of invariants per integration point is the number of layers for either shell or beam elements.
The ELSTO parameter stores element information on an auxiliary storage device, rather than in main memory. You
can invoke this option, or Marc invokes it automatically. In this option, Marc:
sets a flag for element storage (IELSTO)
specifies how many words are used per element (NELSTO)
specifies the number of elements per buffer (MXELS)
specifies the total amount of space needed to store this information
The number of elements per buffer should be at least two, and you can change the buffer size in the ELSTO parameter
option to increase the number of elements which will be in the buffer.
Marc reports the memory for different parts as it is allocated.

For contact bodies:
allocated workspace of 4567 words of memory for body 1
For incremental backup:
space needed for incremental backup: 555313
For boundary conditions:
allocated 600 words of memory due to kinematic boundary conditions
We also get printouts like
total workspace needed with in-core matrix storage = 893281
This printout refers to the memory in general memory (see Marc Workspace Requirements in Chapter 2 in this
volume) and is not the total memory used.
The solution of the equations of equilibrium result in the formation of the operator matrix and the
decomposition/inversion of the matrix. In an mechanical analysis, this is the global stiffness matrix. For some solvers,
there is an intermediate stage where the matrix is restructured from a sparse format to a more optimal format for
numerical calculations. The allocation of this memory is reported in the memory summary (see below) either under
solver:first part or under the allocated separately: category. This is summarized as follows:
For solvers 0, 2, and 4, all the memory for the matrix solution is allocated within the general memory. For solver 6 and
8, the workspace for the restructuring stage (solver 6 only) and decomposition part is allocated separately outside of
the general memory.
Solvers 0, 4, and 8 have an out-of-core option for both storing the operator matrix and for inverting the matrix. Solver
6 (SGI) and solver 11 (PARDISO) has an out-of-core option for inverting the matrix only. For this solver, there must
be enough memory to assemble the operator matrix in-core.
When using solver 0, if the out-of-core solver is invoked, one observes:
total workspace needed with in-core matrix storage = 89686633
out-of-core matrix storage will be used
core allocation based on 50498 nodal entries per assembly buffer.
...file 14-- maximum record length= 40960
approximate no. of words on file= 3686400
Solver
Operator
Matrix Restructured Matrix
Decomposition or
Inverted Matrix
0 -Profile direct first part first part first part
2 - Sparse Iterative first part N/A N/A
4 - Sparse Direct first part first part first part
6 - Hardware Provided first part allocated separately allocated separately
8 - Multi-front Direct Sparse first part N/A allocated separately
10 - Mixed Direct Iterative first part N/A allocated separately
11 - Pardiso Direct first part N/A allocated separately
12 - MUMPS direct first part N/A allocated separately
737
CHAPTER 12
Output Results
If solver 4 were used, one would observe:
total workspace needed with sparse in-core matrix storage = 34288801
out-of-core matrix storage will be used
Note that until 14 contains the true sparse matrix, and, as the same problem is used, the I/O is the same for both solvers
0 and 4. For solver 0, this matrix gets expanded on unit 11 using a row blocking system, and decomposed to unit 12.
For solver 4, it gets expanded and decomposed using unit 2. Solver 4 uses substantially less I/O than solver 0.
For solver 8, we observe
solver workspace needed for out-of-core matrix storage = 6389794
solver workspace needed for in-core matrix storage = 31339042
matrix solution will be out-of-core
approximate disk space for out-of-core matrix storage = 62678084
At the end of the run, Marc prints out a summary of the memory used. This typically looks something like
memory usage: mbyte words % of total
within general memory:
element stiffness matrices: 0 64982 0.2
solver: first part 4 963934 3.6
overallocation initial allocation 49 12957302 48.1
other: 0 15713 0.1
allocated separately:
incremental backup: 5 1253317 4.7
solver 8 10 2581002 9.6
nodal vectors: 1 349427 1.3
contact: 1 261625 1.0
tyings: 0 13362 0.0
transformations: 0 25182 0.1

kinematic boundary conditions: 0 4200 0.0
element storage: 5 1324844 4.9
executable and common blocks: 27 7000000 26.0
miscellaneous 0 121763 0.5
---------------------------------------------------------------
total: 103 26936653
general memory allocated: 53 14001931
general memory used: 4 1044629
peak memory usage: 110 28936653
In a parallel analysis, Marc also prints out the memory usage summed over the domains. The memory printout, during
the execution, can also be obtained by means of the control file. Suppose the job defined in the input file
jobname.dat is running. The memory summary printout is obtained by creating a file called jobname.cnt with a
single line containing the word memory. The information is output at the beginning of the next increment.
If the ALLOCATE parameter is to be used for initial memory allocation, the value to look for is the one for the line
general memory used: 4 1044629
In this example, one would choose a value of 5 for ALLOCATE to insure that no dynamic reallocation of memory is
performed.
The top part defines the memory in general memory. The overallocation initial memory specifies the amount of
general memory specified with the ALLOCATE option that is not used. The above example is very small and there is
a large overallocation due to the default used (although the total amount is small). For larger problems, this part should
be small. The line other specifies the part of the general memory that is not categorized.
Further down are the parts that are allocated separately from the general memory, listed for each category. Similar to
other above, miscellaneous specifies memory not categorized.
The two lines for general memory is for highlighting the memory allocated and the memory actually used. The
difference between the two is reflected in the overallocation line in the table. executable and common blocks is
an estimate of the memory needed for loading the executable and storing values in common blocks. This is added so
the total better agrees with the amount of memory the operative system reports for the program. Finally, the peak
memory denotes the maximum amount of memory allocated at any instant during the job. In some cases, memory is
allocated and then released. The total: in the summary table may then show a smaller value than the maximum
during the run.
Increment Information
A variety of information is given for each increment, and often for each iteration, of the analysis. This information is
useful in determining the accuracy and the stability of the analysis.
739
CHAPTER 12
Output Results
Summary of Loads
The printout entitled Load Increments Associated with Each Degree of Freedom allows you to check the load input
quickly. It represents the sum, over all nodes in the mesh, of the point loads and the equivalent forces obtained after
distributed loads (pressures, body forces) are applied. For example, you can easily check a pressure because the total
force in a global coordinate direction would be the projected area normal to that direction of the surface upon which
the pressure is applied, multiplied by the pressure magnitude.
Timing Information
The amount of CPU necessary to reach the given location in the analysis is indicated by the following output:
start of increment
start of assembly
start of matrix solution
end of matrix solution
end of increment
As an example, subtract the time associated with the start of the matrix solution from the time associated with the end
of the matrix solution to determine how much time was spent in the equation solver.
Singularity Ratio
The singularity ratio is a measure of the conditioning of the matrix (see Chapter 11 Solution Procedures for Nonlinear
Systems for further details). This ratio is printed each time there is a solution of the matrix equations. You can measure
the influence of the nonlinearities in the structure by examining the change in singularity ratios between increments.
The singularity ratio is not available when using the iterative solvers, and when using the Pardiso and MUMPS direct
solver.
Convergence
Several messages are printed that concern the convergence of the solution. These messages indicate the displacement,
velocity, or residual error and are very important because they provide information concerning the accuracy of the
solution procedure. These messages also indicate the ratio of the error and its relative quantity. This ratio must be less
than that given in the CONTROL option for convergence to occur. See Convergence Controls in Chapter 11for details
on convergence testing.
When the iterative solver is used, additional messages are printed regarding the convergence of the solution
procedures.

Selective Printout
Marc gives you several options and user subroutines for the control of the Marc output.
Options
PRINT CHOICE allows you to select how much of the element and nodal information is to be printed. The possible
selections are
group of elements
group of nodes
which layers (form beam and shell elements)
which integration points
increment frequency between printouts
The data entered through this option remain in control until you insert a subsequent PRINT CHOICE set. You can
include such a set with either the model definition or with the history definition set. To obtain the default printout after
a previous PRINT CHOICE, invoke PRINT CHOICE using blank entries.
You can use the PRINT ELEMENT capability as a replacement for, or in conjunction with, the PRINT CHOICE option.
The enhancements this option offers over the PRINT CHOICE option are the following:
1. You have a choice of integration and layer points for each element to be printed, which is especially useful
when several different element types are used in one analysis.
2. You have a choice of the type of quantity to be printed, for instance, stresses could be printed for all elements,
but strains for only a few.
3. The types of output quantities that can be selected are stresses, strains, creep strain, thermal strain, cracking
strain, Cauchy stress, state variables, strain energy, or all nonzero quantities.
4. The output can be placed in a file other than the standard output file.
5. The PRINT NODE option can be used to obtain element quantities such as stresses and strains at the nodal points
of each element. These are obtained by extrapolating the integration point values to the nodes.
The PRINT NODE option is an alternative to the PRINT CHOICE option for controlling the output of nodal quantities.
The additional capabilities of this option compared to the PRINT CHOICE are as follows:
1. You can choose which of the following nodal quantities are to be printed:
incremental displacements
total displacements velocities
accelerations
reaction forces
generalized stresses
2. Different quantities can be printed for different nodal points.
3. The output can be placed in a file other than the standard output file.
Note:
The PRINT CHOICE option has no effect on the restart or post file.
741
CHAPTER 12
Output Results
The PRINT CONTACT and the NO PRINT CONTACT options control the output of the summary of the loads on the
contact bodies and their positions.
The SUMMARY option allows you to get a quick summary of the results obtained in the analysis. This option prints
the maximum and minimum quantities in tabular form. The is designed for direct placement into reports. You control
the increment frequency of summary information and the file unit to which the information is written.
The SUMMARY option reports on the physical components and the Tresca, von Mises, and mean values of stress,
plastic strain, and creep strain. It also reports on such nodal quantities as displacements, velocities, accelerations, and
reaction forces.
The sort options (ELEM SORT and NODE SORT) allow you to sort calculated quantities in either ascending or
descending order. These quantities may either be sorted by their real magnitude or their absolute magnitude. The sort
options also allow you to control the type of quantity to be sorted, for example, equivalent stress and the number of
items to be sorted. The sort options print the sorted values in tabular form. The is designed for direct placement into
reports. You may control the increment frequency of sorted values and the file unit to which these values are written.
The PRINT VMASS option allows you to selectively choose which elements and associated volumes, masses, first and
second moment of inertia, and strain energy are to be printed. In order to have correct mass computations, mass density
for each element must be given. The volumes, masses, and energies can be written on either standard output file unit
6, or user specified unit. The center of mass is also calculated and output.
Grid Force Balance
The GRID FORCE option allows the user to output the different contributions to the total force applied on an element
or a node.
On an element level, the grid force balance is based upon the
Internal forces
Distributed Loads
Foundation Forces
Reaction Force
On a nodal basis it is much more complete and includes
Note:
All quantities are saved on the restart file, and you can obtain them at a later time by using the RESTART
option. Quantities that were not printed out are available for later use.
Internal Forces Distributed + Point Forces
Foundation Forces Spring Forces
Contact Normal Forces Contact Friction Forces
Tying/MPC Forces Inertia Forces
Damping Forces DMIG Forces
Reaction Force

User Subroutines
User subroutines can be used to obtain additional output, which can be accessed at each time/load increment. The
available subroutines are the following:
IMPD obtains nodal quantities, such as displacements, coordinates, reaction forces, velocities, and
accelerations
ELEVAR obtains element quantities such as strains, stresses, state variables, and cracking information
ELEVEC outputs element quantities during harmonic subincrements
INTCRD obtains the integration point coordinates used for forming the stiffness matrix
Restart
One capability of the RESTART option allows you to recover the output at increments where printout was suppressed
in previous runs. This option can be used to print time/load increments for a number of consecutive increments. Under
this option, Marc does not do any analysis.
In conjunction with the RESTART option, you can use either the PRINT CHOICE or PRINT ELEMENT, or PRINT
NODE option to select a region of the model for which you want to obtain results.
Element Information
The main output from an increment includes element information followed by nodal information. The system provides
the element data at each integration point. If you use the CENTROID parameter, the system provides the element data
at the centroidal point. You can control the amount of printed output by using either the PRINT CHOICE or PRINT
ELEMENT option. All quantities are total values at the current state (at the end of the current increment), and the
physical components are printed for each tensor quantity (stress, strain, and generalized stress and strain). The
orientation of these physical components is generally in the global coordinate system; however, their orientation
depends on the element type. (Marc Volume B: Element Library explains the output for each particular element type.)
Also, the physical components may be printed with respect to a user defined preferred system.
In addition to the physical components, certain invariants are given, as follows:
1. Tresca intensity - the maximum difference of the principal values, and the measure of intensity usually required
for ASME Code standards.
von Mises intensity - defined for stress as
(12-1)
where is a stress tensor and is the deviatoric stress tensor. For beam and truss elements,
, and additional shears are zero; for plane stress elements, including plates and shells,
.
S
i j

i j
1
3
---
i j
kk
=
3
2
--- S
i j
S
i j
=
S
22

33
0 = =
33
0 =
743
CHAPTER 12
Output Results
2. von Mises intensity - calculated for strain type quantities as
(12-2)
Marc uses these measures in the plasticity and creep constitutive theories. For example, incompressible metal
creep and plasticity are based on the equivalent von Mises stress. For beam, truss, and plane stress elements,
an incompressibility assumption is made regarding the noncalculated strain components.
a. For beam and truss elements, this results in
(12-3)
b. For plane stress elements
(12-4)
c. For beam and plane stress elements taken as a whole
(12-5)
3. Mean normal intensity - calculated as
(12-6)
Equation (12-6) represents the negative hydrostatic pressure for stress quantities. For strain quantities, the
equation gives the dilatational magnitude. This measurement is important in hydrostatically dependent theories
(Mohr-Coulomb or extended von Mises materials), and for materials susceptible to void growth.
4. The principal values are calculated from the physical components. Marc solves the eigenvalue problem for the
principal values using the Jacobi transformation method. Note that this is an iterative procedure and may give
slightly different results from those obtained by solving the cubic equation exactly.
5. State variables are given at any point where they are nonzero.
Mentat Computed Element Results
These element results are computed based on the Marc output. By default, Marc always outputs the elemental results
at the integration points. In Mentat, by default, these data are extrapolated linearly to the nodal positions. The data are
then averaged with the contributions from the neighboring elements. Principle values and equivalent values are
computed based on these nodal data in Mentat. If a stress or strain tensor is output in the post file, Mentat will compute
their principal values and the equivalent items (such as, equivalent stress and strain). Note that the equivalent items
may be different from ones directly output by Marc because of the way they are computed in Mentat. For example,
the equivalent stress computed in Mentat may show a slightly different result compared to the equivalent von Mises
stress from Marc output.
2
3
---
i j
i j
=
22

33
1
2
---
11
= =
33

11

22
+ ( ) =
kk
0 =
11

22

33
+ + ( ) 3
or
11

22

33
+ + ( ) 3

It is also possible to turn off the extrapolation procedure in Mentat. This is recommended in problems of high
gradients. It can be done through the RESULTS>scalar plot SETTING>EXTRAPOLATION menu by picking
TRANSLATE.
Solid (Continuum) Elements
For solid (continuum) elements, stress, strain, and state variables are the only element quantities given. Marc prints
out stresses and total strain for each integration point. In addition, it prints out thermal, plastic, creep, and cracking
strains, if they are applicable. Note that the total strains include the thermal contribution.
Solid-shell Elements
Marc prints the solid-shell element output in the output file using the integration point identification. But in the post
file, the element output is presented in the layer identification conforms to the standard shell convention.
For non-stacked element (only one element in the thickness direction), users can request element section outputs
(curvature, moment, stretch and section force) to put in the post file (see Marc Volume C: Program Input for a detailed
description of the POST option).
The section force and moment are calculated using Equations 12-9 and 12-10, respectively. The stresses and y-
coordinate (thickness direction) are referring to element coordinate system.
The stretch and curvature are calculated as follows.
(12-7)
(12-8)
The curvatures and moments are calculated using the element mid-plane as reference.
Shell Elements
Marc prints generalized stresses and generalized total strains for each integration point. For thick shell (Type 22, 72,
75) elements interlaminar shear stresses are printed at the interface of two laminated layers if the TSHEAR parameter
is included.
The generalized stresses printed out for shell elements are
(12-9)
E
i j
i j
t op
i j
bot
+
2
---------------------------- =
i j
i j
t op
i j
bot
2t
--------------------------- =
1
t
---
i j
y d
t 2
+ t 2
}
Section Force per Unit Thickness ( )
745
CHAPTER 12
Output Results
(12-10)
The generalized strains printed are
(12-11)
(12-12)
Physical stress values are output only for the extreme layers unless you invoke either PRINT CHOICE or PRINT
ELEMENT. In addition, thermal, plastic, creep, and cracking strains are printed for values at the layers, if applicable.
Although the total strains are not output for the layers, you can calculate them using the following equations.
(12-13)
(12-14)
(12-15)
where h is the directed distance from the midsurface to the layer; are the stretches; and

are the curvatures as
printed. You can obtain these quantities by using the ELEVAR subroutine.
Beam Elements
The printout for beam elements is similar to shell elements, except that the section values are force, bending and
torsion moment, and bimoment for open section beams. For beam element type 45, interlaminar shear stresses are
printed at the interface of laminated layers if the TSHEAR parameter is used.
These values are given relative to the section axes (X, Y, Z) which are defined in the COORDINATES or the
GEOMETRY option.
Before a beam member can be designed, it is necessary to understand the section forces distribution along the axial
direction of the beam. For example, if variations of shear force and moment along axial direction are plotted, the
graphs are termed shear diagram and moment diagram, respectively. To postprocess these diagrams, post code 261
needs to be specified along with the post codes (shown in Table 12-1) which define the corresponding section forces
of the beam element.
Section moment per Unit Thickness ( )
1
t
--- y
i j
y d
t 2
+ t 2
}
, 1 2 , = E
;
St ret ch ( )
, 1 2 , =
;
Curvature ( )
11
E
11
h
11
+ =
22
E
22
h
22
+ =
12
E
12
h
12
+ =
E
i j

i j

The stresses at the section points in the cross section are only output if the PRINT CHOICE or PRINT ELEMENT option
are included. If any nonlinearity exists at the section points, they will be provided in the output file.
Gap Elements
Marc prints the gap contact force, friction forces, and the amount of slip.
Linear and Nonlinear Springs
Marc prints the spring forces.
The element temperatures are provided at the integration points. When requested, the temperature gradient and the
flux is also provided.
Joule Heating Analysis
In addition to the temperatures, the voltage, current, and the heat generated at the integration points is output.
Table 12-1 Post Codes for Beam Section Forces
Element
Type 2-D/3-D
Post Code
264
(Axial
Force)
265
(Moment xx
or Moment
ip)
266
(Moment yy
or Moment
op)
267
(Shear
xz or
Shear ip)
268
(Shear yz
or Shear
op)
269
(Torque)
270
(Bimoment)
5 2-D X X
16 2-D X X
45 2-D X X X
13 3-D X X X X X
14 3-D X X X X
25 3-D X X X X
31 3-D X X X X X X
52 3-D X X X X
76 3-D X X X X
77 3-D X X X X
78 3-D X X X X
79 3-D X X X X
98 3-D X X X X X X
747
CHAPTER 12
Output Results
The lubricant thickness, pressure, and pressure gradient are provided.
The electrical potential ( ), electric field intensity vector ( ), and electrical displacement vector/flux ( ) are
provided.
Piezoelectric or Electrostatic-Structural Analysis
In addition to the mechanical stresses and strains, the electrical field intensity vector ( ) and electrical
displacement/flux ( ) are provided.
The magnetic potential ( ), the magnetic field density ( ), and the magnetic flux density ( ) are provided. For two-
dimensional problems, only the component is given.
Magnetodynamics Analysis
The electric field intensity vector ( ), the electrical displacement vector/flux ( ), the magnetic flux density ( ), and
the current density ( ) are provided.
Acoustic Analysis
The pressure and the gradient of the pressure is provided. Note in an Acoustic-Structural analysis, the pressure is only
provided as a nodal variable.
Nodal Information
Marc also prints out the following quantities at each nodal point.
Stress Analysis
Incremental displacements - the amount of deformation that occurred in the last increment
Total displacements - the summation of the incremental displacements
Total equivalent nodal forces (distributed plus point loads) - the total force applied to the model through
distributed loads (pressures) and point loads
Reaction forces at fixed boundary conditions
Residual loads at nodal points that are not fixed by boundary conditions
V E D
E
D
A B H
A
z
E D H
J

Reaction Forces
Marc computes reaction forces based on the integration of element stresses. This is the only way to compute total
reaction forces in a nonlinear analysis. Since such integration is only exact if the stresses are known at each integration
point, the reaction forces are not printed if you use the CENTROID parameter.
In a nonlinear analysis, you should check that the reaction forces are in equilibrium with the external forces. If they
are not in equilibrium, the analysis will be inaccurate, usually due to excessively large incremental steps. In most cases,
the equilibrium is automatically ensured due to the convergence testing in Marc.
Residual Loads
The residual loads are a measure of the accuracy of the equilibrium in the system during analysis. This measure is very
important in a nonlinear analysis and should be several orders of magnitude smaller than the reaction forces.
Dynamic Analysis
In a dynamic analysis, Marc prints out the total displacement, velocity, and acceleration at each time increment.
In a heat transfer analysis, Marc prints out nodal temperatures, externally applied fluxes, and calculated nodal heat
fluxes.
Joule Heating Analysis
In Joule heating analysis, in addition to the thermal vectors, the nodal voltage and the applied and calculated currents
are provided.
Rigid-Plastic Analysis
In steady-state rigid-plastic analysis, Marc prints nodal velocities.
In a hydrodynamic bearing analysis, Marc gives the mass flux at the nodal points.
In an electrostatic analysis, Marc prints the scalar potential and the applied and calculated charge.
Note:
Each value listed above is given at each nodal point, unless you invoke the PRINT CHOICE or PRINT
NODE option. If you invoke the TRANSFORMATION option, Marc prints the nodal information relative to
the user-defined system, rather than the global coordinate system.
749
CHAPTER 12
Output Results
In a magnetostatic analysis, Marc prints the vector potential and the applied and calculated current.
In an magnetodynamic analysis, Marc prints out the vector and scalar potential, the applied and calculated current and
charge.
Piezoelectric or Electrostatic-Structural Analysis
In a piezoelectric or electrostatic-structural analysis, in addition to the structural quantities, such as displacement and
reaction force, the electric potential, and reaction charges.
Acoustic Analysis
In an acoustic analysis, Marc prints the pressure and the source.
Supplementary Information
Contact Analysis
The standard output includes, at the end of each increment, a summary of information regarding each body. This
information reports the increments rigid body velocity, the position of the center of rotation, and the total loads on the
body. These last values are obtained by adding the contact forces of all nodes in contact with the rigid body.
Deformable bodies being in equilibrium have no load reported.
Additional information can be obtained by means of the PRINT,5 parameter. In such cases, all the contact activity is
reported. Namely, every time a new node touches a surface, or separates from a surface, a corresponding message is
issued.
Contact with rigid surfaces entails an automatic transformation. Displacement increments and reactions in the
transformed coordinates - tangent and normal to the contact interface - are also reported for every node that is in
contact in the usual Marc manner.
If the EMRESIS or EMCAPAC options are invoked, a summary is output providing the electrical resistance or
capacitance for the conducting body.
Post File
The post file contains results of the analysis performed. This information can be viewed using Mentat or Patran to
observe the displaced mesh, contours of element quantities, principal quantities, time history behavior, response
gradients, design variables, etc. The post file contains the finite element mesh and any changes to the finite element
mesh due to either rezoning or adaptive mesh refinement. Basic nodal quantities (such as displacements, velocities,
acceleration, applied loads, and reaction forces) are automatically placed on the post file. It is also possible to include

user-defined nodal vectors by use of the UPOSTV user subroutine. Element quantities, such as stress, strain, plastic
strain) must be explicitly requested by using the POST model definition option. It is possible to have user-defined
element quantities placed on the post file for subsequent display by using the PLOTV user subroutine.
Forming Limit Parameter (FLP)
For continuum or shell elements, the Principal Engineering Strains can be calculated based on the true strain values.
Correspondingly, the Forming Limit Parameter can be obtained for shell/membranes according to the data of the
Forming Limit Diagram (FLD).
The principal engineering strains are calculated based on the true strain according to following relation:
(12-16)
where are the principal engineering strain components and are the principal true strain components.
The principal engineering strain values can be selected for postprocessing. In addition, if shell/membrane elements are
used, the Forming Limit Parameters (FLP) can also be selected. The FLP is defined as the ratio of the major principal
engineering strain to the maximum allowable major principal engineering strain given by the Forming Limit Diagram.
Based on this definition, it is calculated by
(12-17)
where means the major principal engineering strain, is the minor principal engineering strain. is the
forming limit corresponding to . FLD is defined by input data in model definition through material properties. By
plotting the major principal engineering strain of each integration point of an element as point ( , ) on the chart,
as shown in Figure 12-1, it is easy to find that for points below the FLD line is safe. Strain rates above the FLD line
imply failure. To consider the FLP value of these fields, one can find that in the field which is above FLD line, the
value of FLP is larger than 1. On the FLD line, FLP equals 1, and FLP is less than 1 if a point located in the field below
the FLD line. Therefore, by plotting the FLP value in Mentat, it can be seen if the sheet metal forming process is
successful.
e
i

i
( ) 1 exp =
e
i

i
FLP e
1
FLD e
2
( ) =
e
1
e
2
FLD e
2
( )
e
2
e
2
e
1
751
CHAPTER 12
Output Results
Figure 12-1 The Definition of Forming Limit Parameter (FLP)
The methods employed in Marc to define the Forming Limit Diagram are:
1. Fitted function definition
2. Predicted function definition
3. Table definition (accepts piecewise linear FLD curves)
The details of the above methods are described below:
1. Fitted function:
By this method, the polynomial functions are utilized to fit the FLD curve. The functions are shown in
Equations 12-18a and 12-19b:
(12-18)
(12-19)
2. Predicted function:
The predicted functions are generated based on the theories of local necking and diffuse necking. Both theories
assume that the material obeys the power-law strain hardening, .
Local necking: The critical strain for the onset of local necking is influenced by strain ratio , where
, . Equation shows the relationship of critical strain versus strain ratio (by Hill).
(12-20)
For uniaxial tension, . For plane strain, .
Diffuse necking: Diffuse necking is the phenomenon that while necking happens, deformation continues. The
critical strain for the diffuse necking to happen is determined by Equation 12-21 (by Swift):
Safe
Fail
Fail
Safe
Major Principal Engineering Strain (e
1
)
Minor Principal Engineering Strain (e
2
)
FLP<1
FLP>1
FLD(e
2
)
FLD(e
2
)
FLD e
2
( ) C
0
D
1
e
2
D
2
e
2
2
D
3
e
2
3
D
4
e
2
4
+ + + + =
e
2
0 ( )
FLD e
2
( ) C
0
C
1
e
2
C
2
e
2
2
C
3
e
2
3
C
4
e
2
4
+ + + + =
e
2
0 ( )
K
n
=

2

1
= 0
*
n
1 + ( )
------------------ = 0
0.5 = 0 =

(12-21)
Experiments (mainly with low carbon steels) showed that the predicted FLD usually is at a level below the
experimental FLD curve. The experimental forming limit diagram is characterized by the value of
corresponding to plane strain. This value is referred at as shown in Figure 12-2. This value increases
with the strain-hardening exponent, , and the strain-rate exponent, . The value of is observed to rise as
the thickness increases. This phenomenon is referred as thickness effect and it is characterized as thickness
coefficient . Experiments by Keeler tended to express this relationship as shown in Equation (12-22).
(12-22)
where if is less than 0.21. Otherwise, . is the thickness of the sheet metal. The
thickness coefficient is set as 3.59 if the unit used to define the thickness is Inch. If the unit used is mm,
is set as 0.141. Similarly, if the unit cm is used, is then set as 1.41, etc.
It should be emphasized that, in most cases, friction is also a very important parameter that will affect the
occurrence of necking. On the other hand, Equation (12-22) is mainly from data on low-carbon steels that have
relatively high strain-rate dependence. In materials with low strain-rate dependence (for example, aluminum
alloys), the thickness effect should be much less.
Finally, it is suggested that for safety purposes, a safety zone, as shown as the shadow area in Figure 12-2, must
be set. For points above the safety zone, material is considered as failure. For points below the safety zone,
material is considered as safe. If a point locates in the safety zone, the material is in the marginal status. Usually,
the thickness of the safety zone is set as 0.1 or 10%.
3. Table definition:
The table function in Marc allows users to define any curves through the TABLE model definition option. For
example, if the user has FLD point of material, it is possible to define the FLD as piecewise linear curve. For
details, refer the Marc Volume C: Program Input, Chapter 3 Model Definition Options, TABLE.
Note:
For the case of , the predicted critical strain by Equation 12-21 is well below the experiment
data. Therefore, for the purpose of prediction, the diffuse necking theory and the local necking
theory are chosen for cases of and , respectively.
*
2n 1
2
+ + ( )
1 + ( ) 2
2
2 + ( )
------------------------------------------------------ = 0
0 <
0 > 0
*
FLD
0
n m *
t
c
FLD
0
Q 0.233 t
c
t + ( ) =
Q n 0.21 = n Q 1.0 = T
t
c
t
c
t
c
753
CHAPTER 12
Output Results
Figure 12-2 Predicted FLD and Experimental FLD Curves
Program Messages
The messages provided by Marc at various points in the output show the current status of the problem solution. Several
of these messages are listed below.
START OF INCREMENT x indicates the start of increment number x.
START OF ASSEMBLY indicates Marc is about to enter the stiffness matrix assembly.
START OF MATRIX SOLUTION indicates the start of the solution of the linear system.
SINGULARITY RATIO prints yyy, where yyy is an indication of the conditioning of the matrix. This value is
typically in the range 10
-6
to 1. If yyy is of the order of machine accuracy (10
-6
for most machines), the
equations might be considered singular and the solution unreliable.
END OF MATRIX SOLUTION indicates the end of matrix decomposition.
END OF INCREMENT x indicates the completion of increment number x.
RESTART DATA at INCREMENT x ON UNIT 8 indicates that the restart data for increment number x has
been written (saved) on Unit 8.
POST DATA at INCREMENT x on UNIT y indicates that the post data for increment number x has been
written (saved) on Unit y.
In addition to these Marc messages, exit messages indicate normal and abnormal exists from Marc. Table 12-2 shows
the most common exit messages.

Experimental
forming limit
Local necking Diffuse necking
0
Minor Principal Engineering Strain e
2
Major Principal Engineering Strain e
1
FLD
0

Marc HyperMesh Results Interface
Marc now outputs results for postprocessing with HyperMesh. The writing of the HyperMesh results file is invoked
by the HYPERMESH model definition option described in detail in Marc Volume C: Program Input. Mentat can also
be used to write the necessary commands into the Marc data file. The related button can be found one level below each
of the ANALYSIS CLASS buttons within the JOBS menu.The binary file created as a result has the title jobid.hmr
where jobid is the name of the data file for the job.
The geometry data may be imported into HyperMesh from a Marc data file. Alternatively, it can be created by
HyperMesh or read in from other formats via the HyperMesh import capability.
The types of data that may be selected for writing into the HyperMesh results file are listed in Marc Volume C:
Program Input under the HYPERMESH model definition option. It is important to note that, in case eigenvectors for
buckling or eigenfrequency analysis are to be written into the results file, the corresponding Marc file should have the
BUCKLE INCREMENT or MODAL INCREMENT model definition option, as appropriate. The BUCKLE, MODAL
SHAPE, and RECOVER history definition options are not to be used in these cases.
Since HyperMesh allows only one deformed shape plot per simulation, each eigenvector of an eigenvalue analysis is
saved as a separate simulation. Thus, when using HyperMesh, these eigenvectors can be plotted by skipping to the
next simulation rather than skipping to the next data type of a simulation. The contour plots can be obtained for all
data types including eigenvectors.
In case the number of requested eigenvalues is more than the number extracted, the data type in the HyperMesh
deformed screen will inform you for those modes that are not extracted. The next button may need to be clicked
to see the data type in the deformed mode of plotting.
Marc SDRC I-DEAS Results Interface
A facility exists in Marc to write out an SDRC I-DEAS universal file of the analysis results, in order for postprocessing
by the I-DEAS program. The writing of the universal file is invoked by the SDRC model definition option described
in detail in Marc Volume C: Program Input. Mentat can also be used to write the necessary commands into the Marc
data file. The related button can be found one level below each of the ANALYSIS CLASS buttons within the JOBS
menu. The formatted ASCII file created as a result has the title jobid.unv, where jobid is the name of the data file
for the job.
The model data is imported into I-DEAS by way of the universal file as well.
Table 12-2 Marc Exit Messages
Message Meaning
MARC EXIT 3001 Normal exit.
MARC EXIT 3004 Normal exit.
MARC EXIT 13 Input data errors were detected by the program.
MARC EXIT 2004 Operator matrix (for example, stiffness matrix in stress analysis) has
become non-positive definite and the analysis terminated.
MARC EXIT 3002 Convergence has not occurred within the allowable number of recycles.
755
CHAPTER 12
Output Results
The types of data that may be selected for writing into the I-DEAS universal file are listed in Marc Volume C: Program
Input under the SDRC model definition option.
Marc - ADAMS Results Interface
ADAMS/Flex allows flexible components to be included into ADAMS models to achieve more realistic simulation
results. Marc is capable of generating a Modal Neutral File (MNF) representing the flexible component to be
integrated into the ADAMS model. The flexible component can undergo linear or nonlinear loading in Marc up to the
point where the MNF is requested to be generated. No more analysis takes place in Marc after generating the MNF.
Generating an MNF from Marc is based on performing the most general method of Component Mode Synthesis
(CMS) techniques, namely the Craig-Bampton method.
Using the Craig-Bampton method, the degrees of freedom, , of the flexible component are partitioned into two sets
of degrees of freedom:
Boundary or Interface Degrees of Freedom, : These degrees of freedom can be used to apply loads in
ADAMS or to connect the flexible component to other rigid bodies.
Interior Degrees of Freedom, : These internal degrees of freedom of the flexible component can be
condensed out using the superelement technique.
Two sets of modes shapes arranged in matrices, and , and their corresponding modal coordinates vectors,
and , are also defined:
Constraint Modes, : These modes are the static shapes obtained by giving each boundary degree of
freedom a unit displacement while holding all other boundary degrees of freedom fixed. Corresponding to the
constraint modes is the set of modal coordinates . The basis of constraint modes completely spans all
possible motions of the boundary degrees of freedom, with a one-to-one correspondence between the modal
coordinates of the constraint modes and the displacement in the corresponding boundary degrees of freedom,
.
Fixed-Boundary Normal Modes, : These modes are obtained by fixing the boundary degrees of freedom
and computing an eigensolution. There are as many fixed-boundary normal modes as the user desires. These
modes define the modal expansion of the interior degrees of freedom. The quality of this modal expansion is
proportional to the number of modes retained by the user. Corresponding to the fixed-boundary normal modes
is the set of modal coordinates .
The Craig-Bampton modes consist of both the constraint modes and the fixed-boundary normal modes. The
relationship between the physical degrees of freedom and the Craig-Bampton modes and their modal coordinates is
given by
(12-23)
u
u
B
u
I
I C

I N
q
C
q
N
I C
q
C
u
B
q
C
=
I N
q
N
u
B
u
I
)

`

I
BC
O
BN
I C

I N
q
C
q
N
)

`

=

where the subscripts B, I, C, and N denote Boundary Degrees of Freedom, Interior Degrees of Freedom, Constraint
Mode and Normal Mode, respectively. The above equation indicates that the Craig-Bampton analysis can be
considered as a transformation from the physical degrees of freedom to the Craig-Bampton modal basis in the form
where is the modal matrix or transformation matrix. The finite element equation of motion is given by
in which is the mass matrix, is the stiffness matrix and is the load vector. Substituting by the transformation
relation into the equation of motion and premultiplying by gives
The above equation defines the generalized mass and stiffness matrices as:
Generalized Mass Matrix, : Projection of the mass matrix onto the modal space given by
Generalized Stiffness Matrix, : Projection of the stiffness matrix onto the modal space given by
Note that the ADAMS MNF interface expects lumped mass matrices to be used. Thus, the LUMP parameter must be
included in the Marc input. The MNF interface distinguishes between two types of loads:
Modal Preload, : A projection onto the modal space of the internal load vector corresponding to the loads
that have already been applied in the finite element analysis up to the point when the MNF is generated.
Modal Loads, : A projection onto the modal space of the loads that may be defined in the Marc input using
the ASSEM LOAD history definition option but not applied in the finite element analysis. They may be scaled
and introduced in the ADAMS simulation. ASSEM LOAD loadcases, if exist, should precede any actual
preloading to be applied to the component in Marc.
Both types of loads can be computed using the relation
Corresponding to each Craig-Bampton mode, a stress and a strain mode can be computed. These are the incremental
nodal stresses and strains found by considering each mode at a time to be the incremental displacement solution and
performing a stress recovery. To request the computation and export of stress and/or strain modes, the MNF parameter
must be used.
q
u q =
Mu
Ku + f =
M K f
T
Mq
T
Kq +
T
f =
M
T
M =
K
T
K =
f
PL
f
L
f
T
f =
757
CHAPTER 12
Output Results
The Craig-Bampton analysis is triggered when the SUPERELEM model or history definition option is reached in the
input. The SUPERELEM option allows direct definition of the boundary or interface DOFs, . The option also allows
automatic definition of interface degrees of freedom of the nodes that get in contact with selected rigid contact bodies.
This is very useful for some nonlinear analyses such as tire footprint analysis in which the interface DOFs are not
known a priori. It also allows the specification of interface degrees of freedom of the control nodes of selected load-
controlled rigid contact bodies. The two control nodes for load-controlled rigid bodies are consolidated into one node
with six degrees of freedom before exporting to the MNF.
A MODAL SHAPE loadcase using the Lanczos method must be present in the input immediately following the
SUPERELEM loadcase. No more analysis can be performed afterwards.
The units used in the Marc model are specified using the MNF UNITS model definition option. Successful MNF
generation produces Marc Exit 3018. The MNF generated will have the same filename as the input file and the
extension .mnf. Assuming the input file is jobname.dat, the MNF will have the name jobname.mnf.
Status File
The status file contains summarized information of the analysis. Each line within this file includes the load case
number, the associated increment number, and the number of cycles, contact separations, and cutbacks associated with
that increment; the accumulated total cycles, increment splittings, separations, cutbacks, and remeshing times numbers
in the analysis as well as the time step size of each increment and the overall time achieved by the analysis. Marc
reports all this information on one line upon completion of each increment. If the increment is partially completed (for
example, the completion of one part of a split increment or one part of the whole increment due to cutback), Marc also
reports one line for each part of the increment.
In the PRINT NODSTS option in used, the maximum and minimum resultant quantities are added to the status files. It
is also possible to specify a node in which case the results of this node are provided. This is an easy way to monitor
the results.
Element Group Information
The finite element model is internally subdivided into element groups. The groups are based upon the element type
and material ID. The division into groups is internal to the program and entirely automatic. The memory allocation is
done separately for each element group, allowing the memory usage to be optimized for cases when different element
types and materials are used. Information about the element groups is printed to the output file. This printout occurs
in two places. The program first prints the amount of memory used for each group. Here is a sample printout for a job
with two element groups
memory usage per element group
group # elements nelsto MByte words
1 1368 1480 8 2024640
2 456 1284 2 585504
----------------------------------------------------------
total 1824 10 2610144
For each groups, it lists the number of elements in the group; nelsto, which is the amount of memory in words used
for each element and the total amount of memory used for all elements in the group, given in Megabytes and words.
u
B

A second printout is written further down in the output file after reading the model definition input data:
number of element groups used: 2
group # elements element type material formulation
1 1368 7 1 S
2 456 140 1 S

formulation:
S: small displacement
This listing shows the number of elements in the group and also the element type, the material ID, and the formulation
used in the group. If composites or rebars are used, it lists the composite or rebar group instead of the material ID. The
formulation column shows a number of characters indicating the type of formulation used. The table below shows the
characters that are used.
References
1. R. R. Craig and M. C. C. Bampton. Coupling of substructures for dynamics analyses. AIAA Journal, 6(7):
1313-1319. 1968.
2. R.R. Craig. Structural Dynamics: An Introduction to Computer Methods. John Wiley & Sons. 1981.
3. ADAMS User Guide: Using ADAMS/Flex. 2002.
Character Description
S small displacement small strain analysis
T total Lagrange formulation
U updated Lagrange formulation
M multiplicative plasticity or hyperelastic formulation
A additive plasticity formulation
F finite strain plasticity for additive formulation
E enhanced/assumed strain formulation
C constant dilatation formulation
Chapter 13 Parallel Processing
13
Parallel Processing
Introduction 760
Different Types of Machines 760
Supported and Unsupported Features 760
Matrix Solvers 762
Contact 763
Domain Decomposition 763


Introduction
Marc can make use of multiple processors when performing an analysis. Most, but not all, features of Marc are
supported in parallel mode. The type of parallelism used is based upon domain decomposition. A commonly used
name for this is the Domain Decomposition Method (DDM). The model is decomposed into domains of elements,
where each element is part of one and only one domain. The nodes which are located on domain boundaries are
duplicated in all domains at the boundary. These nodes are referred to as inter-domain nodes below. The total number
of elements is thus the same as in a serial (nonparallel) run but the total number of nodes can be larger. The
computations in each domain are done by separate processes on the machine used. At various stages of the analysis,
the processes need to communicate data between each other. This is handled by means of a communication protocol
called MPI (Message Passing Interface). MPI is a standard for how this communication is to be done and Marc makes
use of different implementations of MPI on different platforms. Marc uses MPI regardless of the type of machine used.
The types of machines supported are shared memory machines, which are single machines with multiple processors
and a memory which is shared between the processors, and cluster of separate workstations connected with some
network. Each machine (node) of a cluster can also be a multiprocessor machine.
Marc also has matrix solvers which can perform the matrix solution in parallel while the rest of the analysis is serial.
The MUMPS solver supports both shared memory machines and clusters. It uses MPI for the communication. The SGI
and Pardiso solvers support shared memory machines. These solvers can also be used with DDM.
Different Types of Machines
As mentioned above, Marc can run on a shared memory machine and on a cluster of workstations. The main reason
for running a job in parallel on a shared memory machine is speed. Since all processes run on the same machine sharing
the same memory, the processes all compete for the same memory. There is an overhead in memory usage so some
parts of the analysis need more memory for a parallel run than a serial job. The matrix solver, on the other hand, needs
less memory in a parallel job. Less memory is usually needed to store and solve several smaller systems than a single
large one.
In the case of a cluster, the picture is somewhat different. Suppose a number of workstations are used in a run and one
process is running on each workstation. The process then has full access to the memory of the workstation. If a job
does not fit into the memory of one workstation, the job could be run on, say, two workstations and the combined
memory of the machines may be sufficient.
The amount of speed-up that can be achieved depends on a number of factors including the type of analysis, the type
of machine used, the size of the problem, and the performance of communications. For instance, a shared memory
machine usually has faster communication than a cluster (for example, communicating over a standard Ethernet). On
the other hand, a shared memory machine may run slower if it is used near its memory capacity due to memory access
conflicts and cache misses etc.
Supported and Unsupported Features
Most of the main features of Marc are supported in parallel mode. For instance, a fully coupled thermomechanical
contact analysis can be performed in parallel. Local adaptivity is fully supported. New elements and nodes that are
created remain in the same domain as the parent element. Global adaptive meshing is supported with DDM, but each
body to be remeshed cannot be in multiple domains.
761
CHAPTER 13
Parallel Processing
The list of unsupported features is:
Ablation
Acoustics
Auto therm creep
Beam-to-beam contact
Bearing
Buckling
Convective terms in heat transfer
Design sensitivity and optimization
Eigenvalue extraction
Electromagnetics
Explicit dynamics
Fluids and its coupled analysis
Gap elements (but the normal contact is supported)
Harmonics
Hydrodynamics
Insert
Pyrolysis
Radiation
Response spectrum
Segment-to-segment contact
Steady state rolling analysis
Superelements
VCCT with crack propagation
Hypermesh
Global remeshing is supported by DDM but with the restriction that each remeshing body must be fully contained in
a domain.
The DDM single input file capability will not support any analyses types and input options that are not currently
supported by the traditional multiple input files DDM.
Moreover, some additional model and history definition options and command line arguments are not supported in the
current release. These are:
ACTIVATE and DEACTIVATE
ACTUATOR
BEGIN SEQUENCE and END SEQUENCE
CHANNEL
CFAST
CONRAD GAP
CWELD
DAMPING
DISP CHANGE if not FIXED DISP format. The DISP CHANGE option is supported only if the 1st entry of the
data block following DISP CHANGE is equal to 0.
EXCLUDE
FXORD
GLOBALLOCAL

GRID FORCE
INCLUDE
INSERT
NEW
POINT LOADS with follower forces
SUPERPLASTIC
QVECT
TIME-TEMP
TYING CHANGE
UDUMP
def command line option
Auxiliary input files
The following notes apply to the use of the single post file option:
Although general DDM jobs support post revisions 7 and higher, the ability to generate a single post file from
a DDM job is restricted to post revisions 9 and higher.
Reading a single post file back into a Marc DDM job using the -pid command line option is only available
for the AXITO3D and PRE STATE options in this release.
In the case of single input file jobs with continuous post files where the POST option follows the RESTART
option, both jobs should either use multiple post files or single post files. If multiple post files are used, then
the number of domains should be the same in both jobs.
Matrix Solvers
All matrix solvers in Marc are supported in conjunction with DDM parallel mode with the exception of the
unsymmetric solver and the mixed direct-iterative. Only the multifrontal solver (solver type 8 and PARDISO solver
type 11) supports out-of-core solution in parallel. The matrix solvers work somewhat different in parallel mode as
compared to serial mode.
For some of the direct solvers (type 0, 4, and 8), a two-stage approach is used. In a mechanical analysis, all inter-
domain nodes are first given fixed displacements and each domain solves the resulting equation system independently.
The second stage operates on the interdomain nodes to relax the extra imposed fixed displacements. This stage is not
necessarily if there are no connections between the meshes in the different domains. An iterative procedure based upon
the conjugate gradient method is performed for this stage. This procedure operates as mentioned on the inter-domain
nodes. The number of conjugate gradient iterations needed to converge to the final solution is reported in the Marc
output file as nn inter-domain iterations.
The conjugate gradient iterative solver (solver type 2) operates simultaneously on the whole model. It works to a large
extent like in a serial run. For each iteration cycle, there is a need to synchronize the residuals from the
different domains.
The third type of solver supported is the hardware solver (SGI), Pardiso, and MUMPS. For this case, the global
stiffness matrix (or the corresponding matrix for nonstructural analysis) is gathered to the parent process (where the
job was started). The original child processes are put to sleep and the solver creates new processes (multiple threads)
to solve the equation system in parallel. When the solution is finished, the child processes are awakened and the data
related to each domain are distributed. The hardware solver, Pardiso solver, or MUMPS may be used in conjunction
763
CHAPTER 13
Parallel Processing
with DDM on a shared memory architecture. For the hardware or Pardiso solver, multithreading is used to obtain the
performance improvement. For distributed memory or cluster environments, the MUMPS solver may be used in
conjunction with DDM.
Contact
Contact analysis poses a special challenge for parallelization due to its global nature. The current version of Marc fully
supports contact analysis including thermal and electrical contact but excluding beam-to-beam contact.
In order to enable a fast contact search procedure, some contact data are duplicated across all domains. A node in one
domain must be able to find contact with a segment in a different domain. The duplicated data includes nodes on the
exterior boundary of deformable contact bodies. This data duplication is not necessary in a pure rigid contact job, so
it is important to define a contact table to indicate that no deformable contact can take place in this case. The default,
if no contact table is used, is that all deformable bodies check for contact with all other bodies including itself.
If a node touches a segment which is located in a different domain, the nodes making up the segment are imported into
the domain of the touching node. If the node separates, the nodes are deleted from the domain. This means that the
number of nodes in a domain can dynamically change. This procedure allows the creation of tyings spanning domains.
This also has implications on how the domains should be created. One should avoid having contact bodies with a large
number of nodes coming into contact across different domains. An extreme case would be two sheets of equal number
of shell elements coming into contact on top of each other. If two domains are used and each sheet is in a separate
domain, the number of nodes in one domain would double.
Domain Decomposition
The way the model is decomposed into domains, in general, affects the performance of the parallel run. Traditionally,
domain decomposition had to be performed in the preprocessor stage using Mentat or MD Patran. Both Mentat and
Patran use Metis to perform the domain decomposition using the multiple file option. For a single input file case, the
domain decomposition can also be performed in Marc.
The general goal for a good decomposition is to minimize the number of interdomain nodes. Keeping this number low
minimizes the overhead in having duplicate nodes and also minimizes the size of the system solved in the interdomain
part of the direct solver. It is important to get a good balance between the workload in the different domains if, for
instance, different types of elements are used or if the nodes of a cluster have different performance.
Another aspect of deformable contact as mentioned above is to avoid having a large number of nodes touching
segments of a different domain.
Several different methods for creating the decomposition are available in Marc (single input), Mentat, and Patran. One
should note that some of them are based on element connectivity so they tend to put separate contact bodies in separate
domains which is not always optimal.
The optimal number of domains to use for a given job is pretty hard to estimate. It obviously depends on the number
of processors available. There is, in general, an optimal number of processors to use for a given job and a given
machine or cluster of machines. A general rule is that the larger the model is, the more it pays off to use more
processors.

Running a Parallel Job
Running a Marc parallel job is not much different from running a serial job. The model can be decomposed into
domains by the preprocessor (Mentat or Patran) and a separate input file is created for each domain. From a users
perspective, the extra steps consist of activating the procedure for decomposing the model and activating the parallel
run.
Marc can also be used to create the domains in case a single input file containing the whole model is used. Two
methods can be used to trigger the single input file option; the first is through the command line option as explained
below and the second is through the use of the PROCESSOR parameter.
If the command line option is used and the PROCESSOR parameter is not defined, the default domain decomposition
method, Metis Best, is used. Currently, the single input file option does not support any additional features that are not
supported by the traditional multiple input file option as described by the list of unsupported features above.
For the case of a run on a cluster, one would additionally have to define which machines make up the cluster (unless
an automatic cluster tool is used). The run is started as usual and, by default, separate results files are created for each
domain. The postprocessor automatically picks up the results from the different domains and combines them into a
single model which looks the same as in a serial run. One can also choose to look at the results of each domain
separately, which is particularly useful for large jobs. As opposed to the default multiple post file option, the user also
has the option to request a single post file for the whole model. For details on how to use this functionality, refer to
Marc Volume C: Program Input for the POST model definition option.
When running a parallel job from Mentat or Patran, one has to activate the button for parallel and start the job as for
a serial job. From the command line, one needs to use an extra argument to specify the number of domains:
run_marc -v n -b n -j jobname -np 5
for running a job with five domains using multiple input files generated by a preprocessor. To indicate that a single
input file will be used, the -np option is replaced with -nps and the command line in this case becomes:
run_marc -v n -b n -j jobname -nps 5
User subroutines are handled in the same way as in a serial run. See Marc Volume D: User Subroutines and Special
Routines for a description on special considerations for user subroutines in a parallel run.
Running on a cluster (a set of connected computers) requires that the machines communicate properly. The actual
connection could be anything from a telephone modem up to a high speed direct connected cluster. A standard 100
Mbit Ethernet network is usually sufficient for reasonable performance, although it can become bottlenecked with fast
processors. Faster connections give better performance on the communication part.
In order to perform an analysis over a network, one needs to specify which machines are to be part of the analysis.
This is done through a so-called host file, which is generic for a Marc run. The Marc startup script converts it into a
format that the current MPI implementation expects. The host file is a simple text file that lists the computers that are
to be used in the run and some info about the run. The host file typically looks something like (using a Unix system)
host1 2
host2 1 /users/joe/run /programs/marc
host3 2 /usr/people/joe /programs/marc/marc
Here host1, host2, and host3 are the host names of the respective machines. It specifies a job with five domains
where two processes are run on host1 and host3 and one on host2. /users/joe/run is the directory where the
job is run on host2. This is where the input files are copied and where the results files are created. This directory could
765
CHAPTER 13
Parallel Processing
be a local directory on a hard drive on host2 or it could be a shared directory (via NFS for example). This directory
must be accessible from host2 (but not necessarily from the other hosts). The fourth entry of the host file specifies
where the Marc installation directory is located. This can also be a local directory (if Marc is installed on the local
machine), or it can point to a shared directory. It is recommended that the run directory be a local directory to avoid
I/O traffic over the network. Not all MPI implementations currently supported uses the fourth entry. Specifically, the
commonly used HP-MPI and Intel MPI (used on Linux, HP-UX and Windows) do not support the specification of
different installation directories. They need to use the same path on all hosts. They can be shared or local but the path
name must be the same.
The input files for the remote hosts are automatically copied to each host in case the run directory is local. When the
run is finished, the post files are copied back. Please note that only the input files and the post files (not output files,
etc.) are copied. This copying can be suppressed by buttons in Mentat and MD Patran or by a command line option if
Marc is manually started. If a user subroutine is used together with local run directories, the new executable is
automatically transferred to all hosts before the Marc job starts.
One can also set up Marc to run via external cluster tools. A useful run time option in this context is the command line
option -dir. It specifies the run directory for the job and is where the scratch files and results files are created. Suppose
Marc is set up to run on a cluster using some kind of cluster tool to distribute the processes and each machine in the
cluster has a local disk called /scratch. By using the -dir /scratch option from the command line, the run uses
the local directory during the run. The input files are still read from where the job is started (which could be a shared
directory on a control workstation which is not even part of the run).
The following command line options related to parallel processing are available for the Marc run_marc start-up
script:
Notes on Windows:
1. The Marc installation directory is to be shared for running in clusters (multiple machines). The UNC naming
convention is to be used for shared directories.
2. The SMPD service is to be registered and kept running in the hosts for running a job in clusters. The service
can be registered and started using the binary files available in the installation directory which can be done as
given below. While registering, the user-name and password is to be provided.
\\host1\marc\intelmpi\win32\bin\ wmpiregister.exe
\\host1\marc\intelmpi\win32\bin\ismpd.exe -install
-np nn specify nn number of processes.
-nps nn specify nn number of processes if using a single input file.
-ci y|n if y, copy input files to local run directories; if n, do not copy.
-cr y|n if y, copy post files from local run directories; if n, do not copy.
-ho hostfile specify the host file that describes the hosts to use in a network run.
-dir rundir specify that the run directory of the run is rundir.
-mpi specifies the MPI version to use if other than the default.

3. The hostfile definition is similar to that for Unix machines. It is to be noted that the directories specified in the
third entry are the scratch directories and are local to the host (first entry). Shared locations can also be used as
scratch directories for which UNC naming convention shall be used. An example of the host file definition is
shown below.
host1 1 E:\host1\
host2 1 E:\host2\
4. An example for running Marc for the above specification is given below.
\\host1\marc\tools\run_marc -b no -j e2x1 -nps 2 -host hostfile
Domain Decomposition Methods
Metis Best - method performs Metis element based, Metis node based, and Vector decomposition and picks the best
decomposition with respect to the number of interdomain nodes as well as the domains balance. (Figure 13-1)
Metis_element_based - performs decomposition of graph constructed from element connectivity.
Metis_node_based - performs decomposition of graph constructed from nodal connectivity.
Figure 13-1 Metis Best Decomposition
Vector - performs decomposition based on IDs of elements. The element ids are sorted by coordinates of the element
centroids. The default direction of sorting is in the maximal dimension of the model bounding box in x-, y-, and z-
directions respectively. User may also specify direction of sorting by supplying the direction vector. See Marc Volume
C: Program Input, Chapter 2, Parameters PROCESSOR. (Figure 13-2)
767
CHAPTER 13
Parallel Processing
Figure 13-2 Vector Decomposition with Specified User Direction (1,1,0)
Radial - performs decomposition based on IDs of elements. The element ids are sorted by distance between the
element centroids and axis of rotation. The default (axis of rotation) is specified by default direction and default point
on the axis of rotation.
Default point for axis of rotation is defined as the centroid of the model bounding box. User may also specify the axis
of rotation by point and direction. See Marc Volume C: Program Input, Chapter 2, Parameters PROCESSOR.
(Figure 13-3)
Figure 13-3 Radial Decomposition with User Direction(.5,.5,.707) and
Point on Axis (2,2,2.141)

Angular - performs decomposition based on ids of elements. The element ids are sorted by angle between the element
centroids, axis of rotation and selected start vector. The default (axis of rotation) is specified by default direction and
default point on the axis of rotation. User may also specify the axis of rotation by point and direction. See Marc Volume
C: Program Input, Chapter 2, Parameters PROCESSOR. (Figure 13-4).
Figure 13-4 Angular Decomposition with User Direction(.5,.5,.707) and Point on Axis (2,2,2.141)
Recursive Coordinate Bisection - performs recursively vector decomposition into two domains. Each domain is
subsequently subdivided into two or more domains until all domains are created (Figure 13-5).
Figure 13-5 Recursive Coordinate Bisection
User - performs decomposition based on element sets in the input file. The element sets must be named domaini
where i is the domain number. The domain sets are usually created by the user in the preprocessor program.
769
CHAPTER 13
Parallel Processing
The Fine Graph flag in the 2nd data block on the PROCESSOR parameter tells Metis to use a fine graph for
decomposition. This results in higher memory requirements and lower speed for the decomposition. In some cases, it
produces better decomposition than the default (Coarse Graph). The Coarse Graph choice might produce more
islands. It is recommended to use the default unless there is a need for better quality of domain decomposition.
Sometimes the domain decomposition produces disconnected domains resembling islands (Figure 13-6). The best
performance usually is obtained when the domains are contiguous. Islands can be added to neighboring domains by
turning on the Island Remove flag in the 2nd data block on the PROCESSOR parameter. This might result in worse
balance of domains as shown in Figure 13-7.
Figure 13-6 Metis Node Based Decomposition without Island Removal
Figure 13-7 Metis Node Based Decomposition with Island Removal
Chapter 14 Code Coupling Interfaces
14
Code Coupling Interfaces
Introduction 771
Code Coupling 771
Coupling Regions 773
References 775
771
CHAPTER 14
Introduction
This chapter describes the coupling interface to external solvers available through the user subroutine programming.
The interface allows code coupling libraries such as MpCCI (see [Ref. 1]) to couple Marc with commercial CFD
codes, such as STAR-CD and Fluent. The interface may also be used for developing dedicated interfaces to external
solvers, for applying complex boundaries to certain regions of the model, or for dedicated postprocessing.
Code Coupling
Analyzing fluid-structure interaction problems has become increasingly important in industrial applications. Effects
of the deforming structure on the fluid flow and, vice versa, of the fluid pressure on the structure, as well as heat
transfer between fluid and structure can be of critical importance for a design, especially if the deformations of the
structure are large.
The fluid-structure capabilities of Marc (see Chapter 6 Nonstructural Procedure Library in this manual) are limited to
nonreactive, incompressible, single phase, laminar flows. This implies that interaction problems such as combustion
reactions in airbag gas generators, involving complex turbulent flows with large deformations of the structure cannot
be solved. An additional drawback is that, at the fluid-structure interface, the fluid and structural meshes must share
the same nodes. This may be inconvenient, as the fluid analysis tends to require a denser mesh than the
structural analysis.
Code coupling software, such as the MpCCI (Mesh-based parallel Code Coupling Interface) software from Fraunhofer
SCAI (see [Ref. 1]), allows structural codes such as Marc to be coupled with commercial CFD codes such as STAR-
CD or Fluent. Fluid-structure interaction problems can then be solved using the capabilities of both solvers in their
respective domains. The two solvers run simultaneously and exchange data regularly during the analysis on the
common boundary between the fluid and the structure.
Figure 14-1 shows a flow chart of such a coupled analysis, in which the CFD solver and the structural solver
alternately solve their respective parts of the coupled problem. The CFD solver solves the fluid flow problem for a
given period of time (the coupling time step) to provide the pressure that the fluid exerts on the structure, the
temperature of the fluid at the common boundary with the structure and the film coefficient to the structural solver.
The structural solver applies the fluid pressure and the film boundary condition to the structure and computes the
deformations and temperatures of the structure for the same period of time. Then it sends the current coordinates of
the nodes and the temperature at the fluid-structure interface to the CFD code. The CFD code uses these coordinates
to update the flow domain and continues with the next coupling time step. Note that the coupling time step may consist
of multiple time steps in each of the solvers.

Figure 14-1 Flow Chart of a Coupled Fluid-Structure Analysis Using Code Coupling Software
CFD
Time Step(s)
Structural Time
Step(s)
CFD
Time Step(s)
Structural Time
Step(s)
CFD Analysis
Structural Analysis
Code Coupling Software
CFD
Mesh
Structural
Mesh
Fluid Pressure, Fluid
Temperature, Film
Coefficient
Current Coordinates,
Structural Temperature
Fluid Pressure, Fluid
Temperature, Film
Coefficient
Initialization
Current Coordinates,
Structural Temperature
Coupling
Time Step
Coupling
Time Step
Initialization
773
CHAPTER 14
The code coupling software serves as a communication layer between the two codes. It handles the data transfer
between the solvers and interpolates the data from the CFD mesh to the structural mesh and vice versa, in case the
meshes do not match at the common boundary (Figure 14-2). To be able to interpolate the data from one mesh to the
other, the code coupling software determines the geometrical relationship between the meshes, using information
(connectivity and coordinates) provided by the two solvers at the start of the analysis.
Figure 14-2 Non-Matching Meshes at the Common Boundary Between Fluid and Structure
The basic stages, from a code coupling point of view, in a coupled analysis are:
1. Initialization. The relationship between the CFD mesh and the structural mesh is established. Both the CFD
solver and the structural solver must provide the code coupling software with the connectivity and the
coordinates of the patches on the common surface between the fluid and the structure.
2. Data exchange. Physical quantities such as the fluid pressure and the temperature of the fluid at the common
surface with the structure, are extracted from the CFD code, interpolated to the structural mesh, transferred to
the structural code and applied through appropriate boundary conditions to the structural mesh. Likewise,
current coordinates and temperature of the structure at the common surface with the fluid are extracted from
the structural code, interpolated to the CFD mesh and transferred to the CFD code. The CFD code uses the new
coordinates to update the flow domain.
3. Finalization.
Coupling Regions
Coupling regions are the basic concept to couple Marc with external solvers via code coupling software such as
MpCCI. A coupling region is that part of the surface or volume of the structural model where the interaction with the
external solver takes place. A surface region consists of a list of edges or geometric curves in 2-D and a list of faces
or geometric surfaces in 3-D. A volumetric region consists of a list of elements or contact bodies. Coupling regions
are defined by the COUPLING REGION model definition option and are accompanied by an extensive application
programming interface (API). The basic mechanical and thermal quantities (current coordinates, displacement, force,
pressure, temperature, heat flux, film coefficient) can be exchanged with an external solver on coupling regions via
API calls. Quantities that are received from the external solver are applied to the coupling region via boundary
conditions.
Structural Mesh
CFD Mesh

The coupling region API consists of three user subroutines and a set of utility routines that can be called from these
user subroutines (see Chapter 9 Special Routines - Marc Post File Processor, Marc Volume D: User Subroutines and
Special Routines). The three user subroutines correspond to the basic three stages of a coupled analysis, as outlined in
the previous section:
1. Initialization: CPLREG_INIT. This subroutine is called once at the start of the analysis and can be used to
extract the connectivity and coordinates of the coupling region and pass it to the code coupling software to
determine the geometric relationship between the coupling region mesh and the mesh of the external solver.
2. Data exchange: CPLREG_EXCHANGE. This subroutine is called twice per coupling time step, once at the start
and once at the end. The call at the start can be used to set the new values of physical quantities that have been
received from the external solver for the next coupling time step. The call at the end can be used to extract the
new values of physical quantities that must be sent to the external solver.
3. Finalization: CPLREG_FINALIZE. This subroutine is called once at the end of the analysis and can be used to
inform the external solver that the Marc analysis has ended.
The utility routines can be called from the user subroutines to obtain the connectivity and coordinates of the coupling
region, to obtain the current values of physical quantities on coupling regions, to set the new values of physical
quantities for the next coupling time step and even to specify the next coupling time step. The available utility routines
are listed in Table 14-1.
Table 14-1 Available Utility Subroutines for Coupling Regions
Utility Subroutine
Available in User
Subroutines Purpose
CPLREG_FIND_NAME CPLREG_INIT
CPLREG_EXCHANGE
Find coupling regions by name.
CPLREG_GET_INFO CPLREG_INIT
CPLREG_EXCHANGE
Get general information about a coupling
region.
CPLREG_GET_QUANTS CPLREG_INIT
CPLREG_EXCHANGE
Get the prescribed physical quantities on a
coupling region.
CPLREG_GET_MESH CPLREG_INIT
CPLREG_EXCHANGE
Get the mesh of a coupling region.
CPLREG_GET_GLOBAL_VALUES CPLREG_EXCHANGE Get the values of a global quantity.
CPLREG_GET_NODE_VALUES
CPLREG_GET_ALL_NODE_VALUES
CPLREG_EXCHANGE Get the values of a node-based quantity at a
coupling region.
CPLREG_PUT_GLOBAL_VALUES CPLREG_EXCHANGE Put the values of a global quantity.
CPLREG_PUT_NODE_VALUES
CPLREG_PUT_ALL_NODE_VALUES
CPLREG_EXCHANGE Put the values of a node-based quantity at a
coupling region.
CPLREG_PUT_EDGE_VALUES
CPLREG_PUT_ALL_EDGE_VALUES
CPLREG_EXCHANGE Put the values of an edge-based quantity at
a coupling region.
CPLREG_PUT_FACE_VALUES
CPLREG_PUT_ALL_FACE_VALUES
CPLREG_EXCHANGE Put the values of a face-based quantity at a
coupling region.
CPLREG_PUT_ELEM_VALUES
CPLREG_PUT_ALL_ELEM_VALUES
CPLREG_EXCHANGE Put the values of an element-based quantity
at a coupling region.
775
CHAPTER 14
Time Step Control
The default coupling time step is equal to the Marc time step, so that the CPLREG_EXCHANGE user subroutine is
called at the start and at the end of each increment. If the AUTO STEP time stepping scheme is used, the coupling time
step may be prescribed by calling the CPLREG_PUT_GLOBAL_VALUES utility routine from the
CPLREG_EXCHANGE user subroutine. In that case, the AUTO STEP algorithm matches the next coupling time, but
multiple increments may be needed to reach this coupling time. If multiple increments are performed within a given
coupling time step, then the CPLREG_EXCHANGE user subroutine is called at the start of the first increment and at
the end of the last increment of the coupling step.
Shell Elements
The top and bottom faces of a shell element are independent surface entities, as far as coupling regions are concerned.
They can both be part of the same coupling region or they can be part of two different coupling regions. The faces
share the same nodes, but the connectivity of the bottom face, as returned by the CPLREG_GET_MESH utility routine,
is exactly the opposite of that of the top face (which is the connectivity of the element).
In heat transfer and coupled thermo-mechanical analyses, the temperatures (and nodal heat fluxes) at the top and
bottom of a shell element are distinct. If a coupling region contains only top faces of shell elements, the
CPLREG_GET_NODE_VALUES utility routine returns the top temperatures (and heat fluxes) at the nodes of the
region. Similarly, the CPLREG_PUT_NODE_VALUES utility routine prescribes the top temperatures (and heat fluxes)
at the nodes of the region. If the coupling region contains only bottom faces of shell elements, these routines return
and prescribe the bottom temperatures (and heat fluxes) at the nodes. If both the top and the bottom faces of shell
elements interact with the mesh of an external solver, two coupling regions must be defined to properly access the
temperatures and nodal heat fluxes on both sides of the shell elements: one containing the top faces and the other the
bottom faces of the shells.
Parallel Processing
In a parallel run with Marc, the finite element mesh is subdivided into domains where each element is part of one
domain. Nodes at the boundary between domains are present in all domains sharing that boundary. Coupling regions
are allowed to span multiple domains. If a physical quantity is prescribed on a such a coupling region, using the
CPLREG_PUT_NODE_VALUES utility routine, the value of the quantity for a given node must be the same in all
domains in which the node appears. Similarly, if the coupling time step is prescribed via
CPLREG_PUT_GLOBAL_VALUES, the value must be the same in all domains.
References
1. Fraunhofer Institute for Algorithms and Scientific Computing SCAI. MpCCI 3.0.4: Manuals and Tutorials.
April 25, 2005, http://www.scai.fraunhofer.de/mpcci.
Appendix A Finite Element Technology in Marc
A
Finite Element Technology in
Marc
Governing Equations of Various Structural Procedures 777
System and Element Stiffness Matrices 779
Load Vectors 780
References 781
777
APPENDIX A
Finite Element Technology in Marc
Governing Equations of Various Structural Procedures
This section describes the basic concepts of finite element technology. For more information, you are urged to refer to
the finite element text [Ref. 5]. Marc was developed on the basis of the displacement method. The stiffness
methodology used in Marc addresses force-displacement relations through the stiffness of the system. The force-
displacement relation for a linear static problem can be expressed as
(A-1)
where is the system stiffness matrix, is the nodal displacement, and is the force vector.
Assuming that the structure has prescribed boundary conditions both in displacements and forces, the governing
Equation (A-1) can be written as
(A-2)
is the unknown displacement vector, is the prescribed force vector, is the prescribed displacement vector,
and is the reaction force. After solving for the displacement vector , the strains in each element can be calculated
from the strain-displacement relation in terms of element nodal displacement as
(A-3)
The stresses in the element are obtained from the stress-strain relations as
(A-4)
where and are stresses and strains in the elements, and is the displacement vector associated with the
element nodal points; and are strain-displacement and stress-strain relations, respectively.
In a dynamic problem, the effects of mass and damping must be included in the system. The equation governing a
linear dynamic system is
(A-5)
where is the system mass matrix, is the damping matrix, Equation (A-6) is the acceleration vector, and is the
velocity vector. The equation governing an undamped dynamic system is
(A-6)
The equation governing undamped free vibration is
(A-7)
Natural frequencies and modal shapes of the structural system are calculated using this equation.
(A-8)
Ku f =
K u f
K
11
K
12
K
21
K
22
u
1
u
2
)
`

f
1
f
2
)
`

=
u
1
f
1
u
2
f
2
u
el
u
el
=
el
L
el
=
el

el
u
el
L
Mu
Du
Ku + + f =
M D u
Mu
Ku + f =
Mu
Ku + 0 =
K
2
M 0 =

The equations governing some other procedures are similar. For example, the governing equation of transient heat
transfer analysis is
(A-9)
where is the heat capacity matrix, is the thermal conductivity matrix, is the thermal load vector (flux),
is the nodal temperature vector, and is the time derivative of the temperature. Equation (A-9) reduces to
(A-10)
for the steady-state problem. Note that the equation governing steady-state heat transfer (Equation (A-10)) and the
equation of static stress analysis (Equation (A-1)) take the same form. Similarly, the hydrodynamic bearing problem
is analogous to a steady-state heat transfer problem. This problem is governed by an equation similar to Equation
(A-10).
The matrix equation of the electrical problem in coupled thermo-electrical analysis is
(A-11)
The equation governing the thermal problem is:
(A-12)
in both Equation (A-11) and Equation (A-12), is the voltage, is the temperature-dependent
electrical-conductivity matrix, I is the nodal-current vector, is the temperature-dependent, heat-capacity
matrix, and is the thermal-conductivity matrix. is the nodal temperature vector, is the heat-flux vector,
and is the internal heat-generation vector that results from the electrical current. Electrical and thermal problems
are coupled through and .
The matrix equations for the thermal-mechanical problem are as follows:
(A-13)
(A-14)
In Equations (A-13) and (A-14) the damping matrix , stiffness matrix , heat-capacity matrix and thermal-
conductivity matrix are all dependent on temperature. is the internal heat generated due to inelastic
deformation. The coupling between the heat transfer problem and the mechanical problem is due to the temperature-
dependent mechanical properties and the internal heat generated. If an updated Lagrangian analysis is performed,
and are dependent upon prior displacement.
The governing equations described above are either sets of algebraic equations (Equations (A-1), (A-10), and (A-11)), or
sets of ordinary differential equations (Equations (A-5) through (A-8) and all equations through Equation (A-4)). The
time variable is a continuous variable for the ordinary differential equations. Select an integration operator (for
C
T
T
K
T
T + Q =
C
T
K
T
Q T
t
K
T
T Q =
K
E
T ( )V I =
C
T
T ( )T
K
T
T ( )T + Q Q
E
+ =
v K
E
T ( )
C
T
T ( )
K
T
T ( ) T Q
Q
E
K
E
T ( ) Q
E
Mu
Du
K T u t , , ( )u + + F =
C
T
T ( )T
K
T
T ( )T + Q Q
I
Q
F
+ + =
D K C
T
K
T
Q
I
K
K
T
779
APPENDIX A
example, Newmark-beta, Houbolt, or central difference for dynamic problems, and backward difference for heat
transfer) to reduce the set of differential equations to a set of algebraic equations. The final form of governing
equations of all analysis procedures is, therefore, a set of algebraic equations.
System and Element Stiffness Matrices
The previous section presented system matrices assembled from element matrices in the system. For example, the
system stiffness matrix is expressed in terms of the element stiffness matrix as
(A-15)
where is the number of elements in the system. The system stiffness matrix is a symmetric-banded matrix. See
Figure A-1 for a schematic of the assemblage of element matrices. When the element-by-element iterative solver is
used, the total stiffness matrix is never assembled.
The element stiffness matrix can be expressed as
(A-16)
where is the volume of the element, is the strain-displacement relation, and is the stress-strain relation.
(A-17)
(A-18)
The mass matrix can be expressed as
(A-19)
K K
i
el
K K
i
el
i 1 =
N
=
N
K
el
T
Ldv
el
v
el
}
=
v
el
L
u =
L =
M
el
N
T
Ndv
v
el
}
=

Figure A-1 Schematic of Matrix Assemblage
The integration in (A-16) is carried out numerically in Marc, and is dependent on the selection of integration points.
The element stiffness matrix can be fully or under integrated. The mass matrix is always fully integrated.
Load Vectors
The nodal force vector f in (A-1) includes the contributions of various types of loading.
(A-20)
where is the point load vector, is the surface load vector, is the body (volumetric) load vector, and
represents all other types of load vectors (for example, thermal and creep strains, and initial stress).
Note:
The matrix is directly associated with the shape functions and geometry of each element. Shape functions
associated with different element types are explained in Marc Volume B: Element Library. Stress-strain
relations L are discussed in Chapter 7 of this volume.
1 2
3 4 5
6
7
8
(a)
(b)
(c)
1 2 3 4 5
{F} [K]
Band
1
4 5
2
3
f f
poi nt
f
sur f ace
f
body
f
+ + + =
f
point
f
surface
f
body
f*
781
APPENDIX A
The point load is associated with nodal degrees-of-freedom and can be added to the nodal force vector directly.
Equivalent nodal force vectors ,
y
must be calculated from the distributed (surface/volumetric) load first
and then added to the nodal force vector. In Marc, the computation of equivalent nodal forces is carried out through
numerical integration of the distributed load over the surface area of volume to which the load is applied. This can be
expressed as
(A-21)
(A-22)
where is the pressure.
Figure A-1 shows the assemblage of the nodal force vector.
References
1. Zienkiewicz, O. C. and R. L. Taylor. The Finite Element Method (4th ed.) Vol. 1. Basic Formulation and Linear
Problems (1989),) Vol. 2. Solid and Fluid Mechanics, Dynamics, and Nonlinearity (1991) McGraw-Hill Book
Co., London, U. K.
2. Bathe, K. J. Finite Element Procedures, Prentice-Hall, Englewood Cliffs, NJ, 1995.
3. Hughes, T. J. R. The Finite Element MethodLinear Static and Dynamic Finite Element Analysis, Prentice-
Hall, Englewood Cliffs, NJ. 1987.
4. Ogden, R. W. Large Deformation Isotropic Elasticity: On The Correlation of Theory and Experiment for
Incompressible Rubberlike Solids, Proceedings of the Royal Society, Vol. A (326), pp. 565-584, 1972.
5. Cook, R. D., D. S. Malkus, and M. E. Plesha, Concepts and Applications of Finite Element Analysis (3rd ed.),
John Wiley & Sons, New York, NY, 1989.
f
surface
f
body
f
surf ace
N
T
pdA
A
}
=
f
body
N
T
pdV
V
}
=
p
Appendix B Finite Element Analysis of NC Machining Processes
B
Finite Element Analysis of NC
Machining Processes
General Description 783
NC Files (Cutter Shape and Cutter Path Definition) 783
Intersection Between Finite Element Mesh and Cutter 785
Deactivation of Elements 785
Adaptive Remeshing 785
Input of Initial Stresses 786

783
APPENDIX B
Finite Element Analysis of NC Machining Processes
General Description
Machining (or Metal Cutting) is a type of material removal processes widely used to produce a part with the desired
geometry. After removal of the machined material, re-establishment of equilibrium within the remaining part of the
structure causes some distortion due to the relief of the residual stresses in the machined materials.
Depending on the stress level and geometry of the part, the severe distortion due to machining can result in high scrap
rates and increased manufacturing costs, especially for structures with thin wall or large thin plates.
A numerical analysis procedure has been developed in Marc to predict the distortion during the 3-D bulk machining
so that engineers can optimize their structure design and machining processes. This procedure is based on the
assumption that effects introduced by the cutting process are local and small compared to those due to pre-existing
residual stresses.
The simulation procedure uses the cutter shape and path information obtained from NC machining data. Cutter motion
information is used to predict the material to be removed by automatically deleting finite elements that are located
within the cutter path. The distortion of the remaining part can be predicted by re-calculating the equilibrium. An
interface has been developed in Marc to convert cutter path data into the finite elements to be removed. The cutter path
data is stored in either the Automatically Programmed Tools (APT) source or the Cutter Location (CL) data format.
The APT source is the NC data output by CAD software CATIA. The CL data is the cutter location data provided by
APT compilers.
For details on how to use this functionality, refer to Marc Volume C: Program Input for the MACHINING parameter
and the DEACTIVATE model definition option and the DEACTIVATE history definition option.
NC Files (Cutter Shape and Cutter Path Definition)
The cutter shape and the cutter path are the key parameters to determining the volume of the material to be cut. The
cutter shape is defined by the CUTTER statement in the APT source or CL files. The expanded form of the CUTTER
statement is as follows:
where the parameters
have the following meaning as shown in Figure B-1.
The tool diameter.
The radius of the corner circle; it can be zero or larger than .
The radial distance from the tool axis to the center of the corner circle.
The distance from the tool endpoint to the center of the corner circle measured parallel with the tool axis.
The angle from a radial line through the tool endpoint to the lower line segment; it is in range of zero to 90,
and positive.
The angle between the upper segment and the tool axis; it is in the ranges of -90 to 90.
The cutter height measured from the tool endpoint along the tool axis.
CUTTER d r e f h , , , , , ,
d
r d 2
e
f
h

Figure B-1 Definition of Cutter Geometry
The cutter path is calculated based on the motion of the cutter and cutter shape. The motion is defined by cutter motion
statements like GOTO/, GODLTA/, CYCLE/, and DRILL/, etc. In Marc, these motion types are translated into point-to-
point, circular, or drilling motion modes, respectively. For details, refer to references by Irvin H. Kral and Dassault
Systems.
Notes:
1. Marc accepts APT data files written by CATIA V4 and the corresponding CL data files by APT
compilers.
2. The cutter can be of either multi-axis or fixed axis.
3. Two cutter motion types: Point-To-Point (for example, GOTO/ and GODLTA/ statements) and Cycle
(DRILL/) are directly supported. Circular motions (either in-plane or out-of-plane) can be interpolated
into Point-To-Point motions. The interpolation can be done inside CATIA by choosing the proper option
when writing out APT source file.
4. In addition, some statements in the APT/CL files are skipped. If the user chooses statements like
TRACUT/, COPY/, /MACRO, or /CALL to define cutter motions, they must be converted into explicit
motion statements like Point-to-Point or Cycle (DRILL/) motions before writing out the APT data file.
The cutting process is stopped if statements, such as /FINI, /STOP, or /END, or the end of APT/CL file
is reached.
5. It is assumed that the part and cutter are defined in the same coordinate system. So, if a part needs to be
flipped over between two machining stages, the user can rotate the cutter but keep the part unchanged
in space.
d
e
f
Upper Line
segment
Lower line
segment
Endpoint
r
Envelope
Tool axis
785
APPENDIX B
Finite Element Analysis of NC Machining Processes
Intersection Between Finite Element Mesh and Cutter
When the cutter moves, its path forms the volume of the material to be cut. This volume is calculated based on the
cutter geometry and its path in space. The finite element mesh is checked against this volume to decide the intersection
between the volume and the finite element nodes and elements.
Deactivation of Elements
The automatic deactivation of finite elements depends on the intersection status between the cutter path and the mesh.
The approach is designed for elements that are fully or partially located inside the cut volume. Elements that are fully
located inside the cut volume are automatically deactivated and the stress that existed before deactivation is released.
For elements that are partially intersected by the cut volume, integration points are checked. In the current version, the
check of integration points is limited to the element centroid only. If element centroid is found inside the cut volume,
the associated element is deactivated. For deactivated elements, the stresses/strains are set to zero permanently.
After the deactivation of elements along the motion of cutter, the part will distort progressively. The updated geometry
of the part is used for the next cutting step. Using this method, the whole machining process is simulated with Marc.
After the machining is finished, the final distortion of the work piece can be calculated by spring back analysis. The
spring back analysis is conducted by removing all fixturing restraints except for those needed to remove rigid
body motion.
Adaptive Remeshing
Generally speaking, the accuracy of the finite element analysis of machining processes depends on, plays a key role
to get the accuracy of initial (residual) stresses. Accurate initial stress the correct deformation of the part. From the
view point of FE analysis, the initial FE mesh may affect the accuracy significantly. Finer mesh is needed for accuracy
but it results in higher analysis cost due to the increased number of finite elements and increased total number of
analysis increments. Also, for most machining processes, the cutter path is complicated enough that it is not possible
to properly define the best suited fine mesh at the beginning of the analysis. The adaptive remeshing technology has
been enhanced in Marc to obtain accurate results for the FE mesh located inside the cutter path.
An element is subdivided if the cutter only cuts part of the element in the current cutting step. For example, one 3-D
brick element generates 8 new elements after first level subdivision and 64 new elements after the second level
subdivision. Therefore, users need to be aware of the potentially large number of new elements with increased levels
of subdivision.
6. If the user wants to change the boundary conditions for some nodes during the cutting process defined
by one APT/CL file, time synchronization should be turned on when defining the machining load case.
7. For visualization purpose, the user can specify a contact surface as cutter surface. During
postprocessing, user will be able to visualize the cutter motion along the cutting process. In order to do
this, you MUST exclude this cutter surface ID from all kinds of contact checking during analysis. The
user can do this by creating CONTACT TABLE and turn off the contact body ID of the cutter surface. In
addition, it is also necessary to make sure that the cutter surface is initially set at the location of the
origin of the coordinate system and the orientation is initially set to be in the direction of Z-axis.

Typical Features for Machining
1. Multiple level subdivision of element in one incremental step:
Typically, Marc subdivides only once per load step. If the 17
th
criterion of ADAPTIVE model definition option
is used (for NC machining), Marc automatically subdivides the element until the maximum specified level has
been reached. This process may generate a huge number of new elements and nodes. Consequently, these
newly created elements and nodes may result in excessive memory usage and CPU time, though most of them
are immediately cut off. For some problems, it is possible to cut the part with a few large elements but without
sacrifice of accuracy if only allowing remeshing to be conducted at some specific locations of the part where
finer meshing is required. In such circumstances, it might be reasonable to conduct one round of rough cutting
followed by fine cutting based on finer mesh after adaptive remeshing. Details are described in Item 3.
2. ADAPTIVE REMSHING at EACH increment:
At the load case of machining, Marc does adaptive remeshing at each increment based on the intersection of
the cuter path and the FE mesh.
3. ADAPTIVE REMSHING at LAST increment:
A cutting analysis in Marc is not done without adaptive remeshing until the cutting process is completed.
However, the cutting process may not be completely finished yet because some elements that are partially cut
by cutter are still retained. In order to cut those elements off, the cutter path is re-calculated and those elements
that are partially cut are subdivided in order to improve the accuracy of the cutting. In this case, extra
incremental steps are added into this machining load case. The added number of extra load steps equal the
maximum subdivision level that is defined by the ADAPTIVE option.
Input of Initial Stresses
There are various approaches to input the initial stresses into the FE model for machining analysis. Generally speaking,
Marc allows user to input the initial stress through the INIT STRESS model definition option. It also allows user to
import the data through the PRE STATE option. With the PRE STATE option, the user can transfer information,
including strain, stress etc., from previous analysis into the model for new analysis.
By using option of INIT STRESS, user may define the stress based on the input values or tables (if the stresses are
defined as function of, say, coordinates). Marc also allows the user to define the initial stresses with an ASCII format
text file. The text file contains the raw initial stress data of points randomly located in a space field. In Marc, the given
initial stress data is processed and interpolated into the finite element mesh according to the location of each element.
For details about this data structure and format, see INIT STRESS model definition option in Marc Volume C: Program
Input.
References
1. Irvin H. Kral, Numerical Control Programming in APT, Prentice-Hall, 1986.
2. Dassault Systems, CATIA Training Guide, Version 4, Release 1.6, April, 1996.
3. Houtzeel Manufacturing systems, Inc., APT Language Reference Manual, March, 1997.
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
MSC.Fatigue Quick Start Guide
I ndex

Index
INDEX
A
ablation 247
ABLATION (parameter) 248
acoustic
analysis 314, 526, 624, 691, 747, 749
element types 314, 331
input options 624
ACOUSTIC (parameter) 315
ACTIVATE (history definition option) 54
ADAMS 755
ADAPT GLOBAL (model definition option) 248, 557
ADAPTIVE (model definition option) 77, 95, 96,
557, 697
ADAPTIVE (parameter) 697
adaptive mesh
contact 569
global 80
local 77
refinement 697
adaptive time control 134
ALL POINTS (parameter) 279, 686
ALLOCATE (parameter) 29
analysis 340, 525
acoustic 314, 526, 624, 691, 747, 749
contact 749
coupled 324, 561
coupled acoustic-structural 328
coupled contact 561
coupled electromagnetic-thermal 343
coupled electrostatic-structural 691
coupled thermo-electrical (Joule heating) 337
coupled thermo-mechanical 326
crack 140
creep 701
design sensitivity 194
dynamic 748
eigenvalue 161
electrical 622
electristatic-structural 526
electromagnetic 305
electrostatic 525, 623, 691, 747, 748
electrostatic-structural 749
electro-structural 747
error 73
fluid 315, 526, 692
fluid/solid interaction 331, 525, 691
Fourier 96
harmonic 310
heat transfer 220, 524, 621, 746, 748
hydrodynamic bearing 285, 525, 622, 691, 748
Joule heating 748
large strain 694
linear 94
magnetodynamic 692, 749
magnetodynamics 747
magnetostatic 293, 526, 691, 747, 749
nonlinear 99
perturbation 117
piezoelectric 310, 525, 624, 692, 747, 749
post buckling 701
rigid cavity acoustic 314
rigid-plastic 183, 694, 748
snap-through 701
soil 188, 500, 692
steady state rolling 204
stress 747
structural zooming 208
thermal contact 563
thermo-electrical 525
analytical contact 575
ANELAS (user subroutine) 95, 348, 350
ANEXP (user subroutine) 95, 350, 495
ANISOTROPIC (model definition option) 95, 281,
311, 348, 350, 354, 381, 433, 456, 457, 493
anisotropic viscoelastic material 483
ANKOND (user subroutine) 221, 525
ANPLAS (user subroutine) 350, 433
Marc Volume A: Theory and User Information 788
application phase 157
APPROACH (history definition option) 538
Arbitrary Eularian-Lagrangian (AEL) formulation
113
arc-length methods 715
ASSEM LOAD (model definition option) 756
ASSUMED STRAIN (parameter) 696
assumed strain formulation 696
ATTACH EDGE (model definition option) 76, 611
ATTACH FACE (model definition option) 611
AUTO CREEP (history definition option) 134, 135,
136, 138, 139, 479, 700, 701
AUTO INCREMENT (history definition option) 67,
116, 500, 614, 700, 701, 702, 703, 718
AUTO LOAD (history definition option) 67, 335,
337, 488, 614, 700
AUTO STEP (history definition option) 138, 170,
222, 226, 335, 337, 343, 344, 483, 488, 500, 575,
582, 619, 700, 701, 702, 703, 704, 705, 706, 707,
709, 710
AUTO THERM (history definition option) 620, 700,
701, 702
AUTO THERM CREEP (history definition option)
137, 138, 700, 701
automatic global remeshing 80
Automatic Thermally Loaded Elastic-Creep/Elastic-
Plastic-Creep Stress Analysis (AUTO THERM
CREEP) 137
Automatically Programmed Tools (APT) 783
AUTOMSET (parameter) 645
AUTOSPC (parameter) 558
AXITO3D (model definition option) 203
B
bandwidth optimization 65
Barlats (1991) yield function 434
beam elements 689, 745
pin code 660
BEAM SECT (parameter) 69, 675, 689
beam-shell offsets 657
adaptive meshing 659
contact 659
dynamics 659
field analysis 659
offset vector specification 658
post file 659
supported elements 659
beam-to-beam intersection tying constraints 634
BEARING (parameter) 286
BEGIN SEQUENCE (history definition option) 620
behavior
incompressible 119
nearly incompressible 121
nonlinear 700
Bergan and Mollestad 709
B-H RELATION (electromagnetic, model definition
option) 310, 526
B-H RELATION (magnetostatic, model definition
option) 294, 295, 296, 526
Bingham fluid 317
block definition 54
body motion 536
boundary conditions
input options 626
loading types 598
time-dependent 177
treatment of 121
breaking glue 557
BUCKLE (history definition option) 116, 117
BUCKLE (parameter) 98, 115
BUCKLE INCREMENT (model definition option) 98,
116, 117
buckling analysis 115
bushing
coordinate system 647
bushings 646
nodal offsets for elements 647
properties 648
C
Cam-Clay model 501
capacitor 291
Carreau model 318
CASE COMBIN (model definition option) 95, 96,
711
CASI iterative solver 183, 726, 728, 729, 730
Index
789
cavity
pressure loading 615
radiating 271
rigid, acoustic analysis 314
surface elements 619, 696
CAVITY DEFINITION (model definition option) 265,
274, 276, 622
central difference operator 171, 174, 175
CENTROID (parameter) 106, 109, 115, 279, 620,
686, 711, 722, 742, 748
CFAST (model definition option) 646, 663, 679, 680,
681, 682, 683, 684
Chaboche 461, 475
CHANGE PORE (model definition option) 191
CHANGE STATE (history definition option) 137,
138, 701
CHANGE STATE (model definition option) 31, 221,
279, 620
CHANNEL (model definition option) 221, 326
classical lamination theory, multi-Layered shells
353
closest point projection procedure 129
code coupling
basic stages 773
software 771
COEFFICIENT (model definition option) 649
COHESIVE (model definition option) 693
COHESIVE (with TABLE Input - Thermal) 693
cohesive zone 516
coil current 299
COIL CURRENT (magnetostatic, model definition
option) 300, 303, 343
coking 241
linear model 243
water drying model 244
combined hardening 456
common material characteristics 117
COMPOSITE (model definition option) 201, 349,
350
composite continuum elements 350
composite material 349
conditioning number 728
CONN FILL (model definition option) 65
CONN GENER (model definition option) 64
CONNECT (model definition option) 59
connection methods 660
CONNECTIVITY (model definition option) 54, 64,
326, 327, 342, 343, 344, 646, 663, 670, 675, 676,
679, 711
Conrad Gap 277
CONRAD GAP (model definition option) 221
constant dilatation elements 695
constitutive relations 459
constraint
rigid link 638
CONSTRAINT (model definition option) 58
contact
adaptive meshing 569
analysis 749
analytical 575
beams 566
bodies 530
constraint implementation 542
coupled 561
deformable 579
detection 538
dynamic 568
friction modeling 544
mathematical aspects 572
neighbor relations 541
optimize ordering 533
quadratic 566
release 558
segment-to-segment 583
separation 558
shell 540, 568
thermal 563
tolerance values 571
CONTACT (model definition option) 113, 114, 121,
171, 221, 264, 335, 336, 530, 532, 555, 566, 568,
573, 598, 644, 650, 704
CONTACT CHANGE (rezoning option) 569
CONTACT NODE (model definition option) 530
contact penetration criteria, remeshing 85
CONTACT TABLE (history definition option) 538,
539, 558, 559
CONTACT TABLE (model definition option) 543,
544, 553, 554, 555, 557, 563, 565, 568, 581
CONTINUE (history definition option) 40, 46, 47
continuum composite elements 695
interlaminar stresses 379
continuum elements 688
CONTROL (electromagnetic, model definition
option) 306
CONTROL (heat transfer, model definition option)
327
CONTROL (history definition option) 137, 222, 653,
658, 700
CONTROL (magnetostatic, model definition option)
294
CONTROL (model definition option) 107, 117, 221,
222, 271, 571, 653, 658, 700, 705, 739
control algorithm 67
convergence controls 163
CONVERT (model definition option) 328, 342, 343,
344
COORD SYSTEM (model definition option) 358,
537, 647
COORDINATES (model definition option) 54, 64,
675, 711, 745
Coulomb friction model 544
arctangent model 546
bilinear model 550
stick-slip (modified step function) model 547
COUPLE (parameter) 326, 342
coupled acoustic-structural analysis 328
acoustic medium 331
divergence theorem of Green 329
fluid pressure 330
harmonic analyses 329
coupled analysis 324, 561
coupled contact analysis 561
coupled electrical-thermal-mechanical 340
coupled electrical-thermal-mechanical analysis 340
coupled electrostatic-structural
analysis 526
coupled emectromagnetic-thermal analysis 343
coupled thermo-electrical analysis (Joule heating)
337
coupled thermo-mechanical analysis 326
coupling interface 771
coupling region
parallel processing 775
shell elements 775
time step 775
coupling regions 773
crack analysis 140
CRACK DATA (model definition option) 498
cracking
closure 500
low tension 499
uniaxial 498
Craig-Bampton 755
creep 133
buckling 137
control tolerances 135
implicit formulation 474
Maxwell model 468
CREEP (model definition option) 139, 468, 469,
474, 479, 494
CREEP (parameter) 137, 138, 139, 467, 474, 478
creep analysis 701
CREEP INCREMENT (history definition option) 700
Crisfield 717, 719
criteria
equivalent values 78
mean strain energy 77
node within a box 79
nodes in contact 79
previously refined mesh 80
remeshing 84
temperature gradient 79
user-defined 79
Zienkiewicz-Zhu 77
Zienkiewicz-Zhu creep strain 78
CRPLAW (user subroutine) 134, 135, 139, 468, 470,
473, 475, 479, 494
CRPVIS (user subroutine) 138, 467, 479
crushing 500
cure kinetics 212
CURE SHRINKAGE (model definition option) 212
CURVES (model definition option) 76, 358
Cutter Location (CL) 783
CWELD (model definition option) 663, 668, 669,
670, 671, 672, 673, 674, 675, 676, 678, 679, 680,
681, 682, 683, 684
CWELD patch-to-patch connections 663
direct 663
indirect 665
cyclic loading 619
Index
791
cyclic symmetry
tying constraints 649
CYCLIC SYMMETRY (model definition option) 650,
651
CYLINDRICAL (model definition option) 53, 658
D
DAMAGE (model definition option) 384, 516
damage models
cohesive zone 516
ductile metals 512
elastomers 514
damping 178, 179
DAMPING (model definition option) 165, 168, 179
DAMPING COMPONENTS (history definition
option) 286
deact glue 557
DEACT GLUE (model definition option) 150, 153,
557
deactivate 54
element method 261
DEFINE (model definition option) 39, 40, 41
deformable contact, solution strategy 580
deformable-deformable contact 579
degree of freedom transformation 627
DELAMINATION (model definition option) 158, 160
DENSITY EFFECTS (model definition option) 439
DESIGN DISPLACEMENT CONSTRAINTS (model
definition option) 201
DESIGN FREQUENCY CONSTRAINTS (model
DESIGN OBJECTIVE (model definition option) 201
design optimization 197
DESIGN OPTIMIZATION (parameter) 201, 348
DESIGN SENSITIVITY (parameter) 201, 348
design sensitivity analysis 194
design space 198
DESIGN STRAIN CONSTRAINTS (model definition
option) 201
DESIGN STRESS CONSTRAINTS (model
DESIGN VARIABLES (model definition option) 201
diffusion 280
DIFFUSION (parameter) 280
dilatational creep 474
direct
constraints 572
integration 170
mesh definition input 49
methods 727
DISP CHANGE (history definition option) 138, 327,
342, 642
DIST CHARGE (electromangetic, model definition
option) 306
DIST CHARGE (model definition option) 311, 335
DIST CHARGES (electrostatic, model definition
option) 290, 335, 344
DIST CURRENT (electromagnetic, model definition
option) 306
DIST CURRENT (history definition option) 342
DIST CURRENT (Joule, history definition option)
337, 344
DIST CURRENT (magnetostatic, model definition
option) 294, 343
DIST CURRENT (model definition option) 342
DIST FLUXES (history definition option) 337, 342,
344
DIST FLUXES (model definition option) 188, 220,
225, 286, 337, 342, 343
DIST LOADS (history definition option) 138, 342
DIST LOADS (model definition option) 188, 311,
335, 336, 342, 612, 614
DIST LOADS (parameter) 201
DIST MASS (model definition option) 281
DIST SOURCE (model definition option) 315
Domain Decomposition Method (DDM) 760, 766
ductile metals 512
dynamic
analysis 748
crack propagation 151, 157
fracture methodology 156
impact 568
DYNAMIC (parameter) 98, 157, 161, 168, 170
DYNAMIC CHANGE (electromagnetic, history
definition option) 306, 309
DYNAMIC CHANGE (history definition option) 168,
170, 176, 315, 332, 700
E
eddy 303
Edge I-J orientation 355
effects of temperature and strain rates on yield stress
451
eigenvalue analysis 161
ELASTIC (parameter) 74, 94, 95
elastic models 501
ELASTICITY (parameter) 95
elastomeric material 418
elastomers 514
electrical
analysis 622
input options 623
ELECTRO (parameter) 290, 334
electromagnetic
analysis 305
input options 625
electrostatic
analysis 525, 623, 691, 747, 748
Coulomb force 334
coupled electrostatic-structural 333, 526, 624,
691
element types 289
input options 623
electrostatic-structural
analysis 747, 749
ELEM SORT (model definition option) 741
element distortion criteria, remeshing 84
elements
acoustic analysis 314, 331
beam 689, 745
beams, contact 566
cavity surface 696
classes of stress-strain relations 351
composite continuum 350
connectivity data 49
constant dilatation 695
contact 565
continuum 688
continuum composite 695
deactivate 261
electrostatic 289
fluid 323
Fourier analysis 97
gap-and-friction 692, 746
groups 757
heat transfer 690
heat transfer convection 279
hydrodynamic bearing 285
incompressible 694
interface 160, 516, 572, 693
magnetodynamic 305
magnetostatic 293
membrane 688
piezoelectric 311
plate 690
quarter point 145
rebar 693
reduced integration 695
semi-infinite 696
shell 568, 690
special 692
truss 687
ELEMENTS (parameter) 160, 663, 676, 679, 686
ELEVAR (user subroutine) 742, 745
ELEVEC (user subroutine) 742
EL-MA (parameter) 306, 344
ELSTO (parameter) 28, 29, 30, 31, 735
EMCAPAC (history definition option) 291, 565
EMINDUC (history definition option) 303
EMISSIVITY (model definition option) 274
EMLAMIN (history definition option) 305, 343
EMRESIS (history definition option) 339, 564
EMWINDING (magnetostatic, model definition
option) 300, 303, 343
END (parameter) 46
END OPTION (model definition option) 38, 46, 47,
201, 315
END REZONE (rezoning option) 47
END SEQUENCE (history definition option) 620
Equivalent Values Criterion 78
error analysis 73
ERROR ESTIMATE (model definition option) 73,
95, 96, 121
Eulerian formulation 111
examples of lists input 41
EXCLUDE (history definition option) 539
exponential cap 438, 440, 446, 450
EXTENDED (parameter) 39
Index
793
F
FAIL DATA (model definition option) 358, 379, 380,
381, 498
FEATURE (parameter) 653, 658
file units 30
FILM (user subroutine) 221, 226, 228, 337, 342, 343,
344, 622
FILMS (model definition option) 220, 226, 236, 264,
337, 342, 343, 344, 622
FILMS (with TABLE input, model definition option)
622
finite element technology
basic concepts 777
FIXED DISP (model definition option) 54, 58, 188,
311, 327, 332, 335, 336, 342, 642
FIXED POTENTIAL (electromagnetic, model
FIXED POTENTIAL (electrostatic, model definition
option) 290, 344
FIXED POTENTIAL (magnetostatic, model
FIXED PRESSURE (model definition option) 188,
281, 286, 315
FIXED TEMPERATURE (model definition option)
220, 225, 274, 324, 327, 337, 342, 343, 344
FIXED VELOCITY (model definition option) 324
FIXED VOLTAGE (model definition option) 337, 342
FLD (Forming Limit Diagram) 750
flow
diagram 731
rule 458
FLP (Forming Limit Parameter) 750
fluid
analysis 315, 526, 692
elements 323
FLUID (parameter) 692
FLUID DRAG (model definition option) 614
fluid drag and wave
input options 614
loads 614
fluid mechanics 315
Bingham fluid 317
Carreau model 318
element types 323
piecewise non-Newtonian flow 317
power law fluid 318
FLUID SOLID (model definition option) 332
fluid/solid interaction analysis 331, 525, 691
FLUX (user subroutine) 221, 225, 227, 286, 290,
294, 311, 315, 335, 337, 342, 343, 344, 622
FOAM (model definition option) 488
FOLLOW FOR (parameter) 106, 115, 612, 630, 696
follow force stiffness contribution 696
FORCDF (user subroutine) 344
FORCDT (user subroutine) 177, 225, 227, 287, 290,
294, 311, 324, 335, 336, 343
force components 598
FORCEM (user subroutine) 75, 177, 306, 311, 335,
337, 343, 344
Forming Limit Diagram (FLD) 750
Forming Limit Parameter (FLP) 750
FOUNDATION (model definition option) 95, 115
FOURIER (model definition option) 95, 96, 98
FOURIER (parameter) 96, 98
Fourier analysis 96
coefficients 97
elements used 97
fractal dimension 728
fracture mechanics 140
friction modeling 544
FXORD (model definition option) 59, 60, 64, 690
G
gap-and-friction element 692, 746
gasket 385
GASKET (model definition option) 388
generalized alpha operator 171, 174
HHT 174
spectral radius 175
generation phase 157
GENT (parameter) 109
geometric domains 686
GEOMETRY (model definition option) 69, 201, 286,
657, 660, 673, 674, 676, 690, 695, 696, 745
GLK (model definition option) 209
global
meshing 569
ZX plane orientation 356
GLOBALLOCAL (model definition option) 31, 209
glue model, friction 557
gradient
objective or response function 194
pressure 188
Gurson model, modified 512
H
harmonic
analysis 310
response 164
HARMONIC (electromagnetic, history definition
option) 310
HARMONIC (electromagnetic, history definition)
344
HARMONIC (parameter) 164, 165, 167
Hashin
fabric failure criterion 366
failure criterion 365
tape failure criterion 367
HEAT (parameter) 279, 343, 344, 691
heat fluxes 561, 621
(special) conditions 621
input options 621
heat transfer
analysis 220, 524, 621, 746, 748
convection 279
convection elements 279
elements 690
temperature effects 224
hereditary integral model 479
Hill
failure criterion 361
yield function 432
history definition options 46
Hoffman failure criterion 362
HOOKLW (user subroutine) 95, 348, 350
HOOKVI (user subroutine) 483
host file 34, 730, 764
Houbolt operator 170
hydrodoynamic bearinh 747
hydrodynamic bearing
analysis 525, 622, 691, 747, 748
elements 285
mass fluxes 622
pump pressures 622
restrictors 622
HYPELA2 (user subroutine) 106, 390, 501
HYPOELASTIC (model definition option) 390
hysteresis 303, 413, 490, 515
I
IBOOC (parameter) 28, 29, 31
IMPD (user subroutine) 742
incompressible
Herrmann elastic formulation 120
material behavior 119
material modeled 120
incompressible elements 694
rigid-plastic flow 694
incremental mesh generators 64
induction
heating 343, 344
magnetic 43, 294
motors 303
inertia relief 180
loads evaluation 181
Support method 180
initial conditions 177
previous analysis result data 203
INITIAL DISP (model definition option) 177
INITIAL PC (model definition option) 191
initial plastic strain 621
INITIAL PORE (model definition option) 191
INITIAL POROSITY (model definition option) 188,
281
initial stress and initial plastic strain input options
621
INITIAL TEMP (model definition option) 220, 225,
337, 342, 343, 344
INITIAL VEL (model definition option) 177
INITIAL VOID RATIO (model definition option) 188
input conventions 38
input options
acoustic analysis sources 624
boundary conditions 626
electrical 623
electromagnetic 625
electrostatic 623
fluid drag and wave loads 614
Index
795
heat fluxes 621
initial stress and initial plastic strain 621
magnetodynamic 306
magnetostatic currents 624
mass fluxes and restrictors 623
mechanical loads 612
thermal loads 620
INSERT (model definition option) 645
INTCRD (user subroutine) 742
interface
ADAMS 755
hyperMesh results 754
SDRC I-DEAS results 754
interference fit 642
interpolation (shape) functions, type of 687
inverse power sweep 162
ISOTROPIC (acoustic, model definition option) 315,
526
ISOTROPIC (electromagnetic, model definition
option) 306, 526
ISOTROPIC (electrostatic, model definition option)
290, 525
ISOTROPIC (fluid, model definition option) 332,
525, 526
ISOTROPIC (heat transfer, model definition option)
337, 344, 524, 525, 620
ISOTROPIC (hydrodynamic, model definition
option) 525
ISOTROPIC (magnetostatic, model definition
option) 294, 295, 526
ISOTROPIC (model definition option) 39, 201, 220,
224, 240, 274, 275, 281, 286, 311, 324, 343, 348,
350, 381, 388, 432, 433, 437, 451, 453, 456, 457,
470, 473, 474, 475, 481, 483, 493, 498, 612
ISOTROPIC (with TABLE input - Stress) 452
isotropic hardening 455
J
J-INTEGRAL (model definition option) 95
Joule
analysis 748
heating 337
structural 340
JOULE (model definition option) 337, 342
JOULE (parameter) 337, 342
K
Kachanov factor 516
Kelvin-Voigt model 478
kinematic
constraints 625
hardening 455
L
Lagrange multipliers 572
Lagrangian formulation 103
Lanczos method 162
large deformations 118
LARGE DISP (parameter) 105, 106, 109, 115, 116,
121, 390, 452, 640
large strain
analysis 694
elasticity 118
plasticity 120
LARGE STRAIN (parameter) 105, 106, 109, 131,
171, 275, 327, 389, 452, 503, 630, 640
linear analysis 94
linear elastic material 347
linear fracture mechanics 140
linear springs and elastic foundations input options
646
linked/unlinked 201
list of items, input 39
load
AUTO STEP scheme 703
automatic incrementation 702
cavity pressure 615
fixed 702
fluid drag and wave 614
history definition options 700
increment size 702
mass fuxes 622
mechanical 611
mechanical and thermal increments 700
piezoelectric 624
pump pressures 622
recycling criterion 703
residual 748
residual load correction 710
restrictors 622
surface/volumetric 599
thermal 620
types 598
user-defined physical criteria 705
vectors 780
LOAD COR (parameter) 612
load incrementation
adaptive scheme 701
fixed scheme 700
Lorentz 335
LORENZI (model definition option) 143, 145, 157
losses 303, 342
M
machining (or metal cutting) 783
MAGNETO (parameter) 294, 336, 343
magnetodynamic
analysis 749
element types 305
input options 306
magnetodynamicic
analysis 692
magnetodynamics
analysis 747
magnetostatic
analysis 526, 624, 691, 747, 749
current input options 624
elements types 293
MANY TYPES (model definition option) 59
Marc
host systems 27
overview 23
mass fluxes and restrictors 622
MASSES (model definition option) 165
material
instabilities 122
low tension 498
mechanical shape memory 413
powder 438
preferred direction 354
material dependent failure criteria 358
Hashin 365
Hashin Fabric 366
Hashin Tape 367
Hill 361
Hoffman 362
maximum strain 360
maximum stress 359
Puck 368
SIFT 371
strain failure 360
Tsai-Wu 364
user-defined 365
material model
composite 349
elastomers 418
gasket 385
linear elastic 347
Mohr-Coulomb (hydrostatic yield dependence)
436
nonlinear hypoelastic 389
time-independent inelastic 428
Mean Strain Energy Criterion 77
mean-normal method 130
mechanical
erosion 237
erosion by other actions 238
loads 611
loads input options 612
mechanical shape memory model
conversation thermo-mechanical to mechanical
SMA 417
experimental data fitting 416
mechanical wear 192
membrane elements 688
MERGE (model definition option) 55
mesh
definition 49
refinement tying constraints 632
severe distortion 121
mesh generation, remeshing techniques 86
advancing front meshing 86
Delaunay Triangulation 87
overlay meshing 86
mesh splitting 158
MESH2D 54
MESH2D (model definition option) 59
mixture model 381
MNF (parameter) 756
MNF UNITS (model definition option) 757
Index
797
modal
stresses and reactions 164
superposition 164
MODAL INCREMENT (model definition option) 98
MODAL SHAPE (history definition option) 162, 163,
315, 332, 757
mode separation 145
model connections
CWELD patch-to-patch 663
model definition options 46
Mohr-Coulomb
linear material 436
parabolic material 437
MOONEY (model definition option) 109, 381, 422,
426, 493
MOONEY (parameter) 109
Mooney-Rivlin 422, 427
MOTION (2-D, user subroutine) 536
MOTION CHANGE (history definition option) 538,
558
MPC-CHECK (parameter) 645
MPI (Message Passing Interface) 760
Mullins effect 514
multiplicative decomposition 125, 428
multistage return mapping 131
MUMPS 34, 726, 760
N
Narayanaswamy model 138, 488
NASSOC (user subroutine) 139, 475
Nastran RBE2 and RBE3 652
Navier-Stokes 315, 318, 324, 326
NC machining data 783
cutter path 784
cutter shape 783
element deactivation 785
intersection 785
nearly incompressible behavior 121
neighbor relations 541
Neo-Hookean 422
Newmark-beta operator 170, 171
Newton Cooling 496
Newton-Raphson method
modified 713
NiTi alloys 404
NLELAST 390
(bimodulus) elasticity model 402
(bimodulus) elasticity model with either no
tension, limited tension, no compression, or
limited compression enhancements 402
basic assumptions and definitions 391
converting engineering strain/stress to true
strain/stress 391
input of uniaxial stress-strain data 391
Nastran like nonlinear elasticity model 392
nonlinear orthotropic elasticity model 404
principal strain space model 400
strain invariant model 396
NLELAST (model definition option) 106, 350, 390,
393, 394
nodal degrees of freedom tying constraints 638
NODE CIRCLE (model definition option) 65
NODE FILL (model definition option) 65
NODE GENER (model definition option) 65
NODE SORT (model definition option) 741
Node Within A Box Criterion 79
Nodes In Contact Criterion 79
nodes, merging of 55
nonlinear
analysis 99
fracture mechanics 142
nonlinear hypoelastic material 389
HYPELA2 user subroutine 390
NLELAST model definition 390
nonlinearities, three types 103
non-Newtonian 317
nonstructural materials 524
NURBS 531, 542, 543, 575, 576
trimmed 531, 570
O
Oak Ridge National Laboratory (ORNL) 134, 451,
473
objective function 194, 195, 200, 201, 202, 203
objective stress rates 123
OGDEN (model definition option) 423, 488, 493,
516
OGDEN (parameter) 109
operator
central difference 171
generalized alpha operator 171
Houbolt 170
Newmark-beta 170
single step Houbolt 171
OPTHOTROPIC (model definition option) 457
optimization algorithm 200
OPTIMIZE (model definition option) 65
Option
ACTIVATE (history definition) 54
ADAPT GLOBAL (model definition) 248, 557
ADAPTIVE (model definition) 77, 95, 96, 557,
697
ANISOTROPIC (model definition) 95, 281, 311,
348, 350, 354, 381, 433, 456, 457, 493
APPROACH (history definition) 538
ASSEM LOAD (model definition) 756
ATTACH EDGE (model definition) 76, 611
ATTACH FACE (model definition) 611
AUTO CREEP (history definition) 134, 135,
136, 138, 139, 479, 700, 701
AUTO INCREMENT (history definition) 67,
116, 500, 614, 700, 701, 702, 703, 718
AUTO LOAD (history definition) 67, 335, 337,
488, 614, 700
AUTO STEP (history definition) 138, 170, 222,
226, 335, 337, 343, 344, 483, 488, 500, 575,
582, 619, 700, 701, 702, 703, 704, 705, 706,
707, 709, 710
AUTO THERM (history definition) 620, 700,
701, 702
AUTO THERM CREEP (history definition) 137,
138, 700, 701
AXITO3D (model definition) 203
BEGIN SEQUENCE (history definition) 620
B-H RELATION (electromagnetic, model
definition) 310, 526
B-H RELATION (magnetostatic, model
definition) 294, 295, 296, 526
BUCKLE (history definition) 116, 117
BUCKLE INCREMENT (model definition) 98,
116, 117
CASE COMBIN (model definition) 95, 96, 711
CAVITY DEFINITION (model definition) 265,
274, 276, 622
CFAST (model definition) 646, 663, 679, 680,
681, 682, 683, 684
CHANGE PORE (model definition) 191
CHANGE STATE (history definition) 137, 138,
701
CHANGE STATE (model definition) 31, 221,
279, 620
CHANNEL (model definition) 221, 326
COEFFICIENT (model definition) 649
COHESIVE (model definition) 693
COIL CURRENT (magnetostatic, model
definition) 300, 303, 343
COMPOSITE (model definition) 201, 349, 350
CONN FILL (model definition) 65
CONN GENER (model definition) 64
CONNECT (model definition) 59
CONNECTIVITY (model definition) 54, 64, 326,
327, 342, 343, 344, 646, 663, 670, 675, 676,
679, 711
CONRAD GAP (model definition) 221
CONSTRAINT (model definition) 58
CONTACT (model definition) 113, 114, 121,
171, 221, 264, 335, 336, 530, 532, 555, 566,
568, 573, 598, 644, 650, 704
CONTACT CHANGE (rezoning) 569
CONTACT NODE (model definition) 530
CONTACT TABLE (history definition) 538, 539,
558, 559
CONTACT TABLE (model definition) 543, 544,
553, 554, 555, 557, 563, 565, 568, 581
CONTINUE (history definition) 40, 46, 47
CONTROL (electromagnetic, model definition)
306
CONTROL (heat transfer, model definition) 327
CONTROL (history definition) 137, 222, 653,
658, 700
CONTROL (magnetostatic, model definition)
294
CONTROL (model definition) 107, 117, 221,
222, 271, 571, 653, 658, 700, 705, 739
CONVERT (model definition) 328, 342, 343,
344
COORD SYSTEM (model definition) 358, 537,
647
COORDINATES (model definition) 54, 64, 675,
711, 745
Index
799
CRACK DATA (model definition) 498
CREEP (model definition) 139, 468, 469, 474,
479, 494
CREEP INCREMENT (history definition) 700
CURE SHRINKAGE (model definition) 212
CURVES (model definition) 76, 358
CWELD (model definition) 663, 668, 669, 670,
671, 672, 673, 674, 675, 676, 678, 679, 680,
681, 682, 683, 684
CYCLIC SYMMETRY (model definition) 650,
651
CYLINDRICAL (model definition) 53, 658
DAMAGE (model definition) 384, 516
DAMPING (model definition) 165, 168, 179
DAMPING COMPONENTS (history definition)
286
DEACT GLUE (model definition) 150, 153
DEFINE (model definition) 39, 40, 41
DELAMINATION (model definition) 158, 160
DENSITY EFFECTS (model definition) 439
DESIGN DISPLACEMENT CONSTRAINTS
(model definition) 201
DESIGN FREQUENCY CONSTRAINTS (model
definition) 201
DESIGN OBJECTIVE (model definition) 201
DESIGN STRAIN CONSTRAINTS (model
definition) 201
DESIGN STRESS CONSTRAINTS (model
definition) 201
DESIGN VARIABLES (model definition) 201
DISP CHANGE (history definition) 138, 327,
342, 642, 644
DIST CHARGE (electromangetic, model
definition) 306
DIST CHARGE (model definition) 311, 335
DIST CHARGES (electrostatic, model
DIST CURRENT (electromagnetic, model
definition) 306
DIST CURRENT (history definition) 342
DIST CURRENT (Joule, history definition) 337,
344
DIST CURRENT (magnetostatic, model
DIST CURRENT (model definition) 342
DIST FLUXES (history definition) 337, 342, 344
DIST FLUXES (model definition) 188, 220,
225, 286, 337, 342, 343
DIST LOADS (history definition) 138, 342
DIST LOADS (model definition) 188, 311, 335,
336, 342, 612, 614
DIST MASS (model definition) 281
DIST SOURCE (model definition) 315
DYNAMIC CHANGE (history definition,
electromagnetic) 306, 309
DYNAMIC CHANGE (history definition) 168,
170, 176, 315, 332, 700
ELEM SORT (model definition) 741
EMCAPAC (history definition) 291, 565
EMINDUC (history definition) 303
EMISSIVITY (model definition) 274
EMLAMIN (history definition) 305, 343
EMRESIS (history definition) 339, 564
EMWINDING (magnetostatic, model definition)
300, 303, 343
END OPTION (model definition) 38, 46, 47,
201, 315
END REZONE (rezoning) 47
END SEQUENCE (history definition) 620
ERROR ESTIMATE (model definition) 73, 95,
96, 121
EXCLUDE (history definition) 539
FAIL DATA (model definition) 358, 379, 380,
381, 498
FILMS (model definition) 220, 226, 236, 264,
337, 342, 343, 344, 622
FILMS (with TABLE input, model definition)
622
FIXED DISP (model definition) 54, 58, 188,
311, 327, 332, 335, 336, 342, 642
FIXED POTENTIAL (electromagnetic, model
definition) 306
FIXED POTENTIAL (electrostatic, model
FIXED POTENTIAL (magnetostatic, model
FIXED PRESSURE (model definition) 188,
281, 286, 315
FIXED TEMPERATURE (model definition) 220,
225, 274, 324, 327, 337, 342, 343, 344
FIXED VELOCITY (model definition) 324
FIXED VOLTAGE (model definition) 337, 342
FLUID DRAG (model definition) 614
FLUID SOLID (model definition) 332
FOAM (model definition) 488
FOUNDATION (model definition) 95, 115
FOURIER (model definition) 95, 96, 98
FXORD (model definition) 59, 60, 64, 690
GASKET (model definition) 388
GEOMETRY (model definition) 69, 201, 286,
657, 660, 673, 674, 676, 690, 695, 696, 745
GLK (model definition) 209
GLOBALLOCAL (model definition) 31, 209
HARMONIC (electromagnetic, history
HYPOELASTIC (model definition) 390
INITIAL DISP (model definition) 177
INITIAL PC (model definition) 191
INITIAL PORE (model definition) 191
INITIAL POROSITY (model definition) 188, 281
INITIAL TEMP (model definition) 220, 225, 337,
342, 343, 344
INITIAL VEL (model definition) 177
INITIAL VOID RATIO (model definition) 188
INSERT (model definition) 645
ISOTROPIC (acoustic, model definition) 315,
526
ISOTROPIC (electromagnetic, model
ISOTROPIC (electrostatic, model definition)
290, 525
ISOTROPIC (fluid, model definition) 332, 525,
526
ISOTROPIC (heat transfer, model definition)
337, 344, 524, 525, 620
ISOTROPIC (hydrodynamic, model definition)
525
ISOTROPIC (magnetostatic, model definition)
294, 295, 526
ISOTROPIC (model definition) 39, 201, 220,
224, 240, 274, 275, 281, 286, 311, 324, 343,
348, 350, 381, 388, 432, 433, 437, 451, 453,
456, 457, 470, 473, 474, 475, 481, 483, 493,
498, 612
ISOTROPIC (with TABLE input - Stress) 452
ISOTROPIC (with TABLE input - thermal) 343
J-INTEGRAL (model definition) 95
JOULE (model definition) 337, 342
LORENZI (model definition) 143, 145, 157
MANY TYPES (model definition) 59
MASSES (model definition) 165
MERGE (model definition) 55
MESH2D (model definition) 59
MNF UNITS (model definition) 757
MODAL INCREMENT (model definition) 98
MODAL SHAPE (history definition) 162, 163,
315, 332, 757
MOONEY (model definition option) 109
MOONEY (model definition) 381, 422, 426, 493
MOTION CHANGE (history definition) 538, 558
NLELAST (model definition) 106, 350, 390,
393, 394
NODE CIRCLE (model definition) 65
NODE FILL (model definition) 65
NODE GENER (model definition) 65
NODE SORT (model definition) 741
OGDEN (model definition) 423, 488, 493, 516
OPTIMIZE (model definition) 65
ORIENTATION (model definition) 348, 350,
354, 433, 483, 686
ORTHO TEMP (model definition) 220, 337,
350, 451, 493
ORTHOTROPIC (electrical, model definition)
525
ORTHOTROPIC (electromagnetic, model
ORTHOTROPIC (electrostatic, model
ORTHOTROPIC (magnetostatic, model
ORTHOTROPIC (model definition) 95, 201,
220, 240, 274, 275, 281, 311, 348, 350, 381,
432, 433, 451, 453, 456, 457, 474, 475, 481,
483, 493, 495, 612
ORTHOTROPIC (thermal, model definition)
337, 344, 524, 525, 620
ORTHOTROPIC (with TABLE input - Stress)
452
ORTHOTROPIC (with TABLE input - thermal)
343
PARAMETERS (history definition) 172
Index
801
PARAMETERS (model definition) 183, 185,
275, 324, 543, 562, 589
PBUSH (model definition) 646, 648, 649
PERMANENT (magnetostatic, model
PFAST (model definition) 646, 648, 663, 679
PIEZOELECTRIC (model definition) 311
PIN CODE (model definition) 660
POINT CHARGE (electrostatic, model
POINT CHARGE (model definition) 290, 311
POINT CHARGE (with TABLE Input, model
POINT CURRENT (electromagnetic, history
definition) 306
POINT CURRENT (history definition) 342
POINT CURRENT (Joule, history definition)
337
POINT CURRENT (Joule, model definition) 337
POINT CURRENT (magnetostatic, model
POINT CURRENT (model definition) 342
POINT CURRENT (with TABLE Input -
Magnetostatic, history definition) 344
POINT FLUX (history definition) 337, 342, 343
POINT FLUX (model definition) 188, 220, 225,
286, 337, 342, 344
POINT LOAD (electrostatic, model definition)
335, 336
POINT LOAD (history definition) 138, 342
POINT LOAD (model definition) 39, 188, 311,
342, 536
POINT MASS (model definition) 281
POINT SOURCE (model definition) 315
POINTS (model definition) 76
POROSITY CHANGE (model definition) 385
POST (model definition) 138, 221, 279, 379,
702, 750
POTENTIAL CHANGE (history definition) 344
POWDER (model definition) 439
PRE STATE (model definition) 31, 203, 204
PRINT CHOICE (model definition) 167, 376,
740, 742, 745, 748
PRINT ELEMENT (model definition) 376, 740,
742, 745
PRINT NODE (model definition) 740, 742, 748
PRINT VMASS (model definition) 741
PROPORTIONAL INCREMENT (history
definition) 67, 138, 139, 614
PSHELL (model definition) 354, 385
PWELD (model definition) 663, 664, 667, 672,
673, 674, 675, 676
RAD-CAVITY (model definition) 236, 274, 275,
622
RADIATING CAVITY (model definition) 265
RBE2 (model definition) 150, 153, 653, 660,
669, 670
RBE3 (model definition) 654, 655, 656, 660,
662, 665, 669, 670, 679, 684
REAUTO (model definition) 134, 711
REBAR (model definition) 498
RECEDING SURFACE (model definition) 248
RECOVER (history definition) 164
RELATIVE DENSITY (model definition) 439
RELEASE (history definition) 558
RESTART (model definition) 30, 95, 134, 711,
741, 742
RESTART INCREMENT (history definition) 711
RESTART LAST (model definition) 711
RESTRICTOR (model definition) 286
REZONE (rezoning) 47
ROTATION A (model definition) 612
RROD (model definition) 150, 153
SERVO LINK (model definition) 597, 631
SHELL TRANSFORMATION (model definition)
627, 630
SHIFT FUNCTION (model definition) 138, 486
SOIL (model definition) 188, 191, 501
SOLVER (model definition) 696, 726, 727, 728
SPECIFIED NODES (model definition) 59
SPLINE (model definition) 249, 531, 579, 580,
583, 587, 592, 593, 594
SPRINGS (model definition) 95, 115, 165
START NUMBER (model definition) 59
STEADY STATE (electrostatic, history
definition) 290
STEADY STATE (history definition) 222, 321,
337, 343, 344
STEADY STATE (magnetostatic, history
definition) 294
STIFFNS COMPONENTS (history definition)
286
STIFFSCALE (model definition) 68
STRAIN RATE (model definition, fluid) 526
STRAIN RATE (model definition) 185, 324,
350, 451, 452
SUMMARY (model definition) 741
SUPERELEM (history definition) 757
SUPERELEM (model definition) 757
SURFACE ENERGY (model definition) 230,
248
SURFACES (model definition) 76
SWLDPRM (model definition) 663, 675, 678,
680
SYNCHRONIZED (history definition) 538
TABLE (model definition) 274, 275, 305, 387,
456, 457, 481, 493, 495
TEMP CHANGE (history definition) 220, 225,
324, 327, 337, 342, 343, 344
TEMPERATURE EFFECTS (coupled thermal-
stress, model definition) 525
TEMPERATURE EFFECTS (heat transfer,
model definition) 337, 525
TEMPERATURE EFFECTS (hydrodynamic,
model definition) 525
TEMPERATURE EFFECTS (model definition)
220, 224, 275, 286, 350, 422, 439, 451, 452,
481, 486, 493, 495, 526
THERMAL CONTACT (2-D) (model definition)
281, 303, 304, 305, 564
THERMAL CONTACT (model definition) 221,
264, 291, 299, 339
THERMAL CONTACT(2-D) (model definition)
564, 565
THERMO-PORE (model definition) 240
THICKNESS (model definition) 286
THICKNS CHANGE (history definition) 286
TIME STEP (history definition) 451, 483, 488
TIME-TEMP (model definition) 496, 497
TRANSFORMATION (model definition) 537,
627, 634, 644, 658, 748
TRANSIENT (history definition) 221, 222, 223,
226, 337, 343, 344, 700, 701
TYING (model definition) 221, 332, 597, 631
TYING CHANGE (history definition) 631
UFCONN (model definition) 54
UFXORD (model definition) 54
UTRANFORM (model definition) 627
VCCT (model definition) 140, 146, 151
VELOCITY (model definition) 221, 279, 286
VELOCITY CHANGE (history definition) 221,
279, 324
VIEW FACTOR (model definition) 268, 622
VISCEL EXP (model definition) 138, 492
VISCELFOAM (model definition) 138, 488
VISCELMOON (model definition) 138, 422, 488
VISCELOGDEN (model definition) 138, 423,
488, 516
VISCELORTH (model definition) 483
VISCELPROP (model definition) 483
VOID CHANGE (model definition) 385
VOLTAGE CHANGE (history definition) 337,
342
WELD FILL (model definition) 257, 261, 262
WELD FLUX (model definition) 257, 262, 263
WELD PATH (model definition) 257, 259
WORK HARD (model definition) 185, 350, 452,
456, 457
ORIENT (user subroutine) 95, 348, 433, 483
ORIENT2 (user subroutine) 354, 355, 358
ORIENTATION (model definition option) 348, 350,
354, 433, 483, 686
ORTHO TEMP (model definition option) 220, 337,
350, 451, 493
ORTHOTROPIC (electrical, model definition
option) 525
ORTHOTROPIC (electromagnetic, model definition
option) 306, 526
ORTHOTROPIC (electrostatic, model definition
option) 290, 335, 336
ORTHOTROPIC (magnetostatic, model definition
option) 294, 295, 526
ORTHOTROPIC (model definition option) 95, 201,
220, 240, 274, 275, 281, 311, 348, 350, 381, 432,
433, 451, 453, 456, 457, 474, 475, 481, 483, 493,
495, 612
ORTHOTROPIC (thermal, model definition option)
337, 344, 524, 525, 620
ORTHOTROPIC (with TABLE input - Stress) 452
ORTHOTROPIC (with TABLE input - thermal) 343
out-of-core storage options 29
Index
803
output results
workspace information 735, 771
overclosure 642
tying 642
P
parallel processing
contact analysis 763
domain decomposition 763
solvers 762
types of machines 760
unsupported features 761
Parameter
ABLATION 248
ACOUSTIC 315
ADAPTIVE 697
ALL POINTS 279, 686
ALLOCATE 29
ASSUMED STRAIN 696
AUTOMSET 645
AUTOSPC 558
BEAM SECT 69, 675, 689
BEARING 286
BUCKLE 98, 115
CENTROID 106, 109, 115, 279, 620, 686, 711,
722, 742, 748
COUPLE 326, 342
CREEP 137, 138, 139, 467, 474, 478
DESIGN OPTIMIZATION 201, 348
DESIGN SENSITIVITY 201, 348
DIFFUSION 280
DIST LOADS 201
DYNAMIC 98, 157, 161, 168, 170
ELASTIC 74, 94, 95
ELASTICITY 95
ELECTRO 290, 334
ELEMENTS 160, 663, 676, 679, 686
EL-MA 306, 344
ELSTO 28, 29, 30, 31, 735
END 46
EXTENDED 39
FEATURE 653, 658
FLUID 692
FOLLOW FOR 106, 115, 612, 630, 696
FOURIER 96, 98
GENT 109
HARMONIC 164, 165, 167
HEAT 279, 343, 344, 691
IBOOC 28, 29, 31
JOULE 337, 342
LARGE DISP 105, 106, 109, 115, 116, 121, 390,
452, 640
LARGE STRAIN 105, 106, 109, 131, 171, 275,
327, 389, 452, 503, 630, 640
LOAD COR 612
MAGNETO 294, 336, 343
MNF 756
MOONEY 109
MPC-CHECK 645
OGDEN 109
PIEZO 311
PORE 280
PRINT 117, 230, 274, 549, 749
RADIATION 264, 268, 525, 622
RBE 670
RESTRICTOR 286
REZONING 66
R-P FLOW 183, 715
SCALE 700
SHELL SECT 690
SIZING 686, 738
STATE VARS 286
STOP 29
STRUCTURAL 335, 336
TSHEAR 377, 745
T-T-T 496
UFLUXMEC 230
UPDATE 171
VISCO ELAS 138
parameters 46
PARAMETERS (history definition option) 172
PARAMETERS (model definition option) 183, 185,
275, 324, 543, 562, 589
Pardiso 726, 760
particle tracking 31
PBUSH (model definition option) 646, 648, 649
penalty 183, 321
methods 572
PERMANENT (magnetostatic, model definition
option) 294, 296
perturbation analysis 117
PFAST (model definition option) 646, 648, 663, 679
piecewise linear representation 493
PIEZO (parameter) 311
piezoelectric
analysis 310, 525, 624, 692, 747, 749
coupling 313
element types 311
loads 624
PIEZOELECTRIC (model definition option) 311
pin code 660
PIN CODE (model definition option) 660
pipe bend element tying constraints 637
plate elements 690
PLOTV (user subroutine) 750
POINT CHARGE (electrostatic, model definition
option) 306, 335
POINT CHARGE (model definition option) 290, 311
POINT CHARGE (with TABLE Input - Electrostatic,
model definition option) 344
POINT CURRENT (electromagnetic, history
POINT CURRENT (history definition option) 342
POINT CURRENT (Joule, history definition option)
337
POINT CURRENT (Joule, model definition option)
337
POINT CURRENT (magnetostatic, model definition
option) 294, 343
POINT CURRENT (model definition option) 342
POINT CURRENT (with TABLE Input -
Magnetostatic, history definition option) 344
POINT FLUX (history definition option) 337, 342,
343
POINT FLUX (model definition option) 188, 220,
225, 286, 337, 342, 344
POINT LOAD (electrostatic, model definition option)
335, 336
POINT LOAD (history definition option) 138, 342
POINT LOAD (model definition option) 39, 188,
311, 342, 536
POINT MASS (model definition option) 281
POINT SOURCE (model definition option) 315
POINTS (model definition option) 76
Poissons 94, 347, 348, 349, 350, 439, 493, 496, 498,
501, 503
PORE (parameter) 280
POROSITY CHANGE (model definition option) 385
POST (model definition option) 138, 221, 279, 379,
702, 750
post buckling analysis 701
post file 749
POTENTIAL CHANGE (history definition) 344
powder 100
POWDER (model definition option) 439
powder material 438
Power Law Fluid 318
PRE STATE (model definition option) 31, 203, 204
preconditioners, iterative solvers 728
presentation of the energy equation 245
Previously Refined Mesh Criterion 80
Principal Engineering Strains 750
PRINT (parameter) 117, 230, 274, 549, 749
PRINT CHOICE (model definition option) 167, 376,
740, 742, 745, 748
PRINT ELEMENT (model definition option) 376,
740, 742, 745
PRINT NODE (model definition option) 740, 742,
748
PRINT VMASS (model definition option) 741
program
initiation 33
messages 753
progressive composite failure 379
PROPORTIONAL INCREMENT (history definition
option) 67, 138, 139, 614
PSHELL (model definition option) 354, 385
PSPG 318
Puck failure criterion 368
PWELD (model definition option) 663, 664, 667,
672, 673, 674, 675, 676
PWL (piece-wise linear) 531
pyrolysis 238
alternative formulation 241
presentation of the mass equation 239
principle of 239
Q
quadratic contact 566
quarter point elements 145
Index
805
R
RAD-CAVITY (model definition option) 236, 274,
275, 622
radial return 129
RADIATING CAVITY (model definition option) 265
radiation 263
computational approaches 264
direct adaptive integration 271
Monte Carlo method 268
pixel based modified Hemi-Cube method 265
viewfactors 265
RADIATION (parameter) 264, 268, 525, 622
radiation viewfactors 265
RBE (parameter) 670
RBE2 (model definition option) 150, 153, 653, 660,
669, 670
RBE2 and RBE3 652
RBE3 (model definition option) 654, 655, 656, 660,
662, 665, 669, 670, 679, 684
reaction forces 748
REAUTO (model definition option) 134, 711
REBAR (model definition option) 498
REBAR (user subroutine) 498
rebar elements 693
RECEDING SURFACE (model definition option)
248
RECOVER (history definition option) 164
reduced integration elements 695
RELATIVE DENSITY (model definition option) 439
RELEASE (history definition option) 558
remeshing
automatic global 80
criteria 84
remeshing techniques 85
adaptive 785
mesh generation 86
outline extraction and repair 87
target number of elements 89
residual load correction 710
resistance 337, 339, 564, 648, 749
resistor 339
RESTART (model definition option) 30, 95, 134,
711, 741, 742
restart analysis 711
RESTART INCREMENT (history definition option)
711
RESTART LAST (model definition option) 711
RESTRICTOR (model definition option) 286
RESTRICTOR (parameter) 286
result evaluation, contact 570
REZONE (rezoning option) 47
rezoning 66, 569
REZONING (parameter) 66
rigid contact, solution strategy 577
rigid link constraint 639, 644
rigid-plastic analysis 183, 694, 748
Riks-Ramm 719, 720
ROTATION A (model definition option) 612
R-P FLOW (parameter) 183, 715
RROD (model definition option) 150, 153
rubber foam model 425
RVCNST 546
S
SCALE (parameter) 700
scaling, linear-elastic solution 700
segment-to-segment 583
selective printout 740
grid force balance 741
options 740
semi-infinite elements 696
sensitivity 194
separation, contact 558
SEPFOR (user subroutine) 578
SEPSTR (user subroutine) 578
Sequential Quadratic Programming 200
SERVO LINK (model definition option) 597, 631
shape memory
mechanical 413
thermo-mechanical material 404
shell
contact 540
elements 568, 690
stiffener tying constraints 637
transformation 627
SHELL SECT (parameter) 690
shell stiffener tying constraints 637
SHELL TRANSFORMATION (model definition
option) 627, 630
shell-to-shell intersection tying constraints 633
shell-to-solid intersections tying constraints 634
SHIFT FUNCTION (model definition option) 138,
486
single input
DDM 34, 761, 763, 764, 765
single step Houbolt operator 171
singularity ratio 724
SIZING (parameter) 686, 738
snap-through analysis 701
SOIL (model definition option) 188, 191, 501
soil analysis 188, 500, 692
solver
conjugate gradient iterative 762
direct 762
hardware 762
matrix 762
multifrontal 762
SOLVER (model definition option) 696, 726, 727,
728
special elements 692
cable 693
curved-pipe 693
Gap-and-Friction 692
pipe-bend 693
rebar 693
shear panel 693
SPECIFIED NODES (model definition option) 59
spectrum response 167
SPLINE (model definition option) 249, 531, 579,
580, 583, 587, 592, 593, 594
SPRINGS (model definition option) 95, 115, 165
springs, linear and nonlinear 746
START NUMBER (model definition option) 59
STATE VARS (parameter) 286
status file 757
STEADY STATE (electrostatic, history definition
option) 290
STEADY STATE (history definition option) 222,
321, 337, 343, 344
STEADY STATE (magnetostatic, history definition
option) 294
steady state rolling analysis 204
Stefan-Boltzmann 622
stick-slip model, friction 549
STIFFNS COMPONENTS (history definition option)
286
STIFFSCALE (model definition option) 68
STOP (parameter) 29
storage methods 729
strain correction method 714
STRAIN RATE (model definition option) 185, 324,
350, 451, 452
STRAIN RATE (model definition, fluid option) 526
strain rate effects on elastic-plastic constitutive
relation 460
stress
analysis 747
intensity factor defined 141
intensity factors 141
STRUCTURAL (parameter) 335, 336
structural zooming analysis 208
substructuring 66
SUMMARY (model definition option) 741
SUPERELEM (history definition option) 757
SUPERELEM (model definition option) 757
SUPG 319
surface energy 230
SURFACE ENERGY (model definition option) 230,
248
SURFACES (model definition option) 76
SWLDPRM (model definition option) 663, 675, 678,
680
symmetries 58
SYNCHRONIZED (history definition option) 538
system and element stiffness matrices 779
T
TABLE (model definition option) 274, 275, 305, 387,
456, 457, 481, 493, 495
TEMP CHANGE (history definition option) 220,
225, 324, 327, 337, 342, 343, 344
TEMPERATURE EFFECTS (coupled thermal-stress,
model definition option) 525
TEMPERATURE EFFECTS (heat transfer, model
TEMPERATURE EFFECTS (hydrodynamic, model
Index
807
TEMPERATURE EFFECTS (model definition
option) 220, 224, 275, 286, 350, 422, 439, 451,
452, 481, 486, 493, 495, 526
Temperature Gradient Criterion 79
temperature-dependent
creep 494
material properties 493
tension softening 499
thermal
loads 620
loads input options 620
THERMAL CONTACT (2-D) (model definition
option) 281, 303, 304, 305, 564, 565
THERMAL CONTACT (model definition option)
221, 264, 291, 299, 339
thermal contact analysis 563
thermal expansion coefficient 495
thermochemical ablation and surface wnergy
balance 230
thermo-electrical 525
thermo-mechanical shape memory model 404
constitutive theory 407
phase transformation strains 407
transformation induced deformation 406
THERMO-PORE (model definition option) 240
Thermo-Rheologically Simple material 484
THICKNESS (model definition option) 286
THICKNS CHANGE (history definition option) 286
TIME STEP (history definition option) 451, 483, 488
time step definition 176
time-dependent
boundary conditions 177
inelastic behavior 464
time-independent, inelastic behavior 428
TIME-TEMP (model definition option) 496, 497
Time-Temperature-Transformation (T-T-T) 495
tolerance values, contact 571
total Lagrangian procedure 106
TRANSFORMATION (history definition option) 644
TRANSFORMATION (model definition option) 537,
627, 634, 644, 658, 748
TRANSIENT (history definition option) 221, 222,
223, 226, 337, 343, 344, 700, 701
TRSFAC (user subroutine) 486
truss elements 687
Tsai-Wu failure criterion 364
TSHEAR (parameter) 377, 745
T-T-T (parameter) 496
tying
overclosure 642
shell-to-solid 639
TYING (model definition option) 221, 332, 597, 631
TYING CHANGE (history definition option) 631
tying constraints
beam-to-beam intersection 634
cyclic symmetry 649
mesh refinement 632
nodal degrees of freedom 638
pipe bend element 637
rigid link 639
shell stiffener 637
shell-to-shell intersection 633
shell-to-solid intersections 634
U
UADAP (user subroutine) 79
UADAPBOX (user subroutine) 263
UBEAR (user subroutine) 287
UBREAKGLUE (user subroutine) 557
UCOHESIVE (user subroutine) 521
UCOHESIVET (user subroutine) 521
UCOMPL (user subroutine) 165
UCONORT (user subroutine)T 556
UCOORD (user subroutine) 76
UCRACK (user subroutine) 498
UCRACKGROW (user subroutine) 152, 155
UCRPLW (user subroutine) 134, 137, 474, 494
UDELAM (user subroutine 159
UELASTOMER (user subroutine) 426
UELDAM (user subroutine) 516
UENERG (user subroutine) 426
UEPS (user subroutine) 290, 294, 306, 525, 526
UFAH (user subroutine) 233
UFAIL (user subroutine) 358, 365
UFCONN (model definition option) 54
UFCONN (user subroutine) 54
UFINITE (user subroutine) 432
UFLUXMEC (parameter 230
UFLUXMEC (user subroutine) 238
UFMEC (user subroutine) 237
UFORMSN (user subroutine) 631, 638
UFOUR (user subroutine) 96
UFRIC (user subroutine) 553, 554, 555
UFTHP (user subroutine) 235
UFXORD (model definition optiion) 54
UFXORD (user subroutine) 54, 60
UGLAW (user subroutine) 235
UGMEC (user subroutine) 237
UGROOV (user subroutine) 286
UHTCOE (user subroutine) 563, 564
UHTCON (user subroutine) 563, 564
UHTNRC (user subroutine) 563, 564
UMOONY (user subroutine) 422
UMU (user subroutine) 306, 526
UNEWTN (user subroutine) 185
uniaxial cracking data 498
UOGDEN (user subroutine) 423
UPDATE (parameter) 171
updated Lagrange Formulation, nonlinear elasticity
427
updated Lagrangian Procedure 109
UPNOD (user subroutine) 184
UPOSTV (user subroutine) 750
URESTR (user subroutine) 286
URPFLO (user subroutine) 185
User subroutine
ANELAS 95, 348, 350
ANEXP 95, 350, 495
ANKOND 221, 525
ANPLAS 350, 433
CRPLAW 134, 135, 139, 468, 470, 473, 475,
479, 494
CRPVIS 138, 467, 479
ELEVAR 742, 745
ELEVEC 742
FFORCEM 343
FILM 221, 226, 228, 337, 342, 343, 344, 622
FLUX 221, 225, 227, 286, 290, 294, 311, 315,
335, 337, 342, 343, 344, 622
FORCDF 344
FORCDT 177, 225, 227, 287, 290, 294, 311,
324, 335, 336, 343
FORCEM 75, 177, 306, 311, 335, 337, 344
HOOKLW 95, 348, 350
HOOKVI 483
HYPELA2 106, 390, 501
IMPD 742
INTCRD 742
MOTION (2-D) 536
NASSOC 139, 475
ORIENT 95, 348, 433, 483
ORIENT2 354, 355, 358
PLOTV 750
REBAR 498
SEPFOR 578
SEPSTR 578
TRSFAC 486
UADAP 79
UADAPTBOX 263
UBEAR 287
UBREAKGLUE 557
UCOHESIVE 521
UCOHESIVET 521
UCOMPL 165
UCONORT 556
UCOORD 76
UCRACK 498
UCRACKGROW 152, 155
UCRPLW 134, 137, 474, 494
UDELAM 159
UELASTOMER 426
UELDAM 516
UENERG 426
UEPS 290, 294, 306, 525, 526
UFAH 233
UFAIL 358, 365
UFCONN 54
UFINITE 432
UFLUXMEC 238
UFMEC 237
UFORMSN 631, 638
UFOUR 96
UFRIC 553, 554, 555
UFTHP 235
UFXORD 54, 60
UGLAW 235
UGMEC 237
UGROOV 286
UHTCOE 563, 564
UHTCON 563, 564
UHTNRC 563, 564
UMOONY 422
Index
809
UMU 306, 526
UNEWTN 185
UOGDEN 423
UPNOD 184
UPOSTV 750
URESTR 286
URPFLO 185
USIGMA 306, 526
USINC 177, 225
USPLIT_MESH 158, 161
USPRNG 115, 646
USSD 167
UTHICK 286, 287
UTIMESTEP 706
UTIMP 237
UTRANS 627
UVELOC 221, 279, 286
UVSCPL 135, 140, 475
UWATERSL 245
UWELDFLUX 257, 259
VSWELL 135, 474, 479
WKSLP 456, 457
YIEL 139, 475
ZERO 139, 475
user subroutine input 54
user-defined Criterion 79
user-defined failure criteria 365
USIGMA (user subroutine) 306, 526
USINC (user subroutine) 177, 225
USPLIT_MESH (user subroutine) 158, 161
USPRNG (user subroutine) 115, 646
USSD (user subroutine) 167
UTHICK (user subroutine) 286, 287
UTIMESTEP (user subroutine) 706
UTIMP (user subroutine) 237
UTRANFORM (model definition option) 627
UTRANS (user subroutine) 627
UVELOC (user subroutine) 221, 279, 286
UVSCPL (user subroutine) 135, 140, 475
UWATERSL (user subroutine) 245
UWELDFLUX (user subroutine) 257, 259
V
VCCT (model definition option) 140, 146, 151
VELOCITY (model definition option) 221, 279, 286
VELOCITY CHANGE (history definition option)
221, 279, 324
VIEW FACTOR (model definition option) 268, 622
virtual crack closure technique (VCCT) 140
VISCEL EXP (model definition option) 138, 492
VISCELFOAM (model definition option) 138, 488
VISCELMOON (model definition option) 138, 422,
488
VISCELOGDEN (model definition option) 138, 423,
488, 516
VISCELORTH (model definition option) 483
VISCELPROP (model definition option) 483
VISCO ELAS (parameter) 138
viscoelastic material
incompressible isotropic 483
isotropic 481
Thermo-Rheologically Simple 484
viscoelasticity 138
viscoplasticity
explicit formulation 475
explicit method 139
implicit method 140
VOID CHANGE (model definition option) 385
VOLTAGE CHANGE (history definition option) 337,
342
von Mises 129, 136, 197, 431, 434, 439, 455, 458,
459, 460, 461, 470, 741, 742, 743
VSWELL (user subroutine) 135, 474, 479
W
weighted mesh 58
WELD FILL (model definition option) 257, 261, 262
WELD FLUX (model definition option) 257, 262, 263
WELD PATH (model definition option) 257, 259
welding 257
weld flux 257
weld path 259
Williams-Landel-Ferry (WLF) 138, 485
windings 299
WKSLP (user subroutine) 456, 457
WORK HARD (model definition option) 185, 350,
452, 456, 457
workhardening
isotropic 455
rules 454
slope 429
workspace allocation 28
Y
YIEL (user subroutine) 139, 475
Youngs 350
Youngs modulus 347, 350, 427, 430, 455, 493, 496,
498, 499, 501
Z
ZERO (user subroutine) 139, 475
Zienkiewicz-Zhu Creep Strain Criterion 78
Zienkiewicz-Zhu Criterion 77
zooming 208

Vol A

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Vol A

Hochgeladen von

Copyright:

Verfügbare Formate

Marc

Marc for Analysis 23

Mentat or Patran for GUI 24

Program Function Library 25

Marc Host Systems 27

Examples of Running Marc Jobs 35

Table Driven Input 41

Model Definition Options 46

History Definition Options 46

User Subroutine Input 54

Incremental Mesh Generators 64

BEAM SECT Parameter 69

Automatic Global Remeshing 80

Fracture Mechanics 140

Inertia Relief 180

Rigid-Plastic Flow 183

Soil Analysis 188

Mechanical Wear 192

Design Sensitivity Analysis 194

Design Optimization 197

Defined Initial State with Result Data from Previous Analysis

Compatible element types for PRE STATE 204

Structural Zooming Analysis 208

Cure-Thermal-Mechanically Coupled Analysis 211

132 Marc Volume A: Theory and User Information

158 Marc Volume A: Theory and User Information

174 Marc Volume A: Theory and User Information

Heat Transfer 220

Hydrodynamic Bearing 285

Electrostatic Analysis 289

Magnetostatic Analysis 293

Magnetodynamic Analysis 305

Piezoelectric Analysis 310

Acoustic Analysis 314

Fluid Mechanics 315

Coupled Analyses 324

234 Marc Volume A: Theory and User Information

330 Marc Volume A: Theory and User Information

Linear Elastic Material 347

Composite Material 349

Nonlinear Hypoelastic Material 389

Thermo-Mechanical Shape Memory Model 404

Mechanical Shape Memory Model 413

Time-independent Inelastic Behavior 428

Time-dependent Inelastic Behavior 464

Temperature Effects and Coefficient of Thermal Expansion 492

Low Tension Material 498

Soil Model 500

Damage Models 512

Nonstructural Materials 524

n = Normal to Surface Tangent Plane

(1) Bilinear Representation

(3) Perfectly Plastic

(5) Strain Softening

(2) Elastic-Perfectly Plastic

(4) Piecewise Linear Representation

468 Marc Volume A: Theory and User Information

500 Marc Volume A: Theory and User Information

Critical State Line

508 Marc Volume A: Theory and User Information

Definition of Contact Bodies 530

Numbering of Contact Bodies 532

Motion of Bodies 536