SESSION TITLE:

Finite element Procedures for Nonlinear Analysis:

Lagrangian and Eulerian Formulations of 1D Continua

Topics
Introduction and overview
Governing equations: total Lagrangian formulation
Weak form: total Lagrangian formulation
Finite element discretizations: total Lagrangian Formulation
Element and Global matrices: total Lagrangian formulation
Governing equations: Updated Lagrangian formulation
Weak form: updated Lagrangian formulation
 SESSION TITLE:
Finite element Procedures for Nonlinear Analysis:
Lagrangian and Eulerian Formulations of 1D Continua

Finite element equations: updated Lagrangian formulation

Governing equations: Eulerian formulation
Weak form: Eulerian formulation
Finite element equations: Eulerian formulation
Solution methods
 Introduction and Overview
•One dimensional finite element for nonlinear analysis of Continua
are described and the corresponding finite element equations are
developed.
•Salient features for both Lagrangian and Eulerian formulations are
demonstrated.
•These developments are applicable to nonlinear rods and 1D fluid
flow.
•Some concepts of finite element discretizations and solution
procedures are also reviewed: Weak and strong forms; Operations of
assembly, gather and scatter; imposition of essential boundary
conditions and initial conditions; continuity requirements for finite
element approximation for solutions.
•While this information may be familiar to those who have studied
Linear Finite Elements, it is worthwhile to skim these to refresh your
understanding and learn the notation.
•In SOLID MECHANICS, Lagrangian formulations are most
popular. The ease with which they handle complicated boundaries
and their ability to follow material points, so that history-
dependent materials can be treated accurately are the attractions.
•In the design of Lagrangian Finite Elements, two alternative
approaches are commonly taken.
•Formulations in terms of the Lagrangian measures of stress and
strain in which derivatives and integrals are taken with respect to
the Lagrangian (material) coordinates X called, Total Lagrangian
Formulations (TLF).
•Formulations expressed in terms of Eulerian measures of stress
and strain in which derivatives and integrals are taken with respect
to the Eulerian (spatial) coordinates x, called Updated Lagrangian
Formulations (ULF).
•Although the Total Lagrangian and updated Lagrangian finite
element formulations are superficially quite different, the underlying
mechanics are identical; Expressions in the TLF can be transformed
to ULF and vice versa. The major difference is in the point of view. In
the TLF, the variables are described in the original configuration, and
in the ULF in the current configurations. However, different stress
and strain measures are used in these two formulations. For example,
the TLF uses a total measure of strain, where as the ULF often uses a
rate measure of strain.
•Until recently, Eulerian elements have not been used much in
SOLID MECHANICS. Eulerian element meshes are most appealing
in problems with very large deformations. Their advantage in these
problems is a consequence of the fact that Eulerian elements do not
deform with the material, therefore regardless of the magnitude of the
deformation, Eulerian elements retain their original shape. So
Eulerian elements are used in metal forming simulations.
•For each of the formulations, a weak form of the momentum
equation, which is known as the Principle of Virtual Work will be
developed. The weak form is developed by integrating the product of
a test function with the momentum equation. The integration is
performed over the material coordinates X for the TLF or over the
spatial coordinates, x, for Eulerian and ULF. This procedure is
identical to that in Linear Finite Element Analysis. The major
difference in nonlinear finite element formulations is the need to
define the coordinates over which the integrals are evaluated and to
specify the choice of stress and strain measures. Another salient
feature of the weak form is how the surface traction boundary
conditions are treated so that the approximation functions (trial
functions; shape functions) need not apriori satisfy surface traction
boundary conditions.
•The discrete equations for a finite element will then be derived.
For problems in which the Inertia forces (mass X acceleration) are
important (often called dynamic problems) or those involving rate-
dependent materials, the resulting discrete finite element equations
are a set of Ordinary Differential Equations (ODE’s). The process
of discretizing in space is called semi-discretization since the finite
element procedure only converts the space differential operators to
discrete form: the derivatives in time are not discretized!!
•For static problems with rate-independent materials, the discrete
equations are independent of time and the finite element
discretization results in a set of nonlinear algebraic equations.
•Examples of TLF and ULF are given for the two-node ROD
element and the three-node ROD element.
•Finally, to enable the student to solve some nonlinear problems a
central difference, Explicit time-integration algorithm is presented.
 GOVERNING EQUATIONS FOR TLF
•Notation
Consider the rod shown in Figure 2.1. The initial configuration,
also called the undeformed configuration, is shown in the lower
part of the figure. This configuration plays an important role in the
large deformation analysis of solids. It is also called the reference
configuration, since all equations in the TLF are referred to this.
The current configuration, also called deformed configuration, is
shown in the upper part of the figure. The spatial (Eulerian)
coordinate is denoted by x and the material (Lagrangian)
coordinates are denoted by X. The current cross-section area is
denoted by A(X,t) and the current density by (X,t). The initial
cross section area of the rod is denoted by Ao(X) and its initial
density by o(X).
The displacement is denoted by u(X,t). The boundary points in the
reference configuration are Xa and Xb.
•Motion And Strain Measure
•The motion of the body (rod) is described by a function of
X and t.
X= (X,t), X(Xa, Xb) ---------------------(1)
where (X,t) is called a map between the initial and current
domains.
•The material coordinates are the initial position, so
X= (X,0) ----------------------------------(2)
•The displacement u(X,t) is given by the difference
between the current position and the original position of a
material point.
U(X,t) = (X,t) – X, or u=x-X ---------------(3)
•The deformation gradient is defined by
F= /X --------------------------(4)
Let J denotes the Jacobian between the current and
reference configurations.
J=Infinitesimale volume in the deformed
body(Ax)/infinitesimale volume of the undeformed body
(AoX).
J = (x/X).(A/Ao) = F.(A/Ao )--------------------(5)
The deformation gradient F is an unusual measure of strain
since its value is unity when the body is undeformed. We
will therefore define the strain measure by
(x,T) = F(x,T) – 1 = (x/X) – 1 = u/X ----(6)
The above vanishes in the undeformed configurations. It is
equivalent to the Engineering Strain.

There are many more measures of strain but this is the most
convenient for this presentation.
•Stress measures:
The stress measure used in the TLF is not the physical stress (also
known as Cauchy stress). Let the total force across a given cross
section be denoted by T and assume that the stress is constant in the
cross-section.
The Cauchy stress is given by
 = T/A ---------------------------------------------(7)
Note that this measure of stress refers to the current area A. In the
TLF, we will use the nominal stress. The nominal stress will be
denoted by P and is given by
P = T/Ao ---------------------------------------------(8)
This is equivalent to the definition of the engineering stress.
However, no definition of the engineering stress is available in
multi-dimensional stress states.
Comparing (7) and (8):

 = (Ao/A).P, P = (A/Ao).  -------------------------(9)

• Governing equations:
The nonlinear behavior of the rod is governed by the following
equations.
1. Conservation of mass
2. Conservation of momentum
3. Conservation of energy
4. A measure of deformation, often called strain-displacement
equation
5. A constitutive equation, which describes material behavior
and relates stress to a measure of deformation.
In addition, we require the deformation to be continuous, which is
often called a compatibility requirement. In processes with no
heat or energy transfer, the energy equation need not be
considered.
• Conservation of mass:
J = oJo ------------------------------------------(10)
OR
 (X,t). J(X,t) = o (X).Jo (X) ------------------------(11)
Where Jo is the initial Jacobian, which is taken as unity.
For the case of ROD, F.A = oAo
• Conservation of momentum:
In terms of the nominal stress P and the Lagrangian coordinates X
gives
(AoP),X + oAob = oAoü ----------------------------(12)
where ü = 2u(X,t)/ t2 and b is the body force.
The subscript following a comma (,) denotes partial differentiation
with respect to that variable.
Equation (12) is called Momentum equation. If the initial cross-
sectional area is constant in space, the momentum equation
becomes
P,X + ob = oü --------------------------------------(14)
When the inertial term ou vanishes or neglected, i.e., when the
problem is STATIC, the momentum equation becomes the
equilibrium equation.
(AoP),X + oAob ≡0 ------------------------------------(15)
• Conservation of energy:
The energy conservation equation for a rod of constant area in the
absence of heat conduction or heat sources is
oŴInt = FP -------------------------------------------(16)
Where ŴInt is the rate of internal work, F is the rate of
deformation gradient and P is the nominal stress.
• Constitutive equations:
The constitutive equations give the stresses resulting from the
deformation. They relate the stress or stress rates to the
measures of strain and/or, strain rate at a material point.
The constitutive equation can be written either in total form, which
relates the current stress to the current deformation.
  
P(X, t)  S F(X, t ), F (X, t ) , t  t        (17)
PF

or in rate form
 
P(X, t)  S PF F (X, t ), F(X, t ) , t  t        (18)
t

Here SPF and StPF are functionals of the deformation history. The
superscripts are appended to the constitutive functionals to
indicate which measures of stress and strain they relate.
The stress is assumed to be a continuous function of strain. As
indicated in equation (17), the stress can depend on both F and
F and on other state variables, such as temperature, etc.
The stress can also depend on the history of deformation, as in
elastic-plastic materials; This is indicated in (17) and (18) by
letting the constitutive functions depend on deformation for all
time prior to t.
The constitutive equation for a solid is customarily expressed in
material coordinates because it depends on the history of
deformation at a material point.
Examples of constitutive equations are
1. Linear Elastic Material
Total form: P(X,t) = EPF(X,t) = EPF (F(X,t)-1)
Rate form: P(X,t) = EPF (X,t) = EPF F(X,t)
2. Linear viscoelastic material:
P(X,t) = EPF [F(X,t)-1) +F(X,t)]
or P = EPF (+)
For small deformations, the material parameter EPF corresponds to
Young’s modulus; the constant  determines the magnitude of
damping.
• Momentum equation in terms of displacements:
A single governing equation can be obtained by substituting the
relevant constitutive equation (17) or (18), into the momentum
equation (12) and expressing the strain measure in terms of the
displacement by (6). For the total form of the constitutive
equation, the resulting equation is

A Pu
0 ,x , u ,X    ,X  ρ0 A0 b  ρ0 A0u    (22)

Which is a nonlinear partial differential equation in the

displacement u(X,t).
For linear elastic material (19), (22) becomes
(AoEPFu,X),X + oAob = oAoü -------------------------(23)
For a rod of constant cross section and modulus with zero body
force, the above yields the well-known Linear Wave Equation.
u,XX + 1/Co2ü where Co2 = EPF/o
If in addition the inertia effect is neglected, the governing
equation is
EPF.u,XX = 0
This equation is elliptic.

• Boundary conditions:
A boundary is called displacement boundary and denoted by u if
the displacement is prescribed. The prescribed value is surface
traction on t. The boundary conditions are
u = ŭ on u
noP = txo on t
Initial conditions are expressed in terms of displacement and
velocity.
u(X,0) = uo(x) for X[Xa,Xb]
ŭ (X,0) = o(X) for X[ Xa,Xb]
If the body is initially undeformed and at rest, the initial conditions
can be written
u(X,0) = 0
ŭ (X,0) = 0
WEAK FORM FOR TOTAL LAGRANGIAN
FORMULATION:
The momentum equation cannot be discretized directly in the FEM.
In order to discretize this equation, a weak form, often called a
variational form is needed. The principle of virtual work, or weak
form, which will be developed next, is equivalent to the momentum
equation and the surface traction on boundary conditions.
Collectively, the latter is called the classical strong form.
To develop the weak form, we require the trial functions u(X,t) to
satisfy all displacement boundary conditions and to be smooth
enough so that all derivatives in the momentum equation are well
defined. So also are the test functions u(X)!
The weak form is obtained by taking the product of the momentum
equation with the test function and integrating over the domain.
Smoothness of test and trial functions:
Kinematic admissibility: The weak form is integrable if the test and
trial functions are Co.
In the weak form for the TLF, all integrals are over material
domains, i.e., the reference configuration.
Each of the terms in the weak form represents a VIRTUAL WORK
due to the virtual displacement u. The test function u(X) is often
called a virtual displacement to indicate that it is not the actual
displacement. The virtual work of the body forces b(X,t) and the
prescribed tractions tox is called the virtual external work since it
results from the external (applied) loads.
The external work is

 
Xb

δW ext   δu ρ  δu 0
o A 0 b dX A 0 t x
Xa Γt
The virtual internal work, which arises from the stresses in the
material can be written as
Xb

δWint   δu,X  PA 0dX

Xa
Xb

  δF  PA 0dX
Xa

The term ρ0A0u can be considered as a body force which acts in the
direction opposite to the acceleration i.e., a D’Alembert force. We
will define the corresponding virtual work by Wkin and call it the
virtual inertial work or virtual kinetic work.
Xb

δW kin   δu ρ 0 A 0 u dX
Xa
• Principle of Virtual Work:
δW δu, u   δW  δW int ext
 δW kin
0
This in fact is the required WEAK FORM of the momentum
equation, the traction boundary conditions, and the stress jump
conditions. The equivalence of the strong and weak forms for the
momentum equation is called the principle of virtual work.
A key step in the development of the weak form is the integration
by parts. This eliminates the derivatives on the stress , without
this step stress  would have to be Co and u would have the C1!
Furthermore, the traction boundary conditions would have to be
imposed on the trial functions. It is therefore convenient to
integrate by parts and reduce the smoothness requirements on the
stress and hence the trial displacements.
FINITE ELEMENT DISCRETIZATION IN TLF:
•The discrete equations for a finite element are obtained from the
principle of virtual work by using finite element interpolants for the
trial and test functions.
•The finite element trial function:
nN
u(X, t)  Σ N I (X) u I (t)
I 1
•The test functions (or virtual displacements)
nN
δu(X)  Σ N I (X) δu I
I 1
• Nodal Forces:
δW int  δu T f int
δW ext  δu T f ext
δW kin  δu T f kin
fint are the internal nodal forces (corresponds to the stresses in the
material)
fext are the external nodal forces (corresponds to externally applied
loads)
fkin are the inertial nodal forces (corresponds to the inertia)
•Nodal forces are always defined so that they are conjugate to the
nodal displacement u in the sense of work so that the scalar product
of an increment of nodal displacement with the nodal forces gives an
increment of work.
• Semi-discrete equations
Ma = fext – fint or f = Ma Where f= fext – fint
The above is the semi discrete momentum equation, which is
called the equation of motion. These equations are called semi-
discrete because they are discrete in space but continuous in
time.
The equations of motion are a system of nN-1 second-order
ordinary differential equations. The independent variable is the
time t.These equations are easily remembered as f = Ma
Newton’s second law of motion.
The mass matrix infinite element discretiztion is often not
diagonal, so the equations of motion differ from the Newton’s
second law of motion in that a force at node I can generate
accelerations at node J if MIJ 0.
A diagonal approximation to the mass matrix is often used. In that
case, the discrete equations of motion are identical to Newton’s
equations for a system of particles inter connected by deformable
element.
 ELEMENT AND GLOBAL MATRICES
•In FEA programs, the nodal forces (internal and external) and
mass matrix are usually computed on an element level.
•The element nodal forces are combined into the global matrix by
an operation called scatter or vector assembly.
•The mass matrix and other square matrices are similarly
combined from the element level to the global level by an
operation called matrix assembly.
•The element nodal displacements are extracted from the global
matrix by an operation called gather.
•The connectivity matrix is a Boolean matrix, i.e., it consists of the
integers 0 and 1.
•The total internal energy is the sum of the element internal
energies.
 TWO-NODE LINEAR DISPLACEMENT ELEMENT
•Consider a two-node ROD element as shown in Figure 2.4.
•Displacement field:
u1 (t) 
uX, t   X 2  X1 , X  X1  
1

l0 u 2 (t) 
•Strain measure:
u1 (t) 
ε(X, t)  u ,X   1  1 
1

l0 u 2 (t) 

The Bo matrix is
Bo = 1/lo [-1, +1]
• Nodal Internal Forces
 1
X2
1
f int
  B P dΩ 0  
Γ
 PA 0 dx
 1
e 0
Ωe
l
X1 0
0

int
 f1   1
    A 0P 
f 2 e  1

• Nodal External Forces: body forces

f eext   ρ0 N T bA 0 dX Note : b  b(X, t)
Ω e0
• Element mass matrix:

M e   ρ 0 N N dΩ 0
T

Ω e0

ρ 0 A 0l0 2 1 
 1 2 
6  
THREE-NODE QUADRATIC DISPLACEMENT ELEMENT
•Consider the three-node element of length lo and cross section
area Ao shown in Figure 2.7.
•The mapping between the material coordinates X and the element
(natural) coordinates ξ is

 X1 
1  
Xξ   Nξ  X e   ξξ  1,1  ξ , ξξ  1 X 2 
2 1

2 2  
X 3 

N(ξ) is the matrix of Lagrange Interpolation or shape functions and

ξ  1,1
•The displacement field is given by the same shape functions
(Interpolants)

 u1 t 
 
u ξ, t   Nξ u e t   Nξ  u 2 t 
u t 
 3 
•Strain measures
ε  B0 u e

B0 
1
2ξ  1,  4ξ 2ξ  1
2X ,ξ
•Internal nodal forces
int
 f1 
 
f eint   B0 P dΩ 0  f 2 
T

Ω e0 f 
 3 e
•The external nodal forces
1 
ξ ξ  1
1  2 

1  1  ξ2 
 ρ 0 bA 0 X ,ξ dξ
1 
 2 ξ ξ  1
 
f eext 
1 
ξ ξ  1
1  2 

  
   1  ξ 2  A 0 t 0x
1  1 
2 ξ ξ  1
  Γ et
•Element mass matrix
1 
ξ ξ  1
1  2 

 
2  1
  
M e    1  ξ   ξ ξ  1, 1  ξ , ξ ξ  1
2 1

1  1 2 2 
 2 ξ ξ  1
 
 ρ 0 A 0 X ,ξ dξ
4 2  1
ρ0A 0L0 
Me  2 16 2 
30 
 1 2 4 
1 0 0
ρ0A0L0 
Me 
diag
 0 4 0
6
0 0 1