Sie sind auf Seite 1von 5

Zbornik radova Fakulteta tehnikih nauka, Novi Sad

UDK (Upisuje biblioteka, ostaviti jedan red)


KO

SUPERELEMENT METHOD IN DYNAMIC ANALYSIS OF NORTSTROSGRUMD LIGHTHOUSE


Ivana Krajnovi, Faculty of Technical Sciences University of Novi Sad
2. SUPERELEMENT
CONCEPT
AND
IMPLEMENTATION
Field of study: Civil Engineering, Structural
Engineering
The superelement method (also known as the
substructuring technique) is a technique used in finite
1
element analysis (FEA) of complex structures since

1960s. It was originally developed within aerospace
.
industry and later adopted by the offshore and

shipbuilding industries [3].
. ,

.
65 06'6 N
22 19'3 E

Abstract This paper shows the use of the superelement


method in dynamic analysis of the lighthouse structure.
Superelement method is a reduction technique used in
Finite Element Analysis (FEA) of complex structures. In
this paper different modeling approaches are presented
and compared against each other. Furthermore, usage of
different software for dynamical analysis is presented.
Key words: Superelement method, Abaqus CAE,
Lighthouse, Structural dynamics
1.

INTRODUCTION

Norstrmsgrund lighthouse, shown in Figure 1, was


deployed in 1971 about 60 km offshore Lule in Sweden,
in the Gulf of Bothnia, in the Baltic Sea. The lighthouse is
a concrete structure with foundation caissons filled by
sand. Total height of the structure is 42.3 m. The diameter
of the foundation is 23 m, and diameter of the structures
body varies between 4.3 and 5.3 m. Diameter of the
structure at the waterline is 7.2 m. The total mass of the
structure is about 9000 t, and the sea dept is about 14 m.
Detail description of the lighthouse is in [5]. During the
winter months, this area is usually covered by the sea ice.
Sea ice is a solid material with crystal structure. In cold
regions, it floats on the water and, occasionally strikes
offshore structures. The floating sheets of ice are crushing
against the lighthouse causing vibrations of the structure.
In order to collect data about properties of the sea ice and
its impact on the lighthouse, the extensive measurements
were performed on this structure. To better understand
data from measurements and learn more about the
behavior of the structure various numerical models of the
lighthouse were implemented [2], [10], [6]. In what
follows the creation of the superelement model is
presented.

1 :
- .

Figure 1. The perspective of Nrstromsgrund lighthouse [7],


left, and location of the lighthouse, right2

The aim is to treat a group of finite elements with


common attributes as a single element (superelement). In
the domain of FEA, this means reduction of the system
matrices, and, also, possibility to design parts of complex
structures independently [1]. The technique is
advantageous for the structures with repetitive patterns,
since mass and stiffens calculations are performed only
once per superelement. Reduction of the processing time
is also one of the advantages of this technique. Here are
the principal steps of creating a superelement model
followed by a graphical example in Figure 1. Firstly, there
is a complete model made of finite elements (FEs), as
shown in Figure 2(a). It is then divided into several
substructures. Every substructure now has internal, and
boundary nodes (Figure 2(b)). Internal nodes (slave
nodes) are connected only to the nodes within its
substructure, and boundary nodes (master nodes) may be
connected to nodes of other substructures. In this point,
the system matrices are calculated within every
substructure, but the elements connected to the master
nodes are only preserved. An elimination of the internal
nodes, and their degrees of freedom (DOFs) leads to the
simpler mathematical model of a substructure. Now,
every substructure can be observed as a single element,
superelement (Figure 2(c)). At the end (Figure 2(d)),
master nodes between different substructures are
2 generated with www.google.de/maps

connected, and the structure is assembled from different


superelements.

Only odd eigenfrequencies and corresponding


eigenmodes are presented and compared in what follows,
since the structure is symmetric. Therefore, even
eigenfrequencies are same as odd and corresponding
eigenmodes are symmetric.
Adjustment of 3D model
In the original model, material density was different along
the structure, to take additional masses of equipment and
thermal insulation into account. For this research, material
properties were uniformed along the 3D model. It has
been done by assigning same material properties to the
entire structure, classic reinforced concrete with the
density of 2500kg m3. Eigenfrequencies of 3D models
with uniformed and different materials are presented in
Table 1. The differences are below 2.5%, which justifies
this simplification.
Table 1. Percentage difference in terms of natural
frequencies between model with various and uniformed
material properties

Figure 2. Principal steps in superelement analysis [1]

By performing dynamic condensation [1], [8] the system


matrices of every superelement are reduced in the way.
2.1. Reduction of 3D model to a superelement model
The objective of this research is to simplify the 3D
structure, but to retain physical properties of it.
Substructuring technique was suitable for the
simplification. It enables condensation of the models
properties in terms of mass, stiffness and dynamical
behavior only in certain points the master nodes.
3D model
The Abaqus software package is a finite element analysis
product suitable for wide range simulations of various
design properties [11]. Version 6.13 is used for this
research. The existing finite element model of the
lighthouse is based on 3D solid elements. It was created
in Fraunhofer IWES [10]. Structural and dynamical
properties were derived from [5]. The foundation plate is
in XZ plane. The structure was supported at the central
node of the foundation plate. Soil properties are
introduced in this node as point mass and inertia,
transitional and torsion springs. Sand filling in caissons
was presented as a point mass/inertia in the point (0 m,
4.58m, 0m). Hydrodynamic added mass and thermal
insulation were presented through an increase of the
density of the material in corresponding areas of the 3D
model. The structure was meshed with ten-node quadratic
tetrahedral elements. Other properties and values of the
attachments 3D model are presented in [10]. The natural
frequencies and corresponding mode shapes of the 3D
model were calculated with no effects of damping and
gravity loads included. These frequencies are presented in
Table 1, and correspond to the limited information about
dynamics of the full-scale structure. It was decided to
keep first ten eigenmodes below 20Hz since they
incorporate 98.9% of the modal mass as proved in [10].

Mod
e
1
3
5
7
9

3D Model
Original
[Hz]
2.826
4.170
5.617
12.974
18.085

3D Model
uniformed [Hz]

Difference
[%]

2.840
4.274
5.685
12.972
18.284

0.475
2.422
1.200
0.015
1.088

Generation of superelement model


The substructure generation procedure contains two steps
in Abaqus CAE. The first step is defining master nodes of
the future superelement. The idea was to reduce the entire
structure to a single superelement, to condense
information about the entire 3D model only in several
points of interest. Therefore, the reasonable selection of
the master nodes is of crucial importance in the reduction
procedure. On the full-scale structure, measurements were
performed only in two points, where the accelerometers
were mounted. Thus, those are the main nodes of interest
on the numerical model. The point on the contact between
ice and structure, at the waterline is also of importance.
That is where the superelement model of the lighthouse
can be subjected to environmental loads, such as sea ice,
or wave loading. To facilitate verifications of the
numerical model, it was decided to retain additional nodes
along the structures main axis and one more in the plane
at waterline. Finally, the master nodes of the superelement
model are presented in Figure 3. The equation of motion
of 3D model is presented with the matrix equation:
(1)
M u +C u + Ku=f
Where: K stiffness matrix, M mass matrix and C the
damping matrix of the model, u vector of displacement
DOFs, u
velocity vector, u acceleration
vector and f vector of external forces. All six DOFs
connected to the master nodes have been retained. In the
second step of the substructure generation system
matrices of the 3D model from the Equation 1 are
changed; stiffens matrix is condensed and mass matrix
and damping matrix are reduced to the matrices of the

superelement. The mathematics behind this condensation


is presented in [1], [9] and many others.
Node
names
N_42,3
N_39,5
5
N_37,1
0
N_34,3
0
N_31,5
0
N_28,9
5
N_25,8
5
N_22,8
0
N_19,6
5
N_16,5
0
N_14,2
5
N_14,2
5

Coordinates

42.3

39.55

37.1

34.3

31.5

28.95

25.85

22.8

19.65

16.5

14.25

-3.6

14.25

3.6

(a) 1st

(2)
C= M + K
and are damping coefficients calculated as it follows:

1 22 1

(3)

2221

2 21 1

(4)

2221

i =2 f i
Where:

1,2

1= 2 .

(5)
-1

angular frequencies [rad s ] calculated

with the equation 5,

1,2

(c) 3rd (c) 5th


(d) 7th
(e) 9th
Figure 4. Eigenmodes of the superelement model

2.2. Comparison of dynamic properties

The damping matrix of the superelement model was


presented with Rayleigh viscous damping, as a
combination of mass and stiffness matrices:

=2

are calculated for the two lowest


f 1 =2.839 Hz and f 2=4.274 Hz
and a damping ratio of 1= 2=0.035 , which is the
and

mean value of damping ratios for 3D model proposed in


[10]. The values of and are, respectively,
0.7503 and 0.0016.

Figure 3. Master nodes of the superelement model

=2 1 2

Herein,
frequencies,

damping ratios and

Prior to further use of the superelement model, it was


necessary to confirm that it contains the properties of the
3D model. The superelement model has the same
eigenfrequencies as the model from which has been
derived. The eigenmodes of the superelement model are
presented in Figure 4. Retained nodes are presented by
black dots. The 1st eigenmode is characterized as bending
of the structures tower. Caisson is also slightly rotated.
The 3rd represents bending of the tower and translation of
the caisson. The 5th eigenmode is also characterized by
bending of the tower and rotation of the caisson. The 7 th
represents bending and a small torsion of the structures
tower with no movement of the caisson. The 9 th mode
represents torsion of the structure. These shapes
correspond to the ones reported by Popko in [10]. The
quantitative comparison between eigenmodes of the
superelement model and 3D numerical model has been
performed in terms of magnitudes of displacements. For
first 6 eigenmodes, the difference is below one percent.
For the last eigenmodes the difference between
displacements magnitudes is considerable, over 90% for
the 9th eigenmode. These modes are characterized with the
torsion of the structure. Thus, to decrease the difference, it
would be necessary to retain more nodes in the
substructure generation procedure. Additional nodes
should be at certain distance of the structures vertical
axis, to take the effects of torsion into account. Retaining
more degrees of freedom would not only consume more
modeling time. It would increase the orders of system
matrices, and, subsequently, computing time during the
simulations. In this particular case, retaining additional
nodes would not significantly affect the accuracy of the
model, since forces that can cause the torsion are not
going to be applied to the structure.
2.3. Exportation procedure
Abaqus has been used so far to create 3D model and
generate superelement model that can be used in analyses
within this software package; however, it can generate a
file with information about the finite element models that
could be analyzed within another simulation environment.
This file is standard input data file (SID) and has .SID
extension. As explained by Wallrapp in [12]: SID files

are based on modal representation of small body


deformations and structure of it is organized using an
object-oriented class description. This means that the
SID file contains information about finite element
models system matrices, herein superelements matrices.
The SID file for the superelement is generated with the
Abaqus software package from a SIM file, using an
undocumented function abaqus tosid [11]. With the
generation of this file, the creation of superelement model
of the lighthouse is accomplished.
Superelement model in Modelica
Modelica is an object-oriented programming language,
for modeling complex systems, used to describe models
for individual components. The SID file of the
superelement, have been imported in Modelica as
ModalBody model. The ModalBody model class is a part
of the commercial DLR FlexibleBodies library,
implemented in Modelica by Heckmann et al. [4].
Verification against superelement model in Abaqus
In order to conclude that model in Modelica behaves in
the same manner as the models in Abaqus three
comparisons were made. The very basic comparison
between the superelement in Modelica and the
superelement in Abaqus was in terms of total mass of the
structure. The total mass of the structure is 9192769.87 kg
in both Abaqus and Modelica. The second comparison
was conducted in terms of eigenfrequencies. The
eigenfrequencies of the model in Modelica fit perfectly to
the ones in Abaqus. The third comparison was in terms of
steady-state displacements and the response of the
structure. For this purpose, an impulse horizontal force of
5 MN was applied on the top of the lighthouses in Abaqus
and Modelica. The responses of these two models were
compared in following points: 37.10 m and 16.50 m from
the sea bed. The responses of the both models correspond
to each other, and are presented in Figure 5.With this, it is
concluded that superelement in Abaqus is the same as the
superelement in Modelica.

0.02
0.02

Displacement [m]
Modelica

0.01
0.01

Abaqus

0
0

10 15 20 25

Time [s]
Figure 5. Responses to an impulse force of 5MN of the models
in Abaqus and Modelica at 16.50 m

3. Conclusions
The existing 3D numerical model of the Norstrmsgrund
lighthouse was reduced to the superelement model in

Abaqus CAE, using the substructuring technique. The


superelement model of the lighthouse inherited
assumptions on which the 3D model is set up. After the
reduction, the superelement model is verifiedagainst the
3D model in Abaqus in terms of eigenfrequencies and
corresponding modal shapes. The differences in terms of
eigenfrequencies between the models are below 2.5%
meaning that superelement model is good enough
representation of the 3D model and, subsequently the fullscale structure. The reason behind the simplification was
the reduction of computing time during the coupled
simulations of the ice-structure interaction. The coupled
simulations of the ice-structure interaction were set up in
Modelica. The superelement model of the lighthouse is
imported from Abaqus to Modelica, making the
importation procedure successful as well.
4. Acknowledgments
My deepest gratitude has Wojciech Popko, for having
high demands, precious comments and corrections, as
well as patience while mentoring me at Fraunhofer IWES.
5. Bibliography
[1]
M. V. Belyi. Superelement method for transient
dynamic analysis of structural systems. International
Journal for Numerical Methods in Engineering, 36:2263
2286, 1993.
[2]
Morten Bjerks, Arne Alberktsen, and Arne
Grtner. Static and dynamic ice actions in the light of new
design codes. In Proceedings of OMAE2010 29th
International Conference on Offshore Mechanics and
Arctic Engineering, pages 733739, Shangai, China, June
2010.
[3]
C. A. Felippa. Introduction to finite element
methods. Department of aerospace Engineering Sciences
and Center for Aerospace Structures, University of
Colorado, Boulder, Colorado, USA, 2004.
[4]
Andreas Heckmann, Martin Otter, Stefan Dietz,
and Hos Daz Lpez. The DLR FlexibleBodies library to
model large motions of beams and of flexible bodies
exported from finite element programs. In Proceedings of
the 5th Modelica Conference, volume 1, pages 8595,
Vienna, Austria, September 2006. The Modelica
Association and arsenal research.
[5]
Jaakko Heinonen, Tuomo Krn, and Chouping
Luo. STRICE REPORT Dynamic behaviour of the
Norstrmsgrund lighthouse. Technical Report D-6.2-A,
VTT Technical Research Center of Finland,
LuleUniversity of Technology, January 2004.
[6]
P. Jochmann. STRICE REPORT: Full Scale
Measurements at Lighthouse Norstrmsgrund winter
2001. Report, Technical D-5.1.1., Hamburgishche
Schiffbau-Versuchsanstalt GmbH, Hamburg, Germnay,
2003.
[7]
K. Kolari. http://virtual.vtt.fi/virtual/proj6/arki/kuvat/strice/Norstromsgrund_Lighthouse_2003.html,
February 2003. Online accesed: January 31, 2015.
[8]
A. Y. Leung. An accurate method of dynamic
condensation in structural analysis. International Journal
for Numerical Methods in Engineering, 12:17051715,
1978.

[9]
A. Y. Leung. An accurate method of dynamic
substructuring with simplified computation. International
Journal for Numerical Methods in Engineering, 14:1241
1256, 1979.
[10]
Wojciech Popko. Comparison of full-scale and
numerical model dynamic responses of Norstrmsgrund
lighthouse. In Proceedings of the ASME 2014 33rd
International Conference on Ocean, Offshore and Arctic
Engineering OMAE, San Francisco, CA, USA, June 2014.
American Society of Mechanical Engineers.
[11]
Simulia. Abaqus FEA user guide. http://www.3ds.com/products-services/simulia/portfolio/abaqus/latest-release/?

xtmc=abaqus_user_manual&xtcr=4, July 2014. Online


accessed on February 2nd, 2015.
[12]
Oscar Wallrapp. Standardization of flexible body
modeling in multibody system codes, part i: Definition of
standard input data. Mechanics Based Designe of
Structures and Machines, 33(2):283304, 1994.
Kratka biografija:
, 1988.

2013.
Fraunhofer IWES.


.