Implicit Newer

LSDYNA IMPLICIT NOTES
Copyright 2000-2004
All rights Reserved

Notes Developed by

Ala (Al) Tabiei, Ph.D.

atabiei@aol.com

Mailing Address

Al Tabiei
4056 Granite Ct.
Mason, OH 45040
USA

Tel: (513) 336-8075
Fax: (513) 336-8075

EMAIL: atabiei@aol.com

ACKNOWELEDGMENTS

The author wishes to thank several people for valuable comments and
discussions on the notes. Firstly, Dr. J. Hallquist advance course notes were
of great help in constructing the lecture notes. Many thanks goes to Drs. B.
Maker and R. Grimes for the valuable comments, suggestions, and
discussions. Thanks also go to the LSTC staff for their support.

TABLE OF CONTENTS

Introduction 1-5

Finite Element Equation & Time Integration 2-14

Equilibrium and Nonlinearity 3-25

Incremental Equations and Linearization 4-25
Automatic Time Step Control
Nonlinear Solution Methods for Implicit Analysis
Linear Equation Solver

Finite Element Modeling Techniques 5-66
1. Engineering a FEA Model:
2. Element Selection
3. Mesh Density
4. Symmetry
5. Modeling for Physical Phenomenon
6. Ad Hoc Guidelines
7. How to Tell If Your Results Are Correct

Contact and Friction 6-97
Contact Algorithm
Friction

Material Nonlinearity 7-112

Damping 8-119

Reduce Integration & Hourglass Phenomenon 9-122
Reduced and Selective Reduce Integration
Hourglassing

Quasi-Static Initialization and Quasi-Static Simulations 10-107
Quasi-Static Initialization
Quasi-Static Simulations With Explicit FE

Adaptive Meshing 11-132

Springback 12-136

Trouble Shooting 13-142

Workshop 14-145
INTRODUCTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________
LS-DYNA IMPLICIT NOTES 1-5

INTRODUCTION

Version 950 (1999) included the first release of implicit analysis
capability using LS-DYNA. Initially targeted at metal forming springback
simulation, this new capability allows static stress analysis without interfacing
to LSTC's implicit software package LS-NIKE3D.

The new implicit features in LS-DYNA were not simply created using
subroutines from LS-NIKE3D. Instead, the existing element and material
routines in LS-DYNA were augmented to add stiffness matrix calculations
when the implicit mode is active. This approach has several advantages:

keyword input and arbitrary ID numbering are available in implicit mode
existing optimized subroutines and data structures are used
parallel processing capability is immediately available
post-processing interfaces remain unchanged

This means that all of LS-DYNA's optimized programming and data
structures are used in the implicit mode, so high speed is preserved.
Furthermore, pre- and post-processing interfaces are virtually unchanged, so
that existing support software can be used with LS-DYNA's new implicit
features. This also provides the first opportunity for LSTC customers to use
an implicit code with keyword or structured format input, and fully arbitrary ID
numbering.

One disadvantage to this low-level, closely coupled implementation is that all
of the functionality for the implicit is not immediately available in LS-DYNA.
Instead, implicit stiffness matrix calculations must be added individually to
activate each LS-DYNA feature for implicit mode. Contact LSTC or your LS-
DYNA distributor if you need a feature which is not shown, since this list is
expanding rapidly.

INTRODUCTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________

ACTIVATING IMPLICIT MODE

LS-DYNA's implicit mode may be activated in two ways. Using one of
the new *CONTROL_IMPLICIT keywords, a simulation may be flagged to
run completely in implicit mode. Alternatively, an explicit simulation may be
seamlessly switched into implicit mode at a specific time using the
*INTERFACE_SPRINGBACK_SEAMLESS feature or using a load curve.
The seamless switching feature is intended to simplify metal forming
springback calculations, where the forming phase can be run in explicit
mode, followed immediately by an implicit static springback simulation. In
case of difficulty, full restart capability is supported.
Eight new keywords are available to support implicit analysis, along with
corresponding new control cards in the structured format input deck. Default
values are carefully selected to minimize input necessary for most
simulations. These are summarized below:

Activating Implicit Analysis

several types of analyses can be performed
fully explicit (default)
fully implicit
explicit followed by implicit ("seamless" springback)
switch between implicit and explicit
eigenvalue analysis
normal modes, etc.

New keywords have been added

*CONTROL_IMPLICIT_GENERAL
*CONTROL_IMPLICIT_SOLVER
*CONTROL_IMPLICIT_SOLUTION
*CONTROL_IMPLICIT_AUTO
*CONTROL_IMPLICIT_STABILIZATION
*CONTROL_IMPLICIT_DYNAMICS
*CONTROL_IMPLICIT_EIGENVALUE
*CONTROL_IMPLICIT_MODES

INTRODUCTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________
not all features available in the explicit are in the implicit mode.

Implicit Keywords

*CONTROL_IMPLICIT_GENERAL (required for implicit)
activates implicit mode
defines implicit time step size (standard LS-DYNA termination time is used
too)
optional if *INTERFACE_SPRINGBACK_SEAMLESS is used

*CONTROL_IMPLICIT_SOLVER (optional)
parameters for linear equation solver, which inverts stiffness matrix:
[K]{x}={f}
does NOT invoke a "linear" analysis

*CONTROL_IMPLICIT_SOLUTION (optional)
parameters for nonlinear equation solver (Newton-based methods)
controls iterative equilibrium search, convergence
"linear" analysis selected here (a special case where no iterations are
performed)

*CONTROL_IMPLICIT_AUTO (optional)
activates automatic time step control
back up, try again if equilibrium iterations fail
default is fixed time step size, error termination if any steps fail to
converge

*CONTROL_IMPLICIT_DYNAMICS (optional)
include inertia terms:
n n
ext
n n n
Ma f f u K a M +
+ +
int
1 1 1

problem "time" must now be real, physical time
can improve convergence, especially when (even weak) rigid body modes
are present

*CONTROL_IMPLICIT_EIGENVALUE (optional)
signals LS-DYNA to perform eigenvalue analysis, then stop
number of eigenvalues/vectors, optional frequency shift

*CONTROL_IMPLICIT_MODES
INTRODUCTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________
to calculate constraint and/or attachment modes for later use in modal
analysis
computed modes are written in a binary file called d3mode

*CONTROL_IMPLICIT_STABILIZATION (optional, metal forming only)
allows multi-step springback by unloading internal stresses over several
steps
automatic springs attached to shell element nodes, slowly removed
stabilization must be "completely removed" for accurate results

User's Manual contains helpful notes on each input parameter

INTRODUCTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________
A Note on Implicit Numerical Methods

Implicit governing equations contain two problems to solve

n n
ext
n n n
Ma f f u K a M +
+ +
int
1 1 1

Nonlinear Problem: *CONTROL_IMPLICIT_SOLUTION
find displacements u which satisfy equilibrium f
ext
=f
int

both K, f
ext
and f
int
can be nonlinear functions of u
iterative search employed using Newton-based method
interactive switch "<ctrl-c> nlprint" toggles diagnostic output

Linearized Problem: *CONTROL_IMPLICIT_SOLVER
solve system of linear algebraic equations
must solve during every nonlinear iteration
great CPU and memory cost makes this problem important
interactive switch "<ctrl-c> lprint" toggles diagnostic output

Implicit Eigenvalue Analysis

Extract "n" eigenvalues by subspace iteration
lowest "n" frequencies
optional frequency shift: "n" eigenvalues nearest to shift frequency
initial stress influences frequencies when S/R Hughes-Liu shell is used

Two additional databases output
d3eigv: binary plot database with each mode shape
eigout: text file summarizing frequencies found

Simple input parameters

activates implicit analysis

*CONTROL_IMPLICIT_EIGENVALUE
activates eigenvalue solver
input number of eigenvalues, optional shift frequency

INTRODUCTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*CONTROL_IMPLICIT_BUCKLE
Implicit buckling analysis is performed at the end of a standard implicit
simulation. You must perform the standard simulation first to apply
load and create stress. Then we build material and geometric stiffness
matricies using the model state at the termination time, and solve the
buckling eigenproblem. The resulting eigenvalues are multipliers which,
when applied to the loading at the termination time, give the buckling
loads.

Implicit Output Databases

Binary Output Databases
plot, time history files in standard LS-DYNA format
iteration plot database "d3iter" activated interactively "<ctrl-c> iter"
deformed geometry each nonlinear iteration
good for debugging convergence problems

ASCII Output Databases
NODOUT, ELOUT, GLSTAT, RCFORC, NODFOR, RBDOUT, etc
same format as explicit

FUTURE DEVELOPMENTS

The list of features available for implicit analysis within LS-DYNA is growing.
Customers are encouraged to notify either LSTC or LS-DYNA distributors
about important features needed for their applications. Development priority
will be strongly customer driven. LSTC anticipates that an important future
application area will be implicit stress initialization, followed seamlessly by an
explicit simulation. This will likely become the most common application of
implicit LS-DYNA.
INTRODUCTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________
PROGRESS OF ITERATIONS

Control parameters for the nonlinear equation solver are input using the
*control_implicit_solution keyword. By defaults, the progress of the
equilibrium search is not shown to the screen. This output can be activated
using the NLPRINT input parameter, or interactively toggled on and off by
entering <ctrl-c> nlprint. The box below shows a typical iteration sequence,
where the norms of displacement and energy are displayed. When these
norms are reduced below user prescribed tolerances, equilibrium is reached
within sufficient accuracy, the iteration process is said to have converged,
and the solution proceeds to the next time step.

INTRODUCTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________
The next box depicts a typical print-out of the defult BFGS nonlinear equation
solver. Several automatic stiffness reformation are performed, initialy due to
divergence, and later when the defult limit of 10 iterations is exceeded.

INTRODUCTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________
IN DYNAMIC SIMULATION---TIME STEPS
IN STATIC SIMULATION---LOAD STEPS

(multiple steps may be used to divide the nonlinear behavior into
manageable pices)

FE EQUATIONS AND TIME INTEGRATION
_______________________________________________________________________________________________
_______________________________________________________________________________________________


The equations of equilibrium governing the dynamic response of the system
are:

[ ]{ } [ ]{ } [ ]{ } { } ) (t R u K u C u M
ext
+ + & & & (1)

where
[M] is the mass matrix,
[C] is the damping matrix,
[K] is the stiffness matrix,
{ } u& & , { } u& , and { } u are the nodal accelerations, velocities, and displacements
vectors respectively,
{ } ) (t R
ext
is the external forces vector.

Equation (1) can be rewritten in the form

{ } { } { } { } ) ( ) ( ) ( ) ( t R t R t F t F
ext int
D I
+ + (2)

where
{ } [ ]{ } u M t F
I
& & ) (
{ } [ ]{ } u C t F
D
& ) (
and
{ } [ ]{ } u K t R
int
) (

are the inertia force, the damping force, and the internal forces vectors all of
which are time dependent. Equations (1), which represent a system of
nonlinear second order differential equations, are solved with the help of the
LSDYNA Explicit or Implicit finite element codes using direct time integration
techniques.

Direct Time Integration
When solving dynamic problems with the finite element method, the solution
is sought by dividing the total response time of the system into much smaller
time intervals called time steps or time increments. The equilibrium equations
are solved and the values of the unknowns are determined at time t+ t
based on knowledge of their values at time t (quasi-linearization).
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Using these values we continue solving the linearized differential equations
at t + 2t, and so on, for the entire response time of the system.
When employing time integration, LSDYNA uses the implicit integration
operator and the explicit integration operator.
The implicit integration operator definition is completed by the Newmark
formulae for displacement and velocity integration

( ) [ ]
( ) [ ], 1
2 / 1
2
t t t t t t
t t t t t t t
u u t u u
u u t u t u u
+ +
+ +
+ +
+ + +
& & & & & &
& & & & &

(3)
where and are parameters of the system. Thus, expressing the velocities
and accelerations at t + t in terms of the displacements at t + t and
substitution into equations (1) yields:

[ ]{ } { }
t t t t
F u K
+ +
(4)

where
[ ] [ ] [ ] [ ] ( ) [ ] t C M K K K , , , is the effective stiffness matrix, and
{ } { } [ ] [ ] { } { } { } ( ) { }
t t t t t t t
u u u t C M F F F & & & , , , , , ,
+ +
is the effective load at time t + t.

The explicit dynamic analysis is based on integrating the equations of motion
for the system using the explicit central difference formula

( )
( )
t t t t t
t t t t t t
u u
t
u
u u u
t
u
+
+
2
1
2
1
2
&
& &
(5)

Thus equations (1) take the form

[ ]{ } { }
t t t
F u M
+ 2 /
(6)
where

[ ] [ ] [ ] ( ) [ ] t C M M M , , is the effective mass matrix, and
{ } { } [ ] [ ] [ ] { } { } ( ) { }
2 /
, , , , , ,
t t t t t
u u t C M K F F F

is the effective load at time t.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Equations (4) are solved at time t + t and the corresponding solution
methods are called implicit, while equations (6) are solved at time t, and the
corresponding methods are called explicit.

The explicit central difference operator is very convenient when lumped mass
matrix can be assumed and velocity-dependent damping can be neglected.
Then [ ] M is a diagonal matrix and the solution is achieved automatically
without having to solve the system of equations.

On the other hand the explicit operator is only conditionally stable, which
means that in order to obtain accurate results, the time step must be smaller
than a certain critical value which is defined by the mass and stiffness
properties of the complete element assemblage.

Depending on the particular problem, the stable time increment of the model
may be very small which would require too many steps to solve and therefore
extremely large CPU time. In such cases the implicit method, which is
unconditionally stable, may provide better efficiency.
_______________________________________________________________________________________________
_______________________________________________________________________________________________

EQUATION OF MOTION FOR TRANSLATIONAL VELOCITIES:

} { } { } { } {
int bod ext
R R R
dt
dv
M + , however in here we have additional terms as
follows:
} { } { } { } { } { } {
int cont hgr bod ext
R R R R R
dt
dv
M + + +
elements
V
J I
dV N N M is the mass matrix
elements
ext
R } { S d p N
S
i I
is the externally applied loads vector
elements
R } {
int
dV
N
V j
I
ij

is the internal forces vector

dV b N R
i
elements
V
I
bod

} {
}
is the body forces vector
elements
hgr hgr
r R } { is the hourglass resistant forces vector
surfaces contact
cont cont
r R } { is the contact forces vector
V and S are the actual volume and external surface

_______________________________________________________________________________________________
_______________________________________________________________________________________________

Flowchart for Explicit Time Integration

Initial conditions ); 0 (
0
t 0 ), 0 (
2
1

n t u u & &
)] 0 ( [
1
) 0 (
int
t u C R R
m
t u
o
ext
o
ii
& & &
Update: Velocities
n
n n
u t u u & & & & +
+
2
1
2
1

Displacements
2
1
1
+
+
+
n
n n
u t u u &
Compute internal nodal forces

Loop over elements: E = 1 to NELE
Velocity-strains:
2 / 1 2 / 1 + +
n
E
n
u B & &
Stress rates: ) (
2
1
2
1
+ +
n n
T &
Update stress:
2 / 1 1 + +
+
n n n
t &
Internal forces:
+
+
V
n T
n
V d B r
1
1
int

Assemble:
1
int
+ n
r into
1
int
+ n
R
Compute external nodal forces
1 + n
ext
R
Accelerations: ] [
1
2 / 1 1
int
1 1 + + + +

n n n
ext
ii
n
u C R R
m
u & & &
Output n n +1; go to 2

_______________________________________________________________________________________________
_______________________________________________________________________________________________

Implicit Governing Incremental Equations

To obtain solution at load increment n+1 given solution at load increment n,
the linearized equations are:

) ( ) ( ) (
int
1 0
..
1
n
n
n ext n
t n
x R x R u x K u M +
+ +

M : lumped mass matrix
t
K : tangent stiffness matrix based on geometry at n
ext
R : external load based on applied load at step n+1, but geometry at n
int
R : internal load based on displacement state and stress at load step n

Using Newmark method lead to the following equation

) ( ) (
1 0
n
n
n ext
x R x R u K

+

where
M
t
K K
t
2
1

] )
2
1
(
1
[ ) (
..
.
int
n
n
n
u
t
u
M x R R

+

Solution of the above equation yields
0
u . The coordinate vector is updated
using

0 1
u x x
n n
+
+

The iteration for equilibrium now begins using:

i j Q x R x R u x K
n
i
n
i n
n
i
ext
i
n
j j

+ +
+
+ 1 1
1
1
) ( ) ( ) (

Where i denote the iteration number.

Convergence is assumed if
e d
and is some defined values.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Advantages as DT becomes small
mass contribution to stiffness matrix grows rapidly: ( )M K K
2
1
t

stabilizing effect to equilibrium iterations

_______________________________________________________________________________________________
_______________________________________________________________________________________________

Implicit Flow Chart

1. Set 0 n , ) 0 ( ), 0 (
. .
0 0
t u u t u u select , , t

2. Form

0 0 0
int
0
, , , , R and K K R M
t

3. Calculate ) (
int
0 0
1 ..
0
R R M u
ext

4. Set 0 i

5. Solve for
0
i
u using ) ( ) ( ) (
int
1 0
n
n
n ext n
x R x R u x K
+

6. Update coordinate vector
0
1
u x x
n n
+
+

7. Calculate ) ( ,
1
1
+
+
n
n
x R

8. Set 1 + i i

9. Solve for
i
u using

10. i j Q x R x R u x K
n
i
n
i n
n
i
ext
i
n
j j

+ +
+
+ 1 1
1
1
) ( ) ( ) (

11. If convergence is not attained and NO divergence

12. Set
i
n
i
n
i
u x x +
+ + 1 1
and go to 8

13. Otherwise 1 + n n and

14. Set
n
i
n
i
x x
+
1
0
and ) ( ) (
1
0
n
i
n
i
x R x R
+

15. Calculate
.
n
u and
..
n
u

16. Go to 4 if t<T, otherwise terminate.

Divergence is defined as when the residual force increases during iterations

_______________________________________________________________________________________________
_______________________________________________________________________________________________

Implicit vs. Explicit FEA

Explicit FEA codes have enjoyed much success
l ease of use
l robustness
l capacity

However, latest generation computers offer new opportunities for implicit
FEA codes
l huge memory is now common (RAM > 1.0 Gbyte)

Several crucial enhancements have been added, allowing implicit codes to
offer a credible alternative to explicit
l material, contact algorithms (rigid tooling)
l fast linear equation solver
l robust nonlinear solver
l keyword input, unified I/O databases

Summery

Conservation of momentum: f u +
tt
,
LS-DYNA - Explicit time integration
int
n
ext
n n
f f Ma
impact, penetration, high rate dynamics
many small time steps
Courant condition limits largest stable time step

l LS-DYNA - Implicit time integration
n n
ext
n n n
Ma f f u K a M +
+ +
int
1 1 1

static, eigenvalue, low rate dynamic analyses
few large time steps
Linear equation solver - stiffness matrix
Nonlinear equation solver - user chooses step size

Quasi-Static analysis: "time" represents a monotonically increasing
parameter which characterizes the evolution of the loading.

LS-DYNA offers several implicit nonlinear solvers:
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Full Newton - reform
K every iteration
Modified Newton - reform
K every j iterations
Quasi-Newton - employ simple, approximate "update" to
K

In every case, accuracy is maintained by re-evaluating
R each iteration
_______________________________________________________________________________________________
_______________________________________________________________________________________________

EQUILIBRIUM AND NONLINEARITIES
_______________________________________________________________________________________________
_______________________________________________________________________________________________

The concept of mechanical equilibrium is:
external
v
?
R d B

Geometric Nonlinearities

external
v
T
R dV B

Geometric nonlinearities are caused by:
Nonlinear relationship between displacement increments and strain
increments
Integration over current volume

Effects:
Stress- stiffening
Bifurcation, buckling, and collapse
Snap-through
Necking
Other large displacement effects

Material Nonlinearities

v
external
T
R dV B

Material nonlinearities are caused by:
Dependence of stress on current strain

Effects:
Plasticity
Plastic hinge formation and plastic collapse
Rubber nonlinear elasticity
Other nonlinear material effects
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Boundary Nonlinearities

external
v
T
R dV B

Boundary nonlinearities are caused by:
Dependence of
external
R on current displacements

Effects:
Contact across body surfaces (by means of load transfer)
Nonlinear external loads
Pressure load nonlinearities
Follower forces

Linear Equilibrium

For each , F there will always be only one solution u (existence and
uniqueness).
If F causes a displacement of u , then ( ) aF causes a displacement ( ) au
(scaling).
If F causes a displacement u and P causes a displacement v then
P F + causes a displacement v u + (superposition).
In a linearized problem, the solution u is determined by the current
value of the load
external
R .

external
R Ku

Nonlinear Equilibrium

For a particular F , there may be none, one, many, or an infinite number
of solutions u (non-existence and non-uniqueness).
If F causes a displacement of u , then ( ) aF probably doesnt cause ( ) au
(no scaling)
If F causes a displacement u and P causes a displacement , v then
( ) P F + probably doesnt cause ( ) v u + (no superposition).
In a real (nonlinear) problem, the unique solution is determined by the
entire load history of
external
R
_______________________________________________________________________________________________
_______________________________________________________________________________________________

v
external
T
R dV B

Examples of non-unique solutions

Shallow spherical caps
Plasticity

Linear vs. Nonlinear FEM

The FEM equation corresponding to linearized equilibrium is a matrix
equation in
dof
N unknowns:
extermal
R Ku
The FEM equation corresponding to true (nonlinear) equilibrium is a
vector equation in
dof
N unknowns:

v
T
R dV B
The solution of a linear FEM model involves the solution of the linear
dof dof
N N

matrix equation:
external
R Ku
The solution of a nonlinear FEA model involves the simultaneous
solution of
dof
N nonlinear vector equations:

v
external
T
R dV B
There are no good, general methods for solving systems of more than
one nonlinear equation. There never will be any good, general methods.

Nonconvergence

What LS-DYNA Does:

Solves the equation iteratively by using some of the Newton methods (or
variations of these methods) to find an approximate solution that minimizes
the residuals!
What are residuals ?

Re-write the equilibrium equations as a function r of u :
( )

v
T
R dV B u r
( ) u r are the residuals (out-of-balance forces) at u . ( ) u r is nonlinear!
In general, ( ) 0 u r , but, in equilibrium, ( ) 0 u r

Interpretation

_______________________________________________________________________________________________
_______________________________________________________________________________________________
The residuals are just the distribution and the magnitude of extra external
force at each DOF needed to bring the structure into equilibrium at u :
( )

v
external
T
R dV B u r 0 , u is not in equilibrium under
external
R
( ) { }
+
v
external
T
u r R dV B 0 , u is in equilibrium under ( ) u r R
external
+

Newton Method

Assume you are not in equilibrium at displaced position u so that 0 ) ( u r
Find u so that u u + is in equilibrium ( ) 0 + u u r
Expand ( ) u r in a Taylors series about current displacement u :
( ) ( ) 0 ... +
+ + u
u
r
u r u u r
u

Throw away higher order terms and solve the resulting equations for u :
( ) u r u
u
r
u

Substitute the equilibrium equation into Newton-Raphson scheme :
( ) u r u
u
r
u

+
v
external
T
gent
R dV B u K
tan
this equation is linear!
The solution of a linear FEA model involves single solution of the linear
dof dof
N N matrix equation:
external
F Ku
The solution of a nonlinear FEA model involves many solutions of the
linear
dof dof
N N matrix equation:
external
v
T
gent
R dV B u K +

tan

INCREMENTAL EQUATIONS & LINEARIZATION
_______________________________________________________________________________________________
_______________________________________________________________________________________________


MATRIX EQUATIONS

( ) ( ) 0 , , ) ( + t x P x x F t x M & & &

x x x , , & & & n dimensional acceleration, velocity, and geometry vectors
M nn mass matrix
P body force and external load vector

e
V
e
t
e
dV B F

NONLINEARITIES

Nonlinearities are due to geometric affects and inelastic material behavior

( ) ( )
e
V
e
t
e
dV B x x F & & , ,

Where and & are the strains and the strain rates. If linear

( ) x K x C x x F
e
+ & & ,

Where K, C and x are the stiffness matrix, damping and displacement
vector. Nonlinearities also arise in P due to the geometry dependant
loadings.

RESIDUAL VECTOR

Regardless of whether an implicit or explicit scheme is used, we require that

0 + P F x M Q & &

If linear

0 + + P x K x C x M Q & & &
_______________________________________________________________________________________________
_______________________________________________________________________________________________

LINEARIZATION

The residual vector becomes an implicit function of
1 + n
x only. We seek he
vector
1 + n
x such that

( ) 0
1
+ n
x Q
Assume an approximation
1 + n
k
x to
1 + n
x for k=1,2,3, ..
In the neighborhood of
1 + n
k
x we use the linear approximation to ( )
1 + n
x Q given
by:

( ) ( ) ( )
k
n
k k
n
k
n
x x J x x Q x Q +
+
+ 1
1
1
1
1

k
n
k
n
k
x x x +
+
+ 1
1
1

( )
1 1
1
1
|
+ +
n
x
n
k
k
x
Q
x J

The Jacobian matrix is expressed as

x
P
x
F
x
x
x
F
x
x
M J
&
&
& &

x
P
K
x
x
C
x
x
M J
t
& & &

where
x
F
C
&
is the tangent damping matrix

x
F
K
t
is the tangent stiffness matrix.

STRAIN CALCULATION

To prevent locking during incompressible flow, we impose a constant
volumetric strain over the element. A consistent B matrix for the constant
pressure assumption is employed in the tangent stiffness matrix. B is defined
by Hughes in: T. R. Hughes, Generalization of selective Integration
Procedures to Anisotropic and Nonlinear Media, Int. J. for Numerical
Methods in Engineering, Vol. 15 no. 9, pp. 1413-1418 (1980).
Based on midpoint geometry

_______________________________________________________________________________________________
_______________________________________________________________________________________________
,
_
+
+
+
+
2
1
2
1
2
1
2
1
2
1
n
i
n
j
n
j
n
i
ij
x
x
x
x
&
&
&

this strain measure is a valid large strain measure , i.e., rigid body rotations
do not cause straining

IMPLICIT ALGORITHM

( ) ( ) ( )
n
n
n
o
n
t
x F x P u x K
+1

t
K = Positive definite tangent stiffness matrix.
o
u = Desired increment in displacements.
( )
1 + n
n
x P = External load vector at n+1 based on the geometry at time n.
( )
n
x F = Stress divergence vector at time n.
Update displacement vector

0 0
1
1
u s x x
n
n
+
+

and begin iterations for equilibrium

( ) ( )
1 1
1
1 + +
+
+

n
i
n
i
n
n
i i t
Q x F x P u K
j

where the subscript i denotes the iterate and j<i.
0
s is a parameter between 0
and 1 found from a line search.

After each iteration, convergence is checked, convergence is assumed if the
conditions:

d
i
u
u
<
max

and
e
i
t
i
t
i
Q u
Q u
<
0

are satisfied.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
If convergence Is not attained , we update the displacement vector

i i
n
i
n
i
u s x x +
+ +
+
1 1
1

and perform anther iteration. Lack of convergence within allowable number of
iteration (default=10) or divergence

1
1
1
0
+
+
+
<
n
i
n
Q Q

Causes
t
K to be reformed. Termination occurs if allowable number of
reformations is reached (default =15).
The foregoing iteration method with
t
K held constant is called the modified-
Newton method. When convergence problems arise the stiffness matrix is
reformed.
Four methods for updating the stiffness matrix are available:
1. BFGS
2. Broydens first method
3. Davidon
4. DFP
These are called quasi-Newton methods.

Quasi-Newton methods involve

Line searches
Stiffness updates

Slightly more expensive than modified Newton but results in more stable
program

QUASI-NEWTON SCHEMES

The secant matrices
i
K are found via the Quasi-Newton equations

i i i
Q u K
1

where
) ( ) (
1 1 1 i i i i i i i
F P F P Q Q Q

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Two classes of matrix updates that satisfy the Quasi-Newton equations are
of interest
Rank one updates
t
i i
zx K K +
1

Rank two updates
t t
i i
vy zx K K + +
1

Recall the Quasi-Newton equation
i i i
u u K
1

Subsitituting

t
i i
zx K K +
1

we find
i i
t
i i
Q u zx u K +
1 1 1

By choosing
1 1
1
/ 1

i i i
i
t
u K Q z
u x

we can satisfy the equation.
Note that x is an arbitrary vector but is restricted such that

0
1

x u
t
i

Broyden set
1

i
u x and obtained the update

t
i
i
t
i
i i i
i i
u
u u
u K Q
K K
1
1 1
1 1
1

+

resulting in non symmetric secant matrices.
Davidon chose

1 1

i i i
u K Q x

and obtained the updated formula

t
i i i
i
t
i i i
i i i
i i
u K Q
u u K Q
u K Q
K K ) (
) (
1 1
1 1 1
1 1
1

+

resulting in symmetric secant matrices.
The inverse forms are found by the Scherman-Morrison formula:

_______________________________________________________________________________________________
_______________________________________________________________________________________________
a A b
A ab A
A ab A
t
t
t
1
1 1
1 1
1
) (

+
+

where A is a nonsingular matrix
We can find the inverse form for Broydens update by letting

1 1
1 1

i i
i i i
u
t
u
u K Q
a
1

i
u b

in the Sherman-Morrison formula. Therefore,

i i
t
i
i
t
i i i i
i i
Q K u
k u Q K u
K K

+

1
1 1
1
1 1
1
1 1 1
1
1
) (

The inverse form for Davidons method can likewise be found:

i
t
i i i
t
i i i i i i
i i
Q Q K u
Q K u Q K u
K K

+

) (
) )( (
1
1 1
1
1 1
1
1 1 1
1
1

Again recall the Quasi-Newton equation
i i i
Q u K
1

Substituting
1 1
+ +
i
t t
i i
u vy zx K K
We find
i i
t
i
t
i i
Q u vy u zx u K + +
1 1 1 1

Let
i
i i
i
t
i
t
Q v
u K z
u y
u x

1 1
1
1
/ 1
/ 1

Here x and y are vectors that are non-orthogonal to
1
i
u i.e.,

0
1

x u
t
i
and
0
1

y u
t
i
and 0
1

y u
t
i

In the BFGS method

_______________________________________________________________________________________________
_______________________________________________________________________________________________
1 1

i i
u K x
i
Q y

Leading to the following updates formula

1 1 1
1 1 1 1
1
1

+
i i
t
i
i
t
i i i
i
t
i
t
i i
i i
u K u
K u u K
Q u
Q Q
K K

That preserves the symmetry of the secant matrix. A double application of
the Sherman-Morrison formula leads to the inverse form.
Special product forms have been derived for the DFP and BFGS updates
and exploited by Matthies and Strang (1979).

) ( ) (
1
1
1 t
i i i
t
i i i
w v I K v w I K + +

The primary advantage of the product form is that the determinant of
i
K and
therefore the change in condition number can be easily computed to control
updates.
Update vectors
Let,

i i i
i i
Q Q
u

1
1

i i
i
i
i i
i i
i i i
w
Q
u
Q Q v

1
]
1
2
1
1
1
) ( 1

Determinant
2
1
) 1 )( det( ) det( v w K K
t
i i
+

Change in condition number

{ }
[ ]
2
4
2
1
2
1
2
1
) 1 ( 4
) 1 ( 4 ) ( ) (
w v
w v w vw v w w v v
c
t
t t t t t
+
1
]
1
+ + +

LINE SEARCHES
_______________________________________________________________________________________________
_______________________________________________________________________________________________

With Quasi-Newton updates line searches are necessary to achieve
convergence.
After u is determined, a line search may be performed
search only if global response is hardening
prevents divergence of Newton method
find multiplier which gives best estimate of equilibrium

compute trial displacement 1 0 , + s u s u u
i
tr

evaluate out-of-balance force ( ) ( ) ( )
tr tr ext tr
u f u f u R
int

search (iterate) to find s which minimizes out-of-balance force

if s falls below 0.001, discard u and reform K

i i
sd u

where

i i i
Q K d
1

i i i i
d s u u +
+1

Assume solution lies in direction
i
d . The line search then component s
i
such
that components of (P-F) in the search direction is zero

[ ] 0 ) ( ) ( .
1 1

+ + i i i i
x F x P d

Broyden:

1
1 1 1 1 1 1
1
) 1 )( 1 )....( 1 (

+ + +
o o o o
k u w u w u w K
t t
i i i i
where
1
1 1
+

+
i i i i i i
Q K Q K u w and ) ( / 1
i i
t
i i
w u u are the update factors.

Broydens Algorithm:

1. Solve
i
Q K d
1
o

2. Do 1 , 1 k i
Recall d u
t
i 1

_______________________________________________________________________________________________
_______________________________________________________________________________________________
1 1
+
i i
w d d
3.
) ( / 1
/
1 1 1 1
1 1 1

+
k k
t
k k
k i k
w u u
d s u u w

d u
t
k 1

4.
1 1 1
+
k k k
w d d

ARC-LENGTH METHOD

It is well known that the foregoing approach fails in the neighborhood of limit
points. Geomtrically, the plane of =constant does not necessarily intersect
the load deflection curve. In the arc length method of Riks/Wempner a
constraint equation is added to limit the load step to a constant arc length,
in load/displacement space.

Consider 0 p F

where is the proportional load factor. Newtons method leads to the
incremental equations

F p x K
i i j i

) (

in classical approach constant and iterations proceed until equilibrium is
reached.

ARC-LENGTH METHOD OUTLINED

1. Nonlinear equilibrium equation
0 p R

2. Linearized equation
) (
i i i
u R p u K
i i i
u x x +
+1

equivalently,
i i i
u u u +
+1

3. Constraint equation
2 2
s u u
i
t
i
+
_______________________________________________________________________________________________
_______________________________________________________________________________________________

4. Load level is split into 3 parts
i i m
+ +
m
= load level representing an equilibrium state at the beginning of a load
increment
i
= increment load level in iteration i
i
= change in load level in iteration i

5. Iterative displacements
( ) ( ) [ ] R P P K R P K u
m i i i
+ +

1 1

II
i
I
i i i
u u u +
P K u
I
i
1

( [ ] R P K u
i m
II
i
+

)
1

I
u represents displacement due to constraint condition
II
u represents usual displacement increment

6. Linearization of the constraint equation
0 ) , (
2 2
+ s u u u f
i
t
i

0 ) , ( +
i i i
i
i
i
u f
f
u
u
f

or 0 + +
i i i i i
f u

7. We can now solve for
i

Substituting
II
i
I
i i i
u u u +
Into 0 + +
i i i i
t
f u
Yields
I
i
t
i i
II
i
t
i i
i
u
u f
+
+

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Input Parameters for Nonlinear Solver

*CONTROL_IMPLICIT_SOLUTION:

NSOLVR: select Newton method
.eq.1: linear aproximation (no equilibrium iterations)
.eq.2: BFGS quasi-Newton method (DEFAULT)

ILIMIT: set limits for number of iterations before re-evaluating
.eq.1: new each iteration = Full Newton Method
.eq.10: use 10 inexpensive BFGS updates, reform if not yet converged

MAXREF: reformation limit before abandoning step
if AUTO is active, dt will be reduced and step will be re-tried, so MAXREF
can be smaller (~5)
if AUTO not active, error termination occurs when MAXREF is reached so
MAXRED should be larger (~15, default)

DCTOL, ECTOL: convergence tolerances
use NLPRINT=1 or "<ctrl-c> nlprint" to monitor progress of iterations

Nonlinear Solver Screen Dump

BEGIN implicit time step 3
============================================================
time = 1.09990E+00
current step size = 3.67821E-01

Iteration: 1 *|du|/|u| = 1.0894498E-01 *Ei/E0 = 1.8731172E+00

DIVERGENCE (increasing residual norm) detected:
|{Fe}-{Fi}| ( 1.0547507E+07) exceeds |{Fe}| ( 9.1389570E+06)
automatically REFORMING stiffness matrix...

Iteration: 2 *|du|/|u| = 3.8969724E-03 *Ei/E0 = 3.3420090E-02

DCTOL=1.E-03

ECTOL=1.E-02
ILIMIT=11
_______________________________________________________________________________________________
_______________________________________________________________________________________________
ITERATION LIMIT reached, automatically REFORMING stiffness matrix...


Equilibrium convergence summary for time step 3 at time = 1.0999005E+00
Number of iterations to converge = 13
Number of iterations to converge = 2

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Example: 1-D Nonlinear Spring

To illustrate the Newton method, consider a simple nonlinear spring:

Iteration #1:
( ) 5 . 2 , 4 0 , 0
4
10
1 0
u k u

If the nonlinear solution method is linear, stop here and assume
solution is correct. Do not bother to check for equilibrium.

In a nonlinear problem, evaluate equilibrium by determining the internal force
and the out-of balance force R:

( )
3
7 , 5 . 2
int
1
1 1

F F R
u F u
ext

Check convergence of equilibrium iterations
by evaluating displacement, energy norms

300 . 0
10 5 . 2
3 5 . 2
000 . 1
5 . 2
5 . 2
1
o o
R u
R u
e
e
u
u

f
o
r
c
e
,

F
(
u
)

displacement, u
1 3 2 4
4
8
12
5
f
o
r
c
e
,

F
(
u
)

displacement, u
1 3 2 4
4
8
12
5
k=4
k=2
k=1
F=10
k(u)
u
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Iteration #2: solve K?u=R, update u=u+?u

( )
0 . 4 , 5 . 1
2
3
2 5 . 2 , 5 . 2
2 1 2 3
2
+

u u u
k
R
u
k u

1 , 9 ) (
int
2 2
F F R u F
ext

evaluate new R, convergence norms

060 . 0
10 5 . 2
1 5 . 1
375 . 0
4
5 . 1
1
o o
R u
R u
e
e
u
u

Iteration #3: solve K?u=R, update u=u+?u

( )
0 . 5 , 0 . 1
1
1
1 0 . 4 , 0 . 4
3 2 3 3
2
+

u u u
k
R
u
k u

. 0 , 10 ) (
int
2 2
F F R u F
ext

evaluate new R, convergence norms

000 . 0
10 * 5 . 2
. 0 0 . 1
200 . 0
0 . 5
50 . 1
1
o o
R u
R u
e
e
u
u

f
o
r
c
e
,

F
(
u
)

displacement, u
1 3 2 4
4
8
12
5
f
o
r
c
e
,

F
(
u
)

displacement, u
1 3 2 4
4
8
12
5
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Selecting Implicit Time Step Size

When should load be applied in more than one step?
to reduce nonlinearity to manageable level
to accurately integrate material behavior
to resolve high frequency response in dynamic problems
input parameters: stepsize DT and termination time TERM

How many steps should be used?
estimate nonlinearity in problem
tradeoff:
few steps with many equilibrium iterations (unreliable)
many steps with few iterations (too expensive)
observe progress of equilibrium search
reduce stepsize until norms decrease monotonically
increase stepsize if very few iterations are needed (<10)
activate automatic time step controller

_______________________________________________________________________________________________
_______________________________________________________________________________________________

AUTOMATIC TIME STEP CONTROL

If convergence fails, LSDYNA backs up to beginning of step with smaller DT.
Solution is retried. If convergence again fails, solution is retried with a smaller
step and so on. Finally DT is set to DTMIN, if fails terminate.

An optional number of iterations, IOPT, per step is chosen. Let ILS be the
number of iterations during the last increment. If

ILS<IOPT
DT= DT+(DTMAX DT) MIN ]} 1 ) / [( , 5 . 0 {
2
1
ILS IOPT

ILS>IOPT
DT=DT-(DT-DTMIN) ] ) / ( 1 [
2
1
ILS IOPT

Where DT, DTMAX, and DTMIN are the current step size, the maximum step
permitted, and the minimum step size permitted, respectively.

Automatic time step control allows specification of the Optimum Number
of Equilibrium Iterations per step.

This indicated how hard LSDYNA should work in each time step.

If equilibrium is reached in fewer than optimum iterations, the size of the
next step is increased.

If equilibrium search requires more than the optimum number of iterations,
then the next step size is decreased.

CONTROL_IMPLICIT_AUTO

Automatic time step control adjusts stepsize during the simulation
very persistent, reliable

_______________________________________________________________________________________________
_______________________________________________________________________________________________
After successful steps
compare iteration count to target value ITEOPT
increase/decrease size of next step if difference exceeds window ITEWIN

After failed steps
decrease step size
back up, repeat failed step with new DT

Exponential algorithm for adjusting step size
increase stepsize by 1/5 decade until DTMAX is reached
decrease stepsize by 1/3 decade until DTMIN is reached
error termination if convergence fails when DT=DTMIN

Default time step control strategy
ITEOPT = 11, ITEWIN=5, DTMIN=0.001*DT0, DTMAX=10*DT0
decrease step if more than 16 iterations were required
increase step if less than 6 iterations were required
problem: stepsize continues to decrease even though every step
converges successfully
some models simply require more than 16 iterations per step

Aggressive step control strategy
ITEOPT=200, conservative value for DTMAX
stepsize always increased if convergence is successful
agressively pushes stepsize toward DTMAX
stepsize still decreases if convergence fails, but increases rapidly
thereafter
recommended for many problems

_______________________________________________________________________________________________
_______________________________________________________________________________________________

NONLINEAR SOLUTION METHODS FOR
IMPLICIT ANALYSIS


Define control parameters for the implicit nonlinear equation solver:
There are two groups for nonlinear solution. One group with four solution
methods. The second group is the same as the first group, however, with the
arclength method.

If NSOLVR is equal to:

EQ 2: BFGS updates
EQ 3: Broyden updates
EQ 4: DFP updates
EQ 5: Davidon updates
EQ 6: BFGS updates + arclength
EQ 7: Broyden updates + arclength
EQ 8: DFP updates + arclength
EQ 9: Davidon updates + arclength

Standard Newton Method:

The nth load increment is applied-then using the tangential stiffness matrix
the iterative displacent are found and hence the residual forces.The new
tangential stifness is then evaluated and the new iterative displacement and
redidual forces are found. The process is repated until convergence is
obtained. The figure below depicts the algorithm.

_______________________________________________________________________________________________
_______________________________________________________________________________________________

Equilibrium is reached when iterations converge: 0
, 0
) (

i
R u

Modified Newton-Raphson Method

The difference here between this method and the one before is that there is
no update of the tangential stiffness matrix after the initial iteration. The
figure below depints the algorithm.

( ) ( )
) ( int ) ( ) (

i ext i i
u f f R u u K
F
o
r
c
e

N
o
r
m

Displacemnt
Norm
u n u ( n + 1 )
1
u ( n + 1 )
2
u ( n + 1 )
3
u ( n + 1 )
f n + 1
ext
f n
ext

R
1

R
2

K u
1
( )

K u n
( )
f
in
t
_______________________________________________________________________________________________
_______________________________________________________________________________________________

BFGS Method

Also called the quasi-newton method (QN). In here an approximation to the
local tangent matrices are found. No factorization is required. The
approximation to the local tangent matrix was defined in the previous section.
The figure below depicts the algorethim.

Snap-through and Collapse

In many collapse problems the equilibrium path could look like the figure
below. In this type of behavior the BFGS method or in general the Newton
method has deficulties with the limit points. For the ascending branch of the
equilibrium curve, load control is greatly the most efficient method. As the
plateau in the response is reached (point A), then difficulties with
convergence are experienced (with the BFGS method the solutiion will jump
to point A). If displacement control is adopted then when point B is reached
again there will be convergence difficulties (solution could jump to point B).
To overcome this problem an algorithen is added to LS-DYNA that provide
the correct structural response. This algorithm is called the Arc-Length
Method.

_______________________________________________________________________________________________
_______________________________________________________________________________________________

Arc-Length Method

The arc-length method basic idea is that a constaint equation is added so
that load level is modified at each iteration rather thann holding it constant.
The method is graphicaly depicted in the figure below. Two arc-length
methods are available in LS-DYNA. These methods are called the Ramm
method (updated normal plane method) and the Crisfield method (spherical
path method)

_______________________________________________________________________________________________
_______________________________________________________________________________________________
The Ramm Method
In here as oppose to forcing the iteration process to follow a plane
perpendicular to the initial tangent for each load step, the iteration process is
constrained to be normal to the current arc. The method is graphicly depicted
in the figure below.

The Crisfield Method
In here the iteration process is constrained to follow a spherical path centred
at point o with radius equal to the arc length. The method is graphicly
depicted in the figure below.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Linear Equation Solver


Sparse Direct Linear Equation Solvers

LS-DYNA has 3 options for direct solution of the sparse systems of
linear equations.
All 3 options are based on the multifrontal algorithm.
Multifrontal is a member of the current generation of sparsity preserving
factorization algorithms that are also have very fast computational
rates.
That is multifrontal works with a sparsity preserving ordering to reduce
the overall size of the direct factorization.

Sparsity Preserving Orderings

In LS-DYNA 970 there are two ordering algorithms for preserving the
sparsity of the direct factorization.
The algorithms are Multiple Minimum Degree (MMD) and METIS.
MMD computes the ordering using locally based decisions and a
bottom-up approach.
It is inexpensive and very effective for small problems, that is problems
with fewer than 100,000 rows.
METIS computes the ordering from a top down approach. While
METIS usually takes more time than MMD to compute the ordering, the
METIS ordering reduces the work for the factorization enough to
recover the additional ordering cost.
METIS is especially effective for large problems, especially those that
are modeling three-dimensional solids.

The user can specify either algorithm using
*CONTROL_IMPLICIT_SOLVER.

Multifrontal Algorithm

The multifrontal algorithm factors a sparse matrix in a way that vastly
reduces the amount of work required to compute the factorization
_______________________________________________________________________________________________
_______________________________________________________________________________________________
compared to methods such as the frontal, profile, skyline, and variable
band.
These older methods counted on clustering the nonzero entries of the
factorization close to the diagonal to keep the size of the factorization
and the amount of work required to compute the factorization to a
minimum.

Three Solver Options

In Version 970 of LS-DYNA there are three direct linear solver options. From
Version 960 to 970 LSTC removed the two older (and less efficient) options.
The three options are

Solver
No.
Method
4 Real*4 implementation of multifrontal which includes automatic out-
of-memory capabilities as well as distributed memory parallelism.
Can use either MMD or METIS orderings. Default method.
5 Real*8 implementation of Solver No. 4.
6 Multifrontal solver from BCSLIB-EXT. Uses Real*8 arithmetic with
extensive capabilities for large problems and some Shared Memory
Parallelism.
Can use either MMD or METIS orderings. If the other solvers
cannot factor the problem in the allocated memory, try using this
solver.

Solver 6 on a single processor computer should be comparable to
Solver 5 but has more extensive capabilities for solving very large
problems with limited memory.
Solvers 4 and 5 should be used for distributed memory parallel
implementations of LSDYNA.
Solver 6 can be used in shared memory parallel.
In an installation of LSDYNA where both integer and real numbers are
stored in 8 byte quantities, then Solvers 4 and 5 are equivalent.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Sparse Eigensolver

LSDYNA now includes the Block Shift and Invert Lanczos eigensolver
from BCSLIB-EXT.
This eigensolver is used in LSDYNA to compute the normal modes and
mode shapes for the vibration analysis problem
The Lanczos algorithm iteratively computes a better and better
approximation to the extreme eigenvalues and the corresponding
eigenvectors.
BCSLIB-EXT uses a sophisticated logic to chose a sequence of shifts
to enable the computation of a large number of eigenvalues and
eigenvectors.
The implementation of BCSLIB-EXT in LSDYNA includes a shared
memory implementation. However only limited parallel speed-up is
available for most problems.
The wall clock time for the eigensolver is as much a function of the
speed of the I/O subsystem on the computer as the CPU time.
Parallelism can only speed up the CPU time and does nothing to
speed-up the I/O time.
The user can request how many and which eigenvalues to compute
using the keyword *CONTROL_IMPLICIT_EIGENVALUE.

Via the parameters on this keyword, the user can request any of the following
problems:

Compute the lowest 50 modes (that is nearest to zero)
Compute the 20 modes nearest to 30 Hz.
Compute the lowest 20 modes between 10 Hz and 50 Hz.
Compute all of the modes between 10 Hz and 50 Hz.
Compute all of the modes below 50 Hz.
Compute the 30 modes nearest to 30 Hz between 10 Hz and 50 Hz.

For example, in running a 240,000 node car body model, one need 1400 Mw
to extract 30 eigenvalues, but only 750 Mw to perform a linear analysis.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
During each nonlinear iteration, the linear system R u K

is solved.

Iterative Methods
(preconditioned) conjugate gradient
potentially low operation count
convergence difficult for some problems
promising future developments

x x
x
x x x x x x
x x x x x
x x x x
x x x x x x
x x x x x
x x x x
x x x
x x
x
x x x x x x
x x x x x
x x x x
x x x
x x
x
x x x
x x
x

Direct Methods
gaussian elimination
inexpensive backsolve (quasi-Newton)
The sparse direct solver looks promising for large linear systems
substantial storage savings
huge operation count reduction

Example: Beam Tower Model (Reverse Engineering, Ltd.)

beam elements
43,590 degrees of freedom

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Linear Equation Solver Data
skyline sparse
memory (Mw) 18 <6
max column height 720 114
avg column height 418 20
HP-735 CPU (sec) 1902 25

Example: Trimmed Inner Panel

Model Size
32,540 node points
31,640 shell elements
195,212 degrees of freedom

Linear Equation Solver Data
skyline sparse
memory (Mw) 176 <95
max column height 1320 54
avg column height 899 52
HP-735 CPU >5 day 88 min

Iterative Linear Equation Solver

LS-DYNA offers six iterative linear equation solvers
*CONTROL_IMPLICIT_LINEAR keyword [K]{x}={f}
LSOLVR = 10: best iterative solver (currently activates #16)
LSOLVR = 11: Conjugate Gradient method
LSOLVR = 12: CG with Jacobi preconditioning
LSOLVR = 13: CG with Incomplete Choleski preconditioning
_______________________________________________________________________________________________
_______________________________________________________________________________________________
LSOLVR = 14: Lanczos method
LSOLVR = 15: Lanczos with Jacobi preconditioning
LSOLVR = 16: Lanczos with Incomplete Choleski preconditioning

All iterative solvers use the sparse matrix storage scheme
eliminates all zero entries inside bandwidth
minimizes total storage requirement
Boeing Harwell format for portability

Iterative solvers outperform direct solvers in some problems
solid, massive structures
narrow range of natural frequencies (good numerical condition
number)

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Example:

20x20x20 brick elements, 26,460 equations
Solver type DIRECT ITERATIVE

LSOLVR 0 (default) 16
Memory 120 Mb 9 Mb
CPU 400 sec 21 sec

Linear Equation Solver Memory and CPU

An optional print-out is available to show memory and CPU requirements of
the linear equation solver: [K]{x}={f}
activated by keyword input *CONTROL_IMPLICIT_SOLVER: lprint
activated interactively <ctrl-c> lprint
memory estimate printed before attempting to allocate
change memory using command line option
ls-dyna3d i=inputfile.k memory=8000000

units are words
1 word stores 1 floating point number
single precision = 4 bytes per word
double precision = 8 bytes per word

F
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Linear Equation Solver Memory and CPU

==========================================================
time = 1.55684E+02
current step size = 7.78421E+01
SPARSE LINEAR EQUATION SOLVER STORAGE data (Mwords)
( 225972 degrees of freedom)
pointer arrays: initial = 11.523
actual = 6.413
stiffness coefficients = 6.187
Factorization Workspace (estimated)
symbolic = 14.015
numeric = 18.335
Final Storage Requirements (10% for pivoting)
incore out-of-core
symbolic factorization = 5.276 5.276
numeric factorization = 69.772 5.292
numeric solution = 65.561 3.145
TOTAL = 87.648 23.168
TOTAL available = 98.196 98.196
an INCORE solution will be performed
Initialization CPU = 7.220E+00 seconds
Symbolic Factorization CPU = 1.065E+01 seconds
Numeric Factorization CPU = 8.539E+02 seconds
Forward/Backward CPU = 5.060E+00 seconds


============================================================
time = 1.55684E+02
current step size = 7.78421E+01

SPARSE LINEAR EQUATION SOLVER STORAGE data (Mwords)
( 225972 degrees of freedom)
pointer arrays: initial = 11.523
actual = 6.413
stiffness coefficients = 6.187
Factorization Workspace (estimated)
symbolic = 14.015
numeric = 18.335
Final Storage Requirements (10% for pivoting)
incore out-of-core
symbolic factorization = 5.276 5.276
numeric factorization = 69.772 5.292
numeric solution = 65.561 3.145
TOTAL = 87.648 23.168
TOTAL available = 98.196 98.196

an INCORE solution will be performed

Initialization CPU = 7.220E+00 seconds
Symbolic Factorization CPU = 1.065E+01 seconds
Numeric Factorization CPU = 8.539E+02 seconds
Forward/Backward CPU = 5.060E+00 seconds

memory required
number of equations
solution method
CPU timings
begin factorization here
memory available
1 Mword =4 Mbytes (single)
1 Mword =8 Mbytes (double)
_______________________________________________________________________________________________
_______________________________________________________________________________________________
REMARKS

In nonlinear problems equilibrium iterations must always be permitted.
Reformation of the stiffness matrix at the beginning of every time step
often speeds convergence of the equilibrium search. Stiffness matrix
reformation normally occurs once at the beginning of each step, and
within a time step if either the equilibrium iteration limit is reached or the
solution diverges.
To suppress the stiffness reformation at the beginning of each step, set
the parameter ISTIF, in *CONTROL_IMPLICIT_SOLUTION, to a value,
which exceeds the total number of steps in the problem. This is often
economical for contact dominated problems, where the solution diverges
immediately after the first equilibrium iteration, causing a (second)
stiffness reformation.
To suppress equilibrium iterations during each step, set the parameter
NSOLVR in *CONTROL_IMPLICIT_SOLUTION to a one. A single linear
solution will be performed each step. This solution may not represent an
equilibrium state, hence results obtained may be inaccurate.
In problems dominated by contact it may be cost effective to limit the
BFGS iterations to five or less between stiffness reformations. This allows
LSDYNA/IMPLICIT to better account for rapidly changing contact areas.
Enter a value of one for ILIMIT in *CONTROL_IMPLICIT_SOLUTION to
obtain Full Newton Method with Line Search.
The convergence strategy for DNORM=2, which is the defult, in
*CONTROL_IMPLICIT_SOLUTION compares the current displacement
increment to the total displacement over the analysis. This criterion
becomes less strict as the analysis proceeds. By setting DNORM=1 the
strategy considers the total displacement over the step, hence is more
consistent as the deformation evolves.
In problems where there is much rigid body motion the displacement
tolerance may be insufficient, and it may be advisable, in some problems,
to tighten the energy tolerance to 10
-3
.
The arc length solver is intended for use with buckling problems where the
load-displacement curve is not monotonic, a problem class where
traditional displacement or energy norm driven method fail. In this method,
the solution is advanced in increments of constant arc length on the load-
displacement curve, regardless of its path. This may require load
reduction or unloading as the problem passes a bifurcation point.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
The user may define a single node whose load-deflection history controls
the solution. If no node is defined, norms of the global force and
displacement vectors are used in the generalized arc length method.

IGS
Geometric nonlinearity is introduced when displacements become large
(example: tip deflection of a cantilever beam is greater than the beam
thickness), or rotation become large (example: beam rotation theta is
large enough that the approximation sin(theta)=theta is not reasonable).
A second geometric nonlinearity is due to finite rigid body rotation. Both of
these nonlinearity are always accounted for in LSDYNA-Implicit.
The flag IGS is an option to include an additional term in the stiffness
matrix which reflects a change in apparent stiffness due to rotation of
stress. This flag does not change the degree of nonlinearity included in
calculation of internal or external forces. IGS does not alter the equilibrium
solution. However, the path taken during the iteration process to reach the
final solution depends on the stiffness matrix. In some cases, adding this
term will accelerate convergence and in other cases this term will create a
singular stiffness matrix and halt the iteration process. For instance
including this term will speed convergence for transversely loaded plates
with clamped boundaries, where transverse deflection generates sharp
increase in membrane stress. It is required by hyperelastic materials.

Two cases where you can benefit from the geometric stiffness are:

1) cantilevered beam type problems where either both ends are clamped
(lots of membrane force), or where the tip deflection is very large so that the
slope of the tip becomes say 30 degrees or more. Here the geometric
stiffness will improve convergence rate of the nonlinear iteration process.

2) Modal analysis where you wish to see the effect of stress on the frequency
response of the structure. This is the guitar string problem.

CNSTN
The consistent tangent flag CNSTN applies only to material type 3 for 2D and
3D solid elements, and to material type 115 for 3D solid elements. In these
materials, an option is available to use an alternative procedure for
computing the material tangent stiffness matrix. In general, the "consistent"
_______________________________________________________________________________________________
_______________________________________________________________________________________________
tangent stiffness will be softer than the default stiffness. This may in some
cases accelerate convergence of the nonlinear equilibrium iterations, but in
other cases may lead to divergence. Smaller load steps may be necessary
to realize a benefit from the consistent tangent stiffness. Since the net benefit
of this feature is unclear, it is not active by default.

LSOLVR
The LSOLVR "linear solver" provides the solution to the linear system of
algebraic equations Kx=f. There are several methods to compute this
expensive solution. LSOLVR selects between them.

What is the best and what situations for both direct and iterative.

Several options are available using parameter LSOLVR on
*CONTROL_IMPLICIT_SOLVER. For most simulations, the default
solver should be used. When using a double precision executable
(recommended), the single and double precison solver options are
equivalent.
The BCS-EXT solver (LSOLVR=6) is superior for very large models
which must run out-of-core. (Activate the linear solver print flag
LPRINT=1 or interactively type "<ctrl-c> lprint" to toggle printing of
memory information to the screen, which will indicate memory
requirements for in-core and out-of-core execution.)
The BCS-EXT solver is more versatile in exporting blocks of the
stiffness matrix to disk files, and can run with the least core memory of
any of the direct solvers. This solver also utilizes a more complex
pivoting strategy, producing superior solutions in cases where negative
eigenvalues nearly exist.
The iterative solvers (LSOLVR=10 to 16) are generally poor for
automotive applications. However, these solvers can dramatically
decrease memory and CPU requirements for large brick element
models, such
as those found in civil engineering soil models. If your model is largely
comprised of brick elements, it is probably worthwhile to test an
iterative solver. LSOLVR=10 activates what we feel to be currently
the best of the iterative solver options. It is presently set to activate
iterative solver 16, but this may change as improved iterative solvers
are added.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
ORDER.
Matrix reordering occurs during the symbolic factorization phase of the
linear equation solution procedure. During reordering, matrix rows and
columns are shifted to minimize the number of non-zero entries in the
inverted stiffness matrix. This is similar to the bandwidth minimization
operation which occurred with the old skyline solvers. Because of
reordering, the node numbering scheme chosen by the user is
arbitrary.
Two reordering options are available, selected with the ORDER
parameter on *CONTROL_IMPLICIT_SOLVER. The Multiple Minimum
Degree (MMD) method (ORDER=1) is the least expensive method, but
may not produce an optimal reordering, leading to more cost during the
numeric factorization (matrix inversion) phase.
The METIS reordering option (ORDER=2) is the most expensive
method, and usually produces the best reordering and fastest numeric
factorization phase. Presently, MMD is used by default for small
models, and METIS is used for large models.
The obvious goal in selecting a reordering method is to minimize the
combined cost of the symbolic plus numeric factorization phases of the
solve. To judge these costs, use LPRINT=1 or type "<ctrl-c> lprint" to
toggle printing of linear solver information to the screen. Pay particular
attention to the CPU costs:

Symbolic Factorization CPU = ...

Numeric Factorization CPU = ...

In general, the cost of symbolic factorization should never exceed 10%
of the numeric factorization cost, and is usually 1% or less. If the
symbolic cost is large, try activating the MMD reordering, Beware that
the numeric factorization cost will probably grow, so the net gain using
MMD may not be positive.

NSOLVR
In contrast, the nonlinear solver controls the iteration process as we seek
to find the displacements which balance internal and external forces.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
There are several nonlinear solvers to choose from, many of which are
based on Newton iteration. The parameter NSOLVR is used to make this
choice.
The simplest nonlinear solver option is "NSOLVR=1", which is to skip the
equilibrium iterations and accept the linear approximation to the solutions.
This choice is commonly referred to as a "linear" solution (not to be
confused with the linear equation solver LSOLVR above).

Screen Messages and Interactive Commands
It helps to have a quick look at the progress of a run. An input parameters
is added to activate printing this information to the screen. The default
values prevent printing. These two parameters are:
NLPRT on *CONTROL_IMPLICIT_SOLUTION and
LPRINT on *CONTROL_IMPLICIT_SOLVER.
We recommend *****ALWAYS***** activating NLPRT, and also
recommend activating LPRINT the first time you run a large problem to
get a sense of how much memory and cpu the stiffness matrix
factorization will require.
Interactive switches are added to activate each type of printing. The
switches have the same name as the input parameters:
Typing "<ctrl-c> nlprt" activates the nonlinear print flag.
Typing "<ctrl-c> lprint" activates the linear print flag.

Typing the switch a second time deactivates the flag, so you can turn them
on and off at will. One caution is that you must enter these interactive
commands during the nonlinear iteration phase of a time step. If you happen
to issue the command during another solution phase, say while the plot file is
being printed, then the command will be ignored.

IMFLAG=6
This is used for explicit and implicit analysis with intermittent eigen-value
extraction. Dynamic explicit analysis can produce stress oscillations and as a
results intermittent eigen-value extraction may lead to miss-leading results
because of the stress oscillations.

There are three methods which can be used to account for stress effects in
eigenvalue analysis. Shell type 6 may be used, but is not required.
_______________________________________________________________________________________________
_______________________________________________________________________________________________

1) Perform a standard eigenvalue analysis:
Use IMFLAG=1 and DT0=1.0 on *CONTROL_IMPLICIT_GENERAL.
Enter the number of eigenvalues to be computed using NEIGV on
*CONTROL_IMPLICIT_EIGENVALUE.
Introduce stress using the *INITIAL_STRESS keywords (probably
computed in a previous simulation which used
*INTERFACE_SPRINGBACK_LSDYNA to output a "dynain" file).
Activate the geometric stiffness terms using IGS=1 on
*CONTROL_IMPLICIT_GENERAL.

2) Perform an implicit intermittent eigenvalue analysis.
In this method, eigenvalues will be extracted at one or more times
during an implicit analysis.
Set up an implicit analysis to apply load to the model.
Enter a negative integer value, NEIGV=(-n) on *CONTROL IMPLICIT
EIGENVALUE. This indicates that curve ID=n will give NEIGV as a
function of time.
Define curve ID=n. Define a point at each time when eigenvalues will
be computed. For each curve point, the curve value gives the number
of eigenvalues to extract at that time.
The geometric stiffness terms will be automatically activated during
eigenvalue calculations. Parameter IGS on
*CONTROL_IMPLICIT_GENERAL can be selected as needed to give
the best convergence behavior during the implicit loading analysis.

3) Perform an explicit intermittent eigenvalue analysis.
In this method, eigenvalues will be extracted at one or more times
during an explicit analysis.
Beware that stress oscillations which normally occur during explicit
dynamic analysis can introduce geometric stiffness effects which may
appear counter-intuitive.
Set up an explicit analysis to apply load to the model.
Select IMFLAG=6 on *CONTROL_IMPLICIT_GENERAL to activate
explicit intermittent eigenvalue extraction.
Enter a negative integer value, NEIGV=(-n) on
*CONTROL_IMPLICIT_EIGENVALUE. This indicates that curve ID=n
will give NEIGV as a function of time.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Define curve ID=n. Define a point at each time when eigenvalues will
be computed. For each curve point, the curve value gives the number
of eigenvalues to extract at that time.

An eigenvalue analysis is a linear analysis, so double precision must be used
for accuracy. Linear also means infinitesimal deformation, so contact gaps
which are "open" stay "open", and gaps which are "closed" stay "closed".
With our penalty type contact, we introduce stiff penalty springs between
parts wherever they contact. So, in an eigenvalue analysis, stiff springs are
introduced wherever contact interface gaps are "closed". This effectively
bonds or ties the parts together. No springs are introduced in areas where
interface gaps are "open", so in these areas the parts are completely free.
So, you could say that the contact was "ignored" in areas where gaps are
"open".

FINITE ELEMENT MODELING TECHNOLOGY
_______________________________________________________________________________________________
_______________________________________________________________________________________________

FINITE ELEMENT MODELING TECHNIQUES

1. Engineering a FEA Model
2. Element Selection
3. Mesh Density
4. Symmetry
5. Modeling for Physical Phenomenon
6. Ad Hoc Guidelines
7. How to Tell If Your Results Are Correct

1. ENGINEERING A FFA MODEL

PRIMARY CONSIDERATIONS

Design Situation
Physical Nature of Design Concerns
Information Required
Most Effective Level of Simulation Detail

DESIGN SITUATION
New Design or Existing Problem

PHYSICAL NATURE OF DESIGN CONCERN (S)
Does previous product knowledge exist that can be used to pinpoint high
risk area
What is the nature of the predominant physical phenomena that must be
simulated
What assumption can be made and what are their effects on simulation
results

INFORMATION REQUIRED
Full scale tests, material characterization, geometry, etc.

MOST EFFECTIVE LEVEL OF SIMULATION DETAIL
Purpose of Analysis/Evaluation Focus
Requirements for time Responsiveness
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Cost/Benefit Trade off
Confidence Level of Model and Loading Assumptions
Accuracy Level of Acceptability Criteria

_______________________________________________________________________________________________
_______________________________________________________________________________________________

2. ELEMENT SELECTION

A General-purpose finite element code can be thought of as a toolkit of
structural entities from which a physical and mathematical description of a
structure can be built.
In order to model a structure efficiently and accurately, it is important that
the correct types of element be used and that the mesh of elements
constructed be dense enough to accurately represent the strain
distribution in the structure.

CHOOSING ELEMENT TYPES FOR MODELING A STRUCTURE

1-D rod like
structure
2-D structure
3-D shell
structure
3-D solid structure

STEPS IN DEFINING A MODEL:

1. Select the appropriate element type:
Most economic element, with least DOF per node which will give the
desired results.
2. Determine the mesh density required based on:
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Overall stress/ strain gradients expected (which depends on loading to be
applied )
Location of areas of stress concentration (fillets, holes, etc.)
3. Tradeoffs:
Economic element and coarse meshes are easy to generate and run,
however, they do not always give the results desired.
4. Rule of thumb:
The time to perform a finite element analysis (preprocessing, solution,
postprocessing) is roughly proportional to the number of nodes in the model
squared, TimeN
2

*ELEMENT_DISCRETE

element ID, part ID
2 nodal point IDs, orientation, scale factor on force
force vs. displacement is defined by *MAT_SPRING
force vs. velocity is defined by *MAT_DAMPER

*MAT_SPRING_option *MAT_DAMPER_option
elastic viscous
elastic-plastic (E, Et, SGMY) nonlinear viscous
nonlinear elastic
inelastic (allow tension or compression only)
general nonlinear (negative or zero slope allowed in the F vs. D curve)
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Maxwell (exponential decay of stiffness)

*SECTION_DISCRETE
transnational or rotational
dynamic magnification factor
tension/compression limits

*ELEMENT_MASS
element ID, node ID, mass value

*ELEMENT_BEAM
element ID
part ID
3 nodal point IDs

*SECTION_BEAM
Beam (axial, bending, torsional deformations)
Hughes-Liu (default)
CROSS SECTION
Rectangular, tubular, arbitrary
Area and inertias

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Hughes-Liu: type 1 (DEFAULT), recommended
l 6 DOF per node: (dx, dy, dz, rx, ry, rz)
l one integration point along length
l integrated cross section (not "resultant" beam)
l transverse shear correction factor
l moveable mid-plane option
l s, t directions defined by
orientation node n3

*ELEMENT_SOLID
element ID
part ID
4-noded tetrahedron
6-noded wedge
8-noded brick

*SECTION_SOLID
FORMULATION
8-noded reduce integration (default)
8-noded fully integrated (selective reduce integration)

r
s
t
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Constant Stress Solid: type 1 (DEFAULT), not recommended
l 3 DOF per node: (dx, dy, dz)
l single point integration with hourglass control
l requires hourglass control type 6
l least expensive

Fully Integrated S/R Solid: type 2, recommended
l 3 DOF per node: (dx, dy, dz)
l 2x2x2 selective/reduced integration (B-bar method)
l most expensive

*ELEMENT_SHELL
element ID
_______________________________________________________________________________________________
_______________________________________________________________________________________________
part ID
4 nodes for quads and 3 nodes for tris
thickness can be defined in element card

*SECTION_SHELL
FORMULATION
Hughes-Liu
Beltschko-Tsay (default)
BCIZ triangular
C-zero triangular
S/R Hughes-Liu
Fast full integration
linear

You can define material orientation for orthotropic and composite materials.

Shell Element Technology

4-Noded 3-Noded
1. H.L. 1. BCIZ
2. B.T. 2. C-zero
3. B.T. Membrane
4. S/R H.L.
5. S/R H.L. (co-rotational)
6. B.L.
7. B.T. Membrane full integration
8. B. Wang Chiang
9. H.L. (co-rotational)
10. Fast fully integrated shell
11. linear

3-Noded Shell Elements

1. BCIZ (due to Zienkiewicz and co-workers 1965)
First triangular element implemented in DYNA3D
Used for compatibility with earlier user models
Co-rotational formulation
Non-conforming shape functions
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Kirchhoff theory
Local nodal rotations are interpolated by flexural shape functions similar to
B. beam element
Three point in-plane integration
No hourglassing
More expensive than C-zero

2. C-zero (Kennedy and B. 1986) (C-zero means constant strain
element)
Used for compatibility and mesh transition
Co-rotational formulation
Mindlin theory
Interpolations functions are the area coordinate
One point gauss

_______________________________________________________________________________________________
_______________________________________________________________________________________________
4-Noded Shell Elements

Belytschko-Tsay: type 2 (DEFAULT), not recommended
6 DOF per node: (dx, dy, dz, rx, ry, rz)
single point integration with hourglass control
requires stiffness form of hourglass control (type 4 or 5)
available w/ or w/o warping stiffness (BWC flag on *CONTROL_SHELL)
option to automatically switch formulation during seamless springback to
fully integrated element #6
degenerates to triangle
least expensive

S/R Hughes-Liu: type 6, recommended
2x2 selective-reduced integration
6 DOF per node: (dy, dz, rx, ry, rz)
default element from LS-NIKE3D should be used in comparisons
use "unique nodal normals" on *CONTROL_SHELL
includes optional geometric stiffness terms
can improve convergence if large membrane stresses present
required for eigenvalue problems with initial stress
most expensive

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Belytschko-Wong-Chiang: type 10, not recommended
single point integration with hourglass control
requires stiffness form of hourglass control (type 4 or 5)
improved stiffness compared to Belytschko-Tsay type 2
inexperienced
inexpensive

Fast Shell: type 16, recommended
2x2 integration with enhanced strain formulation
Bathe-Dvorkin transverse shear correction eliminates W-mode
hourglassing
hourglass type 8 adds warping stiffness, may improve convergence
least expensive of 2x2 elements: 2x-3x more than Belytschko-Tsay (not
4x)

best for springback

*ELEMENT_TSHELL
element ID
part ID
8-noded

*SECTION_TSHELL
one point integration or
selective reduce integration
Similar to solid element with enhancement based on H.L. and B.T. shell.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
To obtain shell like behavior it is necessary to have several integration
point through the thickness while employing plan stress constitutive
equations.
Uses reference coordinate system similar to B.T. shell (co-rotational).
To ensure plan stress condition the normal stress is updated as a penalty
condition to maintain the shell thickness.

Implicit 2D Solid Elements

Plane Strain: type 13 , Axisymmetric: type 15
l special case of 4-node shell element
l construct in X-Y plane
l Y-axis is axis of rotation for axisymmetric type 15
l 2 DOF per node (dx,dy)
l MUST constrain remaining DOF (dz, rx, ry, rz)
l MUST select NIP=4
l very fast, good convergence behavior

X
Y
_______________________________________________________________________________________________
_______________________________________________________________________________________________
GUIDELINES FOR MODELING WITH BEAMS

Element is a slender rod like structural component.
Element geometry is described by its length and cross-sectional
properties (area, moments of inertia, etc.)
Orientation of the cross section principal axes with respect to the global
(X, Y, Z) coordinate directions.
Suspension system components in vehicle crashworthiness simulations.
Explicit modeling of skeletal frames structures such as building steel
frames in Civil Engineering design.
As connector element in complex mechanical structures to simulate

Bolts or Rivets
Eccentric Connections
Stiffening Ribs in Shell Structures

GUIDELINES FOR MODELING WITH 2-D PLANE ELEMENTS

In many situations, the analysis can be simplified to a 2D situation. The types
of 2D continuum elements, which are generally provided, are:

Plane Strain/Stress Elements
Axisymetric Elements

GUIDELINES FOR MODELING WITH SHELLS

Thickness to span ratio, flat shell:
h<< a, h << b, generally
b
h
and
a
h
< 0.1
A shell resists lateral and in-plane loads by bending, transverse shear
stresses and in-plane action, not through block-like compression or tension in
thickness direction.

Stress in the Z, or thickness, direction is zero.
Element thickness and loads can vary from node to node.
Thickness to span rations must be considered at transition joints.

_______________________________________________________________________________________________
_______________________________________________________________________________________________

GUIDELINES FOR MODELING WITH THICK SHELLS

Based on shell theory like a thin shell, but is like a 3D element and has 6
DOF per node (plane stress constitutive equations).
Provides a convenient way of transitioning from shell like to solid-like
structures.
Not all general-purpose FE programs provide this type of element.
The element is in general too stiff and in many cases can be replaced by
solid element.
The types of elements available are topologically similar to solid elements.

GUIDELINES FOR MODELING WITH SOLID ELEMENT

Solid elements are the most versatile type of element available in general
finite element program.
Virtually any type of structure can be modeled with this type of element;
however its use is primarily for:

Fully 3 dimensional structures for which a simplification to one of the
previously described situations not judged to be valid due to:

Complex 3D geometry.
Stress fields, which are fully 3D.
Loadings which are not symmetric.

Detailed analysis of stresses in critical areas such as:

Faster locations
Welds
Fillets

_______________________________________________________________________________________________
_______________________________________________________________________________________________
3. MESH DENSITY

In this section we will discuss some of the factors which influence how the
user of finite element analysis can estimate how fine a mesh is required to
yield accurate results.
Obviously, since the finite element technique involves discretization of some
type of physical continuum into a gridwork (i.e. mesh), the way in which the
users chooses to do this influences the outcome of the analysis to some
extent. Furthermore, there are basically an infinite variety of valid ways of
discretizing a given physical phenomenon and arriving at approximately the
right answer.

GUIDELINES FOR MESHING

Element size must be small enough that it adequately matches the shape
of the structure boundaries.
Element internal strain distribution must approximate the actual strain field
in the physical component.
The model must be constructed so that stresses are sampled at or near
where expected concentration of stress actually exists in the model.
For example, if we construct a cantilever beam from 8-noded brick solid
elements, and apply a uniform tensile load to it, we obtain:

Fixed
end
P

Cantilever Beam Model with a pure Tensile Load

We would find that the predicted stress and strain are exactly the correct
values.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
But if we apply a lateral shear loading to this model we would find that

Fixed
end
P

We would find that the predicted MAXIMUM stress and strain values are
not the maximum values in this structure.
Why?
The answer is simply that the error is in the modeling of the structure.
A liner hexahedral solid can in fact predict the linear strain distribution.
And as a result, this model dose in fact predicts the CORRECT
MAXIMUM DISPLACEMENTS for the tensile load.
As we know from simple beam theory the maximum bending stress is
actually at c= b/2 and is thus:
I
Mb
2
max

But for stresses and strains:
The optimal stress recovery point for a linear solid element is the
centroid of the element.
In this model, the stress recovery point lie a long the centroidal axis of
the beam structure, thus, the maximum stress at this point is:

I
MC
max
And c = 0, thus, stress is zero

GUIDELINES:

The following general guidelines can be stated relative to determining an
appropriate mesh density:

_______________________________________________________________________________________________
_______________________________________________________________________________________________
The mesh density should be based upon the degree of accuracy required
from the analysis.
If only displacements or mode shapes are required, it is not necessary to
mesh finely around areas of very localized stress concentration.
When it is desired to model local features in detail, the size of the
elements around that detail should be an order of magnitude less than
that of the detail itself. Must have smooth mesh size transition.
Spacing of nodes should be closer near the detail. Some studies based on
analytical solutions of stress concentrations around holes and fillets
suggest that a logarithmic spacing is a good guideline.
If the structure is very complex, a two step procedure is often
appropriate:
Make a reasonable overall model of the structure with a fairly coarse
grid.
Then make a detail model of the region in question and apply
displacements from the overall model to the detail model as boundary
conditions.

4. SYMMETRY

TYPES OF SYMMETRY
1. Symmetry in the Geometry
2. Symmetry in Loading

The types of symmetry, which can be (separately or in combination), are:

Rotational Symmetry
The structure is symmetric about some axis of revolution (here we say
the structure is axisymmetric).
Reflective Symmetry
The structure has one or more planes about which a mirror image
exists.
Antisymmetry
The structure itself is symmetric but the loading or restraints are
reversed on opposite sides of the plane (s) of symmetry.
Cyclic Symmetry
Simple cyclic symmetry, in which the smallest repetitive piece of the
structure can be modeled, loading is symmetric.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Dihedral cyclic symmetry in which a repetitive segment exists which
is in itself reflectively symmetric about its own midsection, loading is
not symmetric.
Rotation cyclic symmetry, in which a repetitive section exists when
rotated about an axis of revolution, loading is not symmetric.

Example: A turbine disk

Boundary Conditions Which Must be Used With Reflective Symmetry:

When symmetry and antisymmetry are used, the proper boundary conditions
(restrained degrees of freedom) must be specified at nodes which lie in the
plane of symmetry.

Boundary conditions for symmetry are:
Translational displacements NORMAL to the plane of symmetry are
zero. Rotation IN the plane of symmetry are zero.
Boundary conditions for antisymmetry are:
Translational displacement IN the plane of antisymmetry are zero.
Rotations NORMAL to the plane of symmetry is zero.

SIMPLE CYCLIC SYMMETRY
The smallest repetitive piece of the structure can be modeled; loading is
symmetric.

Example: A thin disk with holes, the section shown is modeled with
symmetry boundary condition applied in nodal local coordinate systems to
the edges.

ROTATIONAL CYCLIC SYMMETRY
_______________________________________________________________________________________________
_______________________________________________________________________________________________
This is the most general case. Cyclic symmetry in which a repetitive
section exists when rotated about an axis of revolution; loading is not
symmetric but is periodic (sinusoidal) with K periods.

DIHEDRAL CYCLIC SYMMETRY
Cyclic symmetry in which a repetitive segment exists which is in itself
reflectively symmetric about its own midsection; loading is symmetric.
There are two mathematical models, called the left model and right
model although the user may only need model one if the program has a
built-in cyclic symmetry capability.

WARNING ABOUT SYMMETRY AND DYNAMICS PROBLEMS!

If symmetry is used in dynamics, remember that due to the boundary
conditions ONLY the symmetric modes will be found if symmetry
boundary conditions are imposed.
If anti symmetry is used in dynamics, remember that due to the boundary
conditions only the antisymetric mode will be found if antisymetric
boundary conditions are imposed.
For rotational cyclic symmetry, only the modes involving the K th nodal
diameter will be found when the cyclic symmetry constraints are applied.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
LS-DYNA INITIAL AND BOUNDARY CONDITIONS

*SET_NODES_option
Define a set of nodes by a set identification number (SID)
Define the nodes to be included (NIDS)

*SET_NODE_LIST
Define the nodes, 8 per line

*SET_NODE_COLUMN
Define the nodes, 1 per line

*DEFINE_BOX
Define a boxshaped volume, everything inside the box can be used as
input. Assign a box identification number (BOXID)
Define two extreme corners of the box
Xmin Xmax
Ymin Ymax
Zmin Zmax

*BOUNDARY_PRESCRIBED_MOTION
Impose nodal motion on a node, set of node, or a rigid body
Applicable to one degree of freedom
Displacement, velocity, and acceleration (nodes only)
Motion prescribed by a load curve

*DEFINE_CURVE
Define a (load) curve to be used as input to other options
Assign a load curve identification number (LCID)
Define the point of the curve (Abscissa x and Ordinate y)
Examples: force vs. time, velocity vs. time, force vs. displacement, etc.

*DEFINE_COORDINATE_option
Used as input to other options
Assign a coordinate system identification number (CID)
Define using:
NODES 3nodes: local origin, along local x-axis, in local x-y plane
SYSTEM x,y,z of three points (same as nodes)
VECTOR 2 vectors: local x-axis, local in plane vector
_______________________________________________________________________________________________
_______________________________________________________________________________________________

*LOAD_option
Define various loads (forces, moments, pressures, body, heat, etc.)
Avoid single concentrated loads sine it is physically unrealistic and
promote hourglass modes
Requires a load curve
Load can be scaled

*LOAD_BEAM
Distributed traction load along any axis of a beam or set of beams
Load is given in force per unit length

*LOAD_BODY
Body force load due to a prescribed base acceleration, angular velocity
Any degree of freedom: X,Y, Z, RX, RY, RZ
All nodes or a subset of nodes

*LOAD_NODE and *LOAD_RIGID_BODY
Apply a load to a node, or a rigid body
X, y, or z force, x, y, or z moment
Follower force
Global or local coordinate system

DISTRINUTED PRESSURE
Apply a distributed pressure to a:
Segment (*LOAD_SEGMENT)
Set of segments (*LOAD_SEGMENT_SET)
Shell (*LOAD_SHELL_ELEMENT)
Set of shells (*LOAD_SHELL_SET)
Positive pressure acts in the negative normal direction of the
shell/segment

_______________________________________________________________________________________________
_______________________________________________________________________________________________

5. MODELING FOR PHYISCAL PHENOMENON

(some examples)

INTRODUCTION
Some situations require special modeling techniques or considerations.
In this portion of the course, we will examine some of these and explain how
to use proper modeling techniques to insure correct results.

MIXING DIFFERENT TYPES OF ELEMENTS
In some physical situations, it is desirable to be able to mix different types
of elements in the same model.
When element types are mixed, it is important that degrees of freedom at
the point that dissimilar element types are joined be properly matched so
that displacement compatibility and proper load transfer paths are
maintained.
As an example, suppose we wish to model a situation where a support
strut is welded to a thin plate structure. In this case, the easiest way to do
this would be to represent the strut by a beam and the plate by shells:

Physical modeling
3-D beam
shell

In this case both the space ( 3D ) beam, and thin shell elements have six
degrees of freedom at each node:
X, Y and Z Transnational Displacement DOFs
X, Y and Z Rotational DOFs.
These two elements are thus entirely compatible
Now suppose we have a similar situation, but this time support strut is to
be welded to a massive, thick plate structure:
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Physical modeling
3-D beam
shell

In this case 3D beam element has six degrees of freedom at each node:
X, Y and Z Transnational Displacement DOFs
X, Y and Z Rotational DOFs

But the solid elements have only the 3 translational DOFs.

These two elements are not entirely compatible, and in fact, modeling this
way will cause a ball joint to exist at the end of the beam attached to the
solid since the solid will not transmit moments to the end of the beam and
vice versa.

To solve this problem, it is necessary to connect the beam end J to more
than one point on the solids structure to establish a moment lever by which
the bending is transmitted to the solids. This can be accomplished by either
the use of rigid connectors.

J

_______________________________________________________________________________________________
_______________________________________________________________________________________________

Or by using a multipoint constraint to describe the relationship of the
displacement, d, in X, Y and Z at the nodes J, K, L, M and N.

GEOMETRIC DISCONTINUITES
As another case where geometric discontinuity can cause problems,
consider the case of a branch in a thin shell structure.

I
K
J

Here the problems that can arise are in the interpretation of stress result
at the intersection. Most finite element programs produce contour plots by
extrapolation of the element interior (i.e. integration point) data to the
nodes and AVERAGING that data for all elements connected to given
node. It is the nodal data that is then contoured.
This is acceptable for planar or 3D solid element types, but is
unacceptable for shells, since: Shell element generally has its own
coordinate system.
In general, and especially when the shells have an intersection angle
greater than say 5 degrees, the like components of stress are in
significantly different directions.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Shell 2
Shell 1
Shell 3
Node 1
x3
x2
x1

Membrane stress components, local shell x direction
3 x

1 x
and
2 x

As can be seen from the above, averaging the X, Y and in-plane shear
components without special consideration would be in error; however,
scalar quantities such as vonmises or maximum principal stresses could
be averaged if they are computed from components averaged in a
common coordinate system.
It is possible to make the program average the stress properly by placing
two sets of nodes at the joining line and constraining the co-located joints
to have the same displacements. This method will give correct nodal
averaged values since the elements on one branch will not have results
averaged with those of the other branch.

MATRIAL DISCONTINUITIES

When two or more materials exist in a given model, the same problem can
occur in stress averaging at the nodes on the material interface.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
steel
brass
Incorrect stress
distribution using
nodal averaging
Actual stress
distribution

One solution is to place two sets of nodes at the same location at the
interface and couple them as was described for the shell problem
previously discussed.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
ECCENTERIC SHELL STEFFENERS

Beam elements are frequently used to represent stiffeners for plate/shell
structures. If beams are directly connected between shell mid-surface node
points then:

Beam cetroidal axis will be locked at the mid-surface of the shell.
This is an incorrect representation for eccentric stiffeners, as shown
above.
Structural behavior is significantly different when stiffener is eccentric with
respect to the shell mid-surface.

Eccentric shell stiffeners can be modeled by:

Modeling shells with nodes at mid-surface of physical shell surface.
Modeling stiffeners as beam elements with separate nodes located along
centroidal axis of the beam cross section.
Connecting corresponding shell and beam node pairs with rigid link
elements.

TIME VARYING LOADS

Frequently, the operating environment of structural components involves
its response to time varying loads (dynamic and impulsive).
The determination of acceptable displacement and stress response can
often be obtained with simple quasi-static analysis.
The choice of performing static or transient dynamic analysis depends on
the frequency characteristic of the applied time varying loads and the
responding structural components.
Rule of thumb: when the frequency of excitation is less than one third of
the lowest natural frequency of the structural then static analysis can be
performed.

_______________________________________________________________________________________________
_______________________________________________________________________________________________

6. AD HOC GUIDELINES

Finite Element mesh should be uniform where this is practical however,
nonuinformity is often necessary to achieve refinement of the gridwork in
regions of rapid changes in loading and /or behavior.
Analyses to obtain displacement (or force) behavior do not require as
refined a gridwork as is required to obtain strain (or stress) behavior.
Prediction of modes of vibration usually requires a more refined
idealization than dose the prediction of frequency of vibration.
The use of constant strain triangle finite element should be avoided
wherever possible.
A mesh planned for static analyses will be satisfactory for dynamic unless
high frequency response content is significant.
Determining that the solution is insensitive to spatial and temporal mesh
size provides an adequate check of modeling except in rare cases.
Rapid changes in element size should be minimized.
For complex structures consider a stage analysis. Analyze the entire
structure with a coarse mesh and then study the areas of interest with a
fine mesh, using as boundary the results of the coarse analysis.
Three elements through the wall thickness of thick-walled structures will
usually give reasonable result.
If the structure and loading posses symmetry, advantage should be taken
into account of this fact. However, one should be careful in buckling and
frequency analyses where asymmetric modes might be important.
When modeling reduced portions of the structure for evaluation of local
effects, make the boundary of this local area far enough away from the
area of interest to avoid boundary effects on the information at the area of
interest.
Aspect ratio limits vary from element to element; however, a general
guideline is to keep element aspect ratios below 10 for deformation
analysis and below 5 for stress analysis.
Element of extreme shape (high aspect ratio rectangles, large angled
triangles, etc.) are unfavorable with respect to minimizing error and should
be avoided.
Solutions for structures with uniform stiffness distribution are more
accurate than those for structure with stiff and flexible regions (e.g., portal
frame in which axial deformation of horizontal member is considered). The
same rule may be worded differently as: solutions for structures with a
_______________________________________________________________________________________________
_______________________________________________________________________________________________
large frequency range are less accurate than those for structures with a
small range.
For convergence studies, one must refine (i.e., subdivide) the original
mesh, otherwise a new approximating sequence is begun.
The recommended first checks on predicted behavior are equilibrium
checks using loads, reactions and forces for quasi-static and energy
balance check for impact problems.
Vibration analyses using lumped mass model usually yield slightly lower
frequency values than are obtained from consistent mass models.
LSDYNA uses lumped mass matrix.
Consistent mass model give frequency values higher than exact solution.
Analyses including nonlinearity in geometry or material usually requires a
more refined idealization than linear analyses.

SOLUTION ACCURACY

There are basically 3 ways to determine if sufficient accuracy is being
achieved:

Compare the results with textbook analytical solutions for similar stress
concentration, using the far field stress in the model and an appropriate
stress concentration factor from an analytical/empirical solution (normally
not available).
Compare the results with test data. This is the best way to gain confidence
in the modeling technique for a specific type of component. Using test
correlation, the user will have a high degree of confidence in future
modeling of similar types of components using the specific modeling
technique.
If neither of the above two methods is viable, a finer mesh of the structure
can be constructed and the results compared to the original mesh. This
process of refining can be repeated until the results converge to an
acceptable tolerance. The drawback to this method is obviously the
amount of repetitive modeling and computer run time, but it can be worth
while for complex components where no test prototype is readily
obtainable.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
7. HOW TO TELL IF YOUR RESULTS ARE CORRECT

One serious error that is commonly made in using finite element analysis
is a failure to adequately verify that the results from the analysis are
reasonable.
There are a number of things that users can do to verify that the model is
yielding reasonable results. Will discuss some of the things you can do to
examine the accuracy of the results.

USER MUST UNDERSTAND ENGINEERING PRINCIPLES

Interpretation of result requires understanding the basic fundamentals of
engineering disciplines.
While color graphics are useful for presenting results, they should not be
used as a substitute for examination of the numbers, which result from the
analysis.
DO NOT USE COLOR FILL CONTOUR PLOTS as means of verifying
correctness of results. While they are very useful for looking at trends and
for impressing people, this type of plot tends to mask some errors, since
the plot will not show localized trend. This is because smoothing the
result for display tends to hide localized anomalies.

WHAT TO LOOK FOR
The first thing to examine would be the FUNDAMENTAL result of the
analysis.
After the fundamental result is found to be acceptable, there are three
basic categories of data which must be examined to assess solution
validity:

Global Summations, Nodal Results, Elemental Results

GLOBAL SUMMATIONS
These give the user a good idea of how well equilibrium requirements
have been met, and are indicator of round-off error in the overall solution.
Types of global sums which can be used in structural analysis are:

GLOBAL ENERGY
GLOBAL FORCE SUM (Quasi-Static Structural Analysis):
_______________________________________________________________________________________________
_______________________________________________________________________________________________

This output provides an absolute check on whether or not the structure is in
equilibrium.

GLOBAL ENERGY SUM (Static or Dynamic Structural Analysis).

NODAL RESULTS
Displacements
Summation of Forces
Examine the sum of forces at free nodes to verify that they are zero
and at restrained nodes to verify that they are reasonable non-zero
values.

ELEMENT RESULTS
As a result, the elemental results are generally less accurate than the
fundamental nodal result because they are DERIVED results, not
FUNDAMENAL results
Assessing the accuracy of elemental results can be done by:
Looking at the variation of values around critical locations (numerically, by
X-Y graph plots or iso-value line contour plots). High variation, shown by
jagged graph or iso-value contour liner, indicates too coarse a mesh.
Qualitatively comparing the stress distribution with that predicted by
previous test or known analytical solutions for similar stress concentration
features (holes, fillets, etc.).

CONTACT AND FRICTION
_______________________________________________________________________________________________
_______________________________________________________________________________________________


CONTACT IMPACT ALGORITHMS

Only two methods currently available for LS-DYNA/IMPLICIT

1. Penalty method (general contact with sliding, separation, and friction).

A contact search algorithm is either nodal based or segment based.
The choice of master and slave surfaces is arbitrary when the symmetric
penalty treatment is employed, otherwise;
More coarsely meshed surface should be master
If there is a large difference in mass density then the side with lightest
material density should be master.

2. Kinematic constrained (tied contact)

In the current version of LSDYNA several contact interfaces are supported
1
:

*CONTACT_SURFACE_TO_SURFACE (3)
*CONTACT_NODES_TO_SURFACE (5)
*CONTACT_ONE_WAY_SURFACE_TO_SURFACE (10)
*CONTACT_FORMING_SURFACE_TO_SURFACE (m3)
*CONTACT_FORMING_NODES_TO_SURFACE (m5)
*CONTACT_FORMING_ONE_WAY_SURFACE_TO_SURFACE (m10)
*CONTACT_AUTOMATIC_SURFACE_TO_SURFACE (a3)
*CONTACT_AUTOMATIC_NODES_TO_SURFACE (a5)

1
WARNING: If one of the unsupported contacts is used in an implicit analysis, no error message is issued at the
beginning of the analysis. The simulation will either fail to converge, or if converged the results will be erroneous. In
fact, only supported contacts are added to the system stiffness matrix.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*CONTACT_AUTOMATIC_ONE_WAY_SURFACE_TO_SURFACE
(a10)
*CONTACT_AUTOMATIC_SINGLE_SURFACE (13)
*CONTACT_TIED_SURFACE_TO_SURFACE (2)
*CONTACT_TIED_NODES_TO_SURFACE (6)
*CONTACT_TIED_SHELL_EDGE_TO_SURFACE (7)
*CONTACT_TIED_SURFACE_TO_SURFACE_FAILURE
*CONTACT_TIED_SURFACE_TO_SURFACE_OFFSET (o2)
*CONTACT_TIED_NODES_TO_SURFACE_OFFSET (o6)
*CONTACT_TIED_SHELL_EDGE_TO_SURFACE_OFFSET (o7)
*CONTACT_TIED_SHELL_EDGE_TO_SURFACE_BEAM_OFFSET
*CONTACT_TIED_SURFACE_TO_SURFACE_CONSTRAINED_OFF
SET
*CONTACT_TIED_NODES_TO_SURFACE_CONSTRAINED_OFFSE
T
*CONTACT_TIED_SHELL_EDGE_TO_SURFACE_CONSTRAINED_O
FFSET
*CONTACT_SPOTWELD (s7)
*CONTACT_SPOTWELD_WITH_TORSION
*CONTACT_2D_AUTOMATIC_SURFACE_TO_SURFACE

User should be aware that, if an analysis involving explicit-implicit switching,
only implicit-supported contacts should be used.
In general, contact works in the same way as in the explicit version, with only
a few but important differences, which are listed in the following paragraphs.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Variable Setting for Implicit Analysis
When an implicit analysis is performed, some contact variables
automatically assume different values than the explicit default value:
Bucket sort frequency variable BSORT on the Optional Card A is set to
1 by default. Since the implicit time step is in general much bigger than
the explicit one, a bucket sort each time step becomes necessary.
Shooting node logic variable SNLOG on the Optional Card B is
disabled (SNLOG=1), since in implicit analyses, an excessive
penetration will result in non-convergence and the step will be
attempted again with a smaller time step. Thereby shooting node logic
in not necessary and it is disabled.

Improved Implicit Convergence for Implicit Analysis
Often contact is very problematic to obtain a convergent solution. This
is mainly due to the large time step used in implicit analyses, which ultimately
result in sudden large penetration, large out of-balance forces and thus non-
convergence. LS-DYNA adopts by default a stiffness-softened behavior as
showed in the following Figure, which cause the stiffness to gradually
increase, thus coupling the contacting nodes before contact takes place.
This means that before contact takes place, displacement in one body
creates forces on the other body. However, this means also that when the
two bodies attempt to separate, a force opposite to separation is developed.
This behavior is called sticky-contact and is the default method (IGAP=1).
Please note that AUTOMATIC CONTACT TYPES do not support this
method, a3, a5, a10, 13.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
penetration
stiffness
0
contact
force
penetration
0

The Figure shows the effect of the variable IGAP=1 on the contact stiffness
in a implicit analysis. This modified stiffness causes the sticky behavior
when parts attempt to separate.

If no sticking contact is desired, the user can switch to the traditional explicit
contact behavior. When using automatic contact, improved implicit
convergence behavior variable IGAP in the Optional Card C is automatically
set to no sticky contact (IGAP=2). This will result in a more difficult
convergence, since no improved-convergence behavior scheme will be used.
An indicator that sticking occurs is a fluctuating energy norm during
equilibrium interactions, while the displacement norm is well behaved.

Eigenvalue Analysis and Contact
All the contacts supported by the implicit version of LS-DYNA are
consequently supported for eigenvalue extraction. In an eigenvalue analysis,
deformations are infinitesimal. This means that the nodes that are initially in
contact stay in contact, while the nodes that are not are free to move.
Therefore, stiff springs are added only to those nodes that are in contact at
the beginning of the analysis.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Penalty Method:

Used in explicit and implicit codes
The method consists of placing normal interface springs between all
penetrating nodes and contact surfaces.
In implicit we add spring stiffness matrix to the global stiffness matrix.
In here a force modulus is computed for each slave and master segment
based on the thickness and bulk modulus of the element in which it
resides.
Results in less mesh hourglassing.
t is unaffected by the existence of the interface since the interface
stiffness is chosen to be approximately the same order of magnitude as
the stiffness of the interface elements.

SLIDING WITH SEPARATION AND FRICTION

Node-to-Surface
One part is referred as the master part and the other as the slave part
A node may belong to several interfaces
An interface node can be in rigid body
Breaks down with bad or disjointed mesh
Penetration is reduced by the penalty method
The algorithm have the following characteristics

Accuracy (very good)
Generality (not as good)
Efficiency (very good)
Compatibility (not as good)

The spatial distribution of forces is not smooth and may induce hourglass
deformation. This Interface simulates impacts between a master surface and
a list of slave nodes.

_______________________________________________________________________________________________
_______________________________________________________________________________________________

Contact surface can be used for:
Contact simulation of a complex fine mesh on a simply convex surface.

ALGORITHM

1. For each slave node
2. Find the closest master node
3. Find the closest master segment
4. Check if slave node has penetrated the master segment
5. Find the contact point
6. Compute the penetration
7. Apply forces to reduce the penetration

Normally when the algorithm fails is at step 2.

REDUCTION OF PENETRATION

Let:
N Slave Node
C Contact point
M
t
Element Corner Node
i
m Masses of Element Corners
_______________________________________________________________________________________________
_______________________________________________________________________________________________
K Spring Stiffness
Penetration

Penetration is reduced by the introduction of a massless spring of stiffness K
between node N and contact point C. A force K F is applied on node N in
direction NC.

Apply
4 3 2 1
, , , F F F F on each master node in opposite direction such that

F F F F F + + +
4 3 2 1

Forces
i
F are functions of the position of contact point C and are evaluated
by

F N F
c c i i
) , (

Where
i
N are the standard shape functions
c c
, are the contact point natural coordinates

_______________________________________________________________________________________________
_______________________________________________________________________________________________

SPRING STIFFNESS K

Depends on master side only
K should be limited for stability consideration
If segment belongs to shell element:

Et K
2
1

Where E is young modulus, t is a nominal shell thickness, and is stiffness
factor. The default value for the stiffness factor is 0.2

If segment belongs to a brick element:

V
A
B K
2
2
1

Where
B: bulk modulus
A: segment area
V: element volume
: stiffness factor (default is 0.2)

If a segment is a shell as well as the face of a brick element, the shell
stiffness is used.

SLIDING WITH SEPARATION AND FRICTION

Surface-to-Surface
Used to simulate contact between two surfaces and works properly if the two
surfaces are simply convex

_______________________________________________________________________________________________
_______________________________________________________________________________________________

On the master side the segments must be connected to solid or shell
elements
It is not allowed to put the same node in the two impact surfaces
Perform two times the algorithm for type node-to-surface contact
Symmetric computation slave and master surfaces play identical rule
First time solve penetration of nodes of surface 1 with respect to
segments of surface 2
Second time solve penetration of nodes of surface 2 with respect to
segments of surface 1
Is not effected by discontinuous (bad) mesh
Penalty stiffness is the minimum of the impacting segments
Precision is improved, however, cost is increased.

SINGLE SURFACE

The single surface contact algorithm evolved from the surface to surface
contact. Post contact searching follows the procedure employed for the
surface to surface contact.
The algorithm loops through the contact segments and computes the
normal segment vectors and accumulates an area weighted average at
the nodal points to determine normal vectors at the nodal points.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Then it loops through the bucket sorts and locates the node, which is
nearest.
It checks to see if nearest node is within a penetration tolerance.
For shell elements, determine if the nearest node approaching the
segment from positive or negative side. Project both the node and the
contact segment along the nodal line vectors.
Check for interpenetrating nodes and if a node has penetrated apply a
nodal point force.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
CONTACT TYPES IN LSDYNA IMPLICIT

All implicit contact interfaces use the penalty method
Disabling the shooting node logic is recommended
SNLOG=1 on optional contact interface card "b"
Oriented normal vectors are recommended
Automatic contact types often fail for implicit

Special parameter settings for implicit
SHLTHK=1 or 2 on *CONTROL_CONTACT (required)
ORIEN=2 on *CONTROL_CONTACT (recommended)
SNLOG=1 on optional contact interface card "B" (recommended)

Contact Interface Normal Vectors
Surface penetration determined using outward normal vector
brick elements: unique outward direction for each face
shell elements: sign of outward normal must be determined
shell elements: right-hand rule on local connectivity gives outward normal

*CONTROL_CONTACT parameter ORIEN activates normal checking
default "on" for input by part ID, "off" for input by segment list

Penetrating nodes are relocated during initialization
severe deformation if normal vectors are incorrectly chosen
observe model geometry at time=0 using postprocessor
check "d3hsp" file for messages concerning reversed normal vectors
This behavior is exactly the same for explicit and implicit simulations

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Normal Vector Trouble

2D Example: penetration detected due to reversed normal vectors

Automatic Normal Vectors

The LS-DYNA "automatic" contact interfaces allow randomly numbered
segments
approaching nodes are tracked, identifying and saving "outward"
directions
tracking algorithm assumes many small steps
excellent for explicit analysis
can fail for large, implicit steps

may be safe for some applications, if tracking is properly activated
tracking is activated for all nodes that are initially in contact
master
slave
Input geometry
slave
master
After penetration check
_______________________________________________________________________________________________
_______________________________________________________________________________________________
local relative velocities can be low even if global step size is large
pillar collapse during roof crush
be ware of shooting nodes, negative energy

GUIDELINES:

Contact type node-to-surface contact is simple and works almost all the
time with if the master surface is a closed surface.
Contact type surface-to-surface is also reliable if both master and slave
surfaces are closed.
The advantages of these contact types is that contact forces can be
monitored in RCFORC with no need to specify a transducer.
In general, fine mesh leads to better representation of the contact area
and is more stable.
Contact normals must point to the opposing surface in these contact
types.
Contact type Automatic Single Surface is the most efficient and reliable
contact for general surfaces if time steps are not too large so the tracking
algorithm can work properly.
Must use *contact_force_transducer to be able to obtain contact forces in
RCFORC.
A contact thickness must be defined in order to check for initial
penetrations.
In implicit analysis since time step is relatively large full searching my
become necessary at least every cycle.
Bucket sorting is done every several cycles (user controlled) and it is an
expensive part of the contact algorithm. In some problems it may be
necessary to reduce this to less frequency for sorting.
Distance to the nodes in the slave node bucket and 27 neighbor buckets
are calculated. The cost is proportional to N*(27N-1) where N is the
number of nodes.
Nodal based search may not be able to find the closest master segment,
segment based search is more reliable.
Nodal based projection is more reliable than segment based projection.
Segment based projection may cause a node to interpenetrate between
contiguous segments. This leads to energy growth.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
FRICTION

Friction phenomena exist whenever contacts occur.
Effect of friction may play very important role, for instance, in metal
forming frictional effect between the work piece and the die will strongly
influence the forming process.
Classical friction law (linear dry friction) and non-classical friction laws (like
nonlinear friction and elasto-plastic friction) exists for modeling friction.
Classical friction law has physical deficiencies since it allows small relative
motion of two contacting bodies even if the friction force is less than N
s
.
The Classical friction law ignores the dependence of friction coefficient on
the relative sliding velocity. This dependency is significant when the
relative sliding velocity is large.

Dry Friction (Coulomb friction)
The actual coefficient of friction
v dc
d s d c
e
.
) (

+ is assumed to depend
on DC and
d s
, ,

Where
s
Coefficient of static fiction (defined as FS in manual)
d
Coefficient of dynamic friction (defined as FD in manual)
DC exponential decay coefficient
V relative velocity of the surface in contact

t
C C
V
o
1

If exponential decay 0 DC then N f and
s s c

If exponential decay
20
0 . 1 e DC then N f and
d d c
.
_______________________________________________________________________________________________
_______________________________________________________________________________________________

MATERIAL NONLINEARITY
_______________________________________________________________________________________________
_______________________________________________________________________________________________


ENGINEERING STRESS-STRAIN CURVE:

Nominal (engineering) stress
0
/
n

Where p applied load and
0
A undeformed cross-sectional area,
Engineering strain:
o o
o
o
l
l
l
l l

True stress:
A
p

Where A is the actual area
Logarithmic strain:
o
l
l
lu

PLASTICITY

At low stress level, the stresses in elastoplastic material depend only on the
state of strain; however, above a certain stress level, called the yield stress,
y
, non-recoverable plastic deformations are obtained.

Post- yielding behavior from uni-axial tension tests typically shows the
following behaviors.
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Hardening, like metals, concrete, rock with small deformations
Ideal, crude idealization for steel, plastics, etc.
Softening plasticity models, like dense sand, concrete with large
deformations

ELASTIC-PLASTIC BEHAVIOR

Plasticity Theory:

Yield criterion
(Von-Mises, typically)

Flow rule
For metals use associated rule (flow rule and yield criterion are the
same functions)
For granular materials like soil use non-associated rule (flow rule and
yield criterion are different functions

Hardening rule
This is the rule that describes how the yield surface is changed by the
history of the flow.
1. Kinematic hardening
2. Isotropic hardening
3. Mixed hardening

_______________________________________________________________________________________________
_______________________________________________________________________________________________

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Uniaxial Representation
y
t
y
0
-
~
y
-
t
y
-
-
y

Isotropic Hardening
Figure (a), yield surface expands uniformly.
Yield reappears when
B
>

Kinematic Hardening
Figure (b), yield surface translates in stress space but do not change size,
shape, or orientation. The elastic range of 2
y
o
is maintained.
y
o
y
t
y
t
2
~
+

Mixed Hardening
Figure (c), yield surface translate in stress space and change size.
y
t
y
t
y
t
y
o
2 2 < + <
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Converting Test Data to LSDYNA Material Data

Assumption: elastic deformation is small compared to total deformation and
plastic flow occurs at constant volume.

Covert force (p) vs. elongation (
o
l l ) curve to engineering stress vs.
engineering strain
0
/
n

o o
o
o
l
l
l
l l

Convert engineering stress to Cauchy (true) stress and engineering strain
to true strain
Using incompresibility
o o
l A Al
) 1 ( ) 1 (
o n
o
o
o o o
l
l l
A
p
l
l
A
p
A
p
+
+
) 1 ( ) 1 ( ) (
o
o
o
o
lu
l
l l
lu
l
l
lu +
+
If one is interested in post necking strain then the following equation can
be used

) (
A
A
lu
o

_______________________________________________________________________________________________
_______________________________________________________________________________________________

_______________________________________________________________________________________________
_______________________________________________________________________________________________
The following figure depicts the applicability of the various strain and stress
mesures:

DAMPING
_______________________________________________________________________________________________
_______________________________________________________________________________________________

DAMPING

In general damping in structure is not viscous. It is due to hysteresis and
usually modeled by viscous damping.
Treatment of damping:

1. Phenomenological damping methods:
Where actual physical dissipation is modeled (seldom used)
Elastic plastic hysteresis loss
Structural joint friction
Material microcracking

2. Spectral damping methods (Rayliegh or proportional damping)
Viscous damping is introduced by means of specified fraction of critical
damping.
Rayleigh damping is defined by

K M C +

Where C, M, and K are damping, mass, and stiffness matrices.
and Are the mass and stiffness proportional constants.
The keyword options are:

*DAMPING_GLOBAL
*DAMPING_PART_MASS
*DAMPING_PART_STIFFNESS

For the first option you can either specify a load curve for damping or a
system damping constant d. With system damping the acceleration is
computed as:

) (
1
int damp ext n
R R R
M
a

where dmv R
damp

DAMPING
_______________________________________________________________________________________________
_______________________________________________________________________________________________
The best damping constant for the system is usually based on the critical
damping factor for the lowest natural frequency of interest.
The last two keyword options assigns mass weighted damping and
stiffness damping by part ID
You can specify mass weighted damping by a load curve.
Mass weighted damping damps all motion including rigid body motions. It
is more effective for low frequencies.
Stiffness proportional damping can be input by specifying the value of .
Using the normal mode methods we can write

ij i i i i
T
i i
T
i
K M C 2 ) (
2
+ +

Where
i
is the ith frequency.
If only stiffness proportional damping is needed, then
i i i
2
2
.
If let say only 10% of critical damping is desired for the
i
frequency then

i
2 . 0

Stiffness proportional damping is effective for high frequencies and it is
orthogonal to rigid body motions.
Stiffness proportional Rayleigh damping is introduced by replacing the
incremental stress computation by

2
1
2
1
2
1
+ + +
+
n
elas
stiff
n n
C
t
C

Where
stiff
is the stiffness proportional damping coefficient, t is the time
step, and
elas
C is the elastic consecutive tensor. The stiffness proportional
damping coefficient is specified by material.
Mass proportional damping is added with

+
+
+ +
1
1 1
int
n
V
e
mass
n
T
n e
u M
t
dV B F

Where
mass
is the proportional damping coefficient, M, is the element mass,
matrix, and u , is the vector of nodal incremental displacements. In
DAMPING
_______________________________________________________________________________________________
_______________________________________________________________________________________________
LSDYNA/IMPLICIT implementation, a unique stiffness proportional damping
coefficient may be specified for each part. However, one mass proportional
damping coefficient operates on the entire model mass. The damping force is
proportional to a midstep velocity approximation.
REDUCED INTEGRATION & HOURGLASSING
_______________________________________________________________________________________________
_______________________________________________________________________________________________

SELECTIVE REDUCED INTEGRATION
&
HOURGLASSING

SELECTIVE REDUCED INTEGRATION

Zero energy mode, hourglass mode, rank deficiency

A lower integration rule is desirable for the following reasons:
Lower computation cost.
Tendency to soften an element, thus countering the overly stiff behavior
associated with an assumed displacement field.
Reduce shear locking in elements with transverse shear deformations (in
shell elements). Softening comes about because certain higher order
polynomial terms happen to vanish at gauss points of a lower order rule,
so that these terms make no contribution to strain energy.

However reduced integration rule produces zero energy deformation mode
(or hourglass mode)
(A pattern of nodal D.O.F. produces a strain that is zero at all integration
points)
If the integration points (gauss point) sense no strain under a certain
deformation mode, the resulting element stiffness matrix will have no
resistant to that deformation mode.
The stiffness matrix for bilinear four nodded elements has 8 eigenvalues
in which lead to 8 independent displacement modes, they are:

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Three rigid body displacement
Two translation and one rotation (will not contribute to the strain energy)

Three constant strain modes
Stretching, uniform extension, and shear

Two bending (flexural) modes

If exact or recommended integration is used the stiffness matrix will have
8-3=5 nonzero eigenvalues
(The 3 zero eigenvalues correspond to rigid body motion)
If reduced integration is used the stiffness matrix will have only 3 nonzero
eigenvalues
_______________________________________________________________________________________________
_______________________________________________________________________________________________
So we will have two additional displacement modes that will not contribute
to the strain energy, they are the spurious zero-energy modes

Example:
Four-nodded plane elements integrated by one-integration point

0
v
cxy u

cxy v
u

0

The above displacement fields will lead to
0
xy yy xx
At gauss points ) 0 , , , , ( 0

v v u u

HOURGLASSING

Out-Of-Plane "W" Hourglass Mode in Shell Elements

The figure below depicts the effect of hourglass on deformation. The vehicle
had the same initial kinetic energy and both runs have the same termination
time.

undeformed deformed
_______________________________________________________________________________________________
_______________________________________________________________________________________________

HOURGLASS CONTROL

Adding hourglass stiffness must provide stability to the element without
stiffening the element to legitimate modes that are already working well. To
suppress the hourglass deformation modes, hourglass viscosity stresses are
added to physical stresses at the local element level.

Two methods are generally used:
1. Hourglass viscous forces
2. Hourglass stiffness forces

VISCOUS FORMS
Standard LSDYNA (default)
Flanagan-Beltyschko (2)
Flanagan-Beltyschko with exact volume integration (3)
STIFFNESS FORMS
Flanagan-Beltyschko (4)
Flanagan-Beltyschko with exact volume integration (5)

Some Comments on the Hourglass Control Methods

Viscous form sometimes works better and is the default.
Stiffness form is more stable in many applications and is preferred for auto
crash and sheet stamping.
The stiffness form may results in a stiffer response.
Flanagan-Beltyschko hourglass control behaves better for large rotations.
Fully integrated elements do not hourglass.
If localized hourglass occurs in area far from design critical areas, then it
is admissible.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Hourglass modes are orthogonal to real deformation and rigid body
motions.
Hourglass energy vs. time appears in GLSTAT and MATSUM.
Hourglass energy should be small relative to internal energy (0%-10% is
good).
The work done by hourglass stabiliztation forces dose not appear in
energy balance equation unless it is signaled in the CONTROL_ENERGY
card.

The following cards are used to control hourglass parameters.

*CONTROL_ENERGY
switch to have hourglass energy calculated

*CONTROL_HOURGLASS
set hourglass type
user can modify hourglass coefficient

*HOURGLASS
set hourglass type and other parameters for use in a specific parts

*PART
can be signaled to use a specific hourglass ID

QUASI STATIC SIMULATION
_______________________________________________________________________________________________
_______________________________________________________________________________________________

QUASI STATIC INITIALIZATIN
AND

QUASI STATIC INITIALIZATIN

In some cases it is desirable to have prestress loading of a structure before a
transeint dynamic analysis is performed. This situation occurs in many
applications. To name few as follows:

Submarine under hydrostatic pressure
Gravity and body force loading
Bird impact on rotating jet engine fan blades
And many others.

It is possible to perform quasi-static initialization by three methods as follows:

1. Monotonic increment of pre loading
2. Dynamic relaxation
3. Implicit step calculation

These methods are described as follows:

1. Monotonic increment of pre loading

In here the applied load for prestress is applied in monotonic increments up
to the final load before transient analysis begins. For instance, if car crash is
desired then, the gravity is applied with in a load curve from time zero to
some other time (say t1). Then this gravity loading is held constant for the
duration of the simulation. The impact velocity of the car can start with a
second load curve that has its starting time to be t1.

2. Dynamic relaxation

A load curve can be used for both prestressing and transient analysis. In the
prestressing (or dynamic relaxation) phase LSDYNA approximates a solution
of the nonlinear quasi-static problem by incorporating damping in the update
of the displacement field.
_______________________________________________________________________________________________
_______________________________________________________________________________________________

3. Implicit step calculation

In here one can solve the prestress problem using an implicit solution phase.
With in this phase one needs to include the following card:

*INTERFACE_SPRINGBACK_DYNA3D

This will cause the code to generate a dynain file at the end of the
simulation. This file contains initial stresses imposed on the elements at the
end of the simulation for the part defined. Then one can take these initial-
stresses and put into the input file for a second explicit phase with
prestressing.


WHY STATIC ANALYSIS WITH AN EXPLICIT CODE:

Direct Solution May Fail When:
The material law is highly nonlinear. Extremely complicated material
behavior is much more easily accommodated with explicit methods

Number of elements is extremely large
The IOs are less important
The memory required is also less
Explicit method does not require matrix inversion

Matrices must be re-evaluated at each time step and for most iterations
The CPU time of an explicit solution becomes competitive

The problem includes several contact surfaces
Contact algorithms are very efficient in explicit codes

Excessive rigid body motion
Explicit algorithm works very well

Direct solution may fail when local buckling occur in large scale finite
element models

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Quasi-static analysis with large deformation and contact (like roof crash and
door intrusion) is expensive and some times impossible using implicit
method. Explicit methods can be used with time and mass scaling

Performing Static Simulations

Dynamic FE codes account for inertia in formulation.
Applied forces in static FE codes results in stress, however, in dynamic
FE codes results in stress waves.
To model quasi-static events using explicit dynamic codes the events
must be scaled in time (from seconds to milliseconds).
Need to apply forces or displacements in ramp curves. Sudden application
of loads will results in stress waves that could be much larger than
stresses at the maximum applied forces when performing static simulation
using implicit LSDYNA.
Note that to reduce kinetic energy
2
2
1
mv E we have two options. Either
reduce velocity v or reduce mass m.
For instance to reduce E by we can either reduce the velocity by
2
1

or the mass by . Velocity reduction has more pronounced effect on
kinetic energy reduction.

Three issues must be addressed to determine suitable loading rate when
simulating quasi-static tests using explicit LSDYNA:

Strain rate effects (must not be included in material model).
Amplitude of stress wave in material should be small
( ; cvA A F / E c ). Nonreflecting boundary can be used to reduce
the effect of stress wave.
Ratio of kinetic energy to strain energy should be small.

Note that when performing quasi-static simulations the higher the rate of
loading the higher the elongation at failure.

_______________________________________________________________________________________________
_______________________________________________________________________________________________
Quasi-Static Post-buckling Analysis Example

Time and Mass Scaling

Time scaling:
*CONTOL_TIMESTEP

Minimum time step size can be specified. For shell elements and the
following material:
*MAT_PLASTIC_KENEMATIC
*MAT_POWER_LAW_PLASTICITY
*MAT_STRAIN_RATE_DEPENDENT_PLASTICITY
*MAT_PIECEWISE_LINEAR_PLASTICITY
The moduli in the element material properties will be modified to keep the
time step as the assigned value.
You need to specify the parameter TSLIMT in the
*CONTROL_TIMESTEP
Dangerous exercise in FE simulation.

Mass scaling:
*CONTOL_TIMESTEP

Minimum time step size can be specified, Preferred in FE simulation. For
all elements and materials;
The mass in the element material properties will be modified to keep the
time step as the assigned value.
_______________________________________________________________________________________________
_______________________________________________________________________________________________
You need to specify the parameter DT2MS in the
*CONTROL_TIMESTEP.
Positive values are for quasi-static analysis or time history analysis with
insignificant inertial effects.
Negative values is the minimum time step size permitted and mass
scaling is done if and only if it is necessary to meet Courant criterion.
You can bound mass increase by the parameter ENDMASS in the
termination card.

LSDYNA EXPLICIT VS IMPLICIT COST

Each increment in IMPLICIT analysis consists of at least one iteration, but
usually more than one.
Each iteration in IMPLICIT requires the solution of a set of simultaneous
equations
The CPU cost per iteration is roughly proportional to the number of
degrees of freedom in the model squared.
Each increment in EXPLICIT consists of one group of equations-there are
no iterations.
The CPU cost per increment is directly proportional to the number of
degrees of freedom in the model.
As the size of the model increases, there is a point at which EXPLICIT
may become more cost effective than IMPLICIT for simulations that could
be solved with either.

ADAPTIVITY
_______________________________________________________________________________________________
_______________________________________________________________________________________________

ADAPTIVITY

Theoretical and Practical Aspects of Mesh Adaptation

Objective
Even with the development of newer and faster computers, it is desired to
make simulation as efficient as possible. Often times, the potential
behavior of a model can only be guessed at, requiring finer meshes and
thus, greater amounts of processing time.

What is adaptive meshing?
Adaptive meshing begins with the solution of an initial and usually coarse
mesh.
Based on error estimators, relative element discretization errors are
quantified for the entire mesh.
In areas of large error, the mesh is refined or enriched
The objective of the adaptation is to equally distribute error over the entire
mesh.

Whats what?
H-method alters the elements of the initial mesh by decreasing the size
(refinement) or increasing the size (de-refinement).
P-method changes the order of the shape function polynomial while
keeping the element geometry constant.
R-method moves the nodes of the elements while keeping their number
and connectivity the same.
Can also have combinations of methods (h/p for example) or total
remeshing.

Where Its At -- H-Adaptivity
H-Adaptivity - many major codes use this form of mesh adaptivity,
including: LS-DYNA

Future Trends
More emphasis seems to being placed on the h/p method because of its
ability to enrich and refine the mesh.

ADAPTIVITY
_______________________________________________________________________________________________
_______________________________________________________________________________________________
LSDYNA ADAPTIVITY: H-adaptive refinement uses error norms to generate
elements where they are needed. Solution errors are minimized without a
manual, trial and error approach. A one side two neighbor rule is used for
ease in implementation.

ADAPTIVE DATA STRUCTURE

Internal software and user databases are an issue in h-adaptivity.
New element connectivities are generated for each adaptive remeshing.
Requires resorting for disjointness and subcycling. Connectivities must be
available for post-processor also. Node numbers do not change.
In parallel computations the domain decomposition is repeated for load
balancing considerations (available soon).
LS-TAURUS and LSPOST will read through and plot the adaptive
database.
Optimized mesh data is available for future run without adaptivity.

REFINEMENT INDICTOR

A deformation based indictor-using angles- no error indicator.
Check angular deformation between elements in plane and out-of-plane.

ADAPTIVITY
_______________________________________________________________________________________________
_______________________________________________________________________________________________

If
1 to
> the refinement, where
1 to
is user determined. Limit for refinement
has to be determined by user as well as the number of refinements.

CONTINUE CALCULATION OR BACK UP?

After computing the changes in deformation one can:
Refine the mesh and continue.
Backup to an earlier state and repeat calculation.

Advantages and disadvantages of continuing:
Speed but for accuracy the frequency of adaptive remeshing must be high
and this may be costly.
Much easier implementation.
Will be inefficient on MPPs.

Advantages and disadvantages of repeating the calculation:
Speed since only very intermittent error checks are made.
Speed due to the fact that all the vector blocking of elements can be
accomplished.
Nearly all options work with adaptive remeshing.
Domain decomposition on MPPs.
Much more difficulty in implementation.

ADAPTIVITY
_______________________________________________________________________________________________
_______________________________________________________________________________________________

TROUBLE SHOOTING
_______________________________________________________________________________________________
_______________________________________________________________________________________________

SPRINGBACK OPTIONS IN LS-DYNA

Springback analysis is important in metal forming, structural design
l Apply load, causing plastic deformation
l remove load, compute residual deformation and "springback"
displacement
l static springback solution needed => explicit dynamics not good
l design problem: find tool shape which gives correct part geometry after
springback
l two modeling strategies are available:

Implicit Method
l run entire simulation using static implicit method
l load completely removed at final time step
l control rigid body modes using BCs with late birth time (if necessary)

Seamless Springback method
l standard explicit simulation used to apply load
l automatic, seamless switch to implicit for static springback
l switch formulation at termination time
l springback for a subset of parts, additional BCs optional
l all contact interfaces automatically killed

Minimal input required
l *INTERFACE_SPRINGBACK_SEAMLESS
part set ID
optional constraints to eliminate RB modes
l *CONTROL_IMPLICIT keywords are optional
change number of steps (default: NSBS=1)
change stepsize (default: DT0 = TERM)
alternate "springback" defaults used: IAUTO=1, IAS=1 (see manual)

TROUBLE SHOOTING
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Implicit Springback

Single vs. multi-step springback
l many springback problems can be unloaded in one nonlinear step
l for large springback, geometric nonlinearity requires several steps

Artificial stabilization
l Allows multi-step springback
l artificial springs constrain motion of every node
l springs soften during simulation, vanish at termination time
l accurate springback only when stabilization is completely removed

Automatic time step control
l modify new step size using iteration count of previous step
l repeat failed steps with smaller time step size
l works in conjunction with artificial stabilization for springback

Nonlinear, Multi-Step Springback Analysis

Single vs. multi-step springback
l many springback problems can be unloaded in one nonlinear step
l for large springback, geometric nonlinearity requires several steps
l seamless default: try single step, activate two features below just in
case...

l adjusts step size automatically to hold convergence "effort" constant
l repeats failed steps with exponentially smaller step size
l very, very persistent

Artificial Stabilization
l artificial springs added to model
l springback over several steps
l springs vanish at termination time
l stiffness multiplier option
l load curve option for k vs. t

TROUBLE SHOOTING
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Static Simulations Require Constraints

Boundary conditions and rigid body modes
static implicit simulation requires boundary constraints
rigid body modes must be eliminated
apply translational constraints to three nodes

a: reference node, dx=dy=dz=0
eliminates all translational modes

b: node along X-axis, dy=dz=0
eliminates rotations about y- and z-axis

c: node along Y-axis, dz=0
eliminates rotation about x-axis

0.0 1.0
0.0
1.0
0.1
10.0
Simulation Time
S
p
r
i
n
g

S
t
i
f
f
n
e
s
s

a
X

Y
Z

c
b
TROUBLE SHOOTING
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Numerical Roundoff in the Linear System

Implicit FEA simulations are susceptible to numerical errors
Force (residual) vector computation
area, volume integration, stress evaluation
similar to explicit methods
single precision usually adequate
Gaussian Elimination
roundoff errors during stiffness factorization
rule of thumb: numerical digits lost = c log(condition number)

In well conditioned model, equilibrium iteration corrects G.E. errors:

Example - Nodal Displacements in Beam Tower Model, mm.

linear nonlinear
double precision 127.7 127.7
single precision 124.2 127.8

In poorly conditioned models, increased sensitivity to errors in force vector
compound the problem.

Condition Number vs. Mesh Density

The condition number of a model is effected by mesh density.

Example: simple beam modeled with n beam elements
Lowest eigenvalue remains virtually unchanged
Highest eigenvalue (axial mode) increases with mesh refinement:

m
k

m
k
2
2

m
k
n
2

one mass two masses n - masses

M
K
M
K
M
K
M
K
TROUBLE SHOOTING
_______________________________________________________________________________________________
_______________________________________________________________________________________________
for n-element lumped mass beam model, element length
n
L
L
0
,

( )
3 1 2 max
n AL
L
EA
n
m
k
n
,
_

so condition number varies with mesh density cubed

Condition Number vs. Boundary Conditions

The condition number is effected by boundary conditions

Example: simple beam modeled with n beam elements
Highest eigenvalue remains virtually unchanged
Lowest eigenvalue (cantilever mode) increases with added constraint

min
4 . 7
min

8 . 39
min

cantilever simple support double clamp

boundary conditions improve conditioning, decrease roundoff error

TROUBLE SHOOTING
_______________________________________________________________________________________________
_______________________________________________________________________________________________

NONLINEAR CONVERGENCE PROBLEMS

Convergence trouble is the most common problem. Error messages
displayed by LS-DYNA :

l iteration limit reached
displacement and energy tolerances were not satisfied, abandon step

l divergence
out-of-balance force R is growing, reform K and continue iterations

l energy explosion
energy norm grows beyond 1,000,000x initial value, abandon step

l negative eigenvalues
error from linear equation solver while computing K
-1

input parameter gives procedure (default: ignore, try to continue)

l line search stepsize zero
new displacement u does not reduce R
reform K if not new last iteration, otherwise abandon step

Nonlinear Convergence Problems

Several diagnostic tools are available
l NLPRINT print flag / sense switch for diagnostic information
l negative eigenvalue flag optionally repeats with smaller step immediately
after encountering negative eigenvalues, avoiding potential inverted
element trouble (DEFAULT=ignore negative EVs)
l <ctrl-c> conv overrides tolerances, forces convergence this step
l <ctrl-c> iter creates d3iter plot database every iteration

TROUBLE SHOOTING
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Procedures for solving convergence problems
l determine reason for termination (examine error messages)
l activate print flags to get more information
l view deformed geometry during iteration process using "d3iter" database
l carefully inspect input deck

Troubleshooting Non-Convergence

Iteration Limit Reached error message
l observe convergence progress of displacement, energy norms
(view d3hsp file or activate NLPRINT flag for screen output)

l if tolerances nearly satisfied:
allow a few more iterations
relax tolerance slightly
interactively force convergence using "<ctrl-c> converge"

l if convergence progress is slow
switch to Full Newton method (ILIMIT=1)
check for incorrect material properties
increase step size

l if convergence progress is erratic
decrease contact penalty stiffness scale factor 10x (monitor
penetration)
decrease step size

Divergence and Energy Explosion error messages:
l some part of model is likely deforming very rapidly
l generate d3iter database using "<ctrl-c> iter"
l observe model during iteration process, watch for locally large
deformation
l check for unbounded applied loads
l check for incorrect yield stress or hardening (temporarily switch to elastic)
l check contact definitions for reversed normals or excessive penetration
l decrease time step size

TROUBLE SHOOTING
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Negative Eigenvalue error message - four possibilities

1) Non-physical material properties
l negative Young's modulus, negative slope of stress-strain curve, etc
l check d3hsp file to be sure LS-DYNA read the input deck properly

2) Inverted elements
l negative volume elements in initial mesh (check in preprocessor)
l elements inverted when nodes relocated during contact initialization
check messages in d3hsp f ile, view mesh at time=0 with postprocessor
l elements inverted due to excessive deformation
activate plot state output to get many pictures just before trouble
begins

3) Rigid body modes
l review boundary conditions and constraints
l watch out for beams that are free to spin about their axes

4) A feature in the model is not available for implicit analysis
l warning messages are not always issued when illegal feature is used
l check keywords against list in Appendix M of User's Manual

TROUBLE SHOOTING
_______________________________________________________________________________________________
_______________________________________________________________________________________________
WAYS TO BATTLE NONCONVERGENCE

1. Allowing more attempts or decreasing the minimum time increment.
2. Loosening up the convergence tolerances.
3. Limiting the maximum number of stiffness reformation in order to try a
smaller step size rather wasting too many iterations on a difficult step.
4. Modifying your model to add stability to an other wise unstable
solution which could include hidden instabilities like:

* Material instabilities
* Compression instabilities
* Contact instabilities
* Rigid body instabilities
* Modeling error

5. Use displacement control when possible. Load control is in general
less stable.
6. Build your way up in a complex model. Start with a simple convergent
model, it is better to add complexity than to subtract them from a
complex non-convergent model.
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Lab Exercise Problems

These example problems are designed with two objectives in mind:

Demonstrate and teach LS-DYNA syntax
Provide understanding of nonlinear implicit FE analysis

_________________________________________________________________________________
LS-DYNA Implicit Workshop

Problem #1-1: Tensile Test

_________________________________________________________________________________

Objectives
Learn how to activate implicit mode.
Learn how to select linear or nonlinear analysis.

Problem Description
A static tensile test is simulated using shell elements and elastic material model. One end of the
specimen is constrained, while concentrated nodal loads are applied at the other end. Uniform
stresses develop in the narrowed center section.

Input Filename: tensile1.k

Procedure
Copy the input file to your local directory. Using an editor, view the input file and answer the
following questions:

1. Which shell element is used? What is the yield stress?________________________
2. How many steps are used to apply the load? ________________________________
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Run the input deck. Record below the total applied load and total tip displacements using lspost
(you can use NODFOR for post processing

3. Applied load = __________ Tip displacements_____________

Modify the input deck to increase the applied load 100x and rerun the simulation.

4. Is the simulation linear or nonlinear? (does the tip displacement scale linearly with the applied
load? ____________________________________________________________

Modify the *CONTRIL_IMPLICIT_SOLUTION to perform a linear simulation. Modify the
element formulation to 21. Rerun and record the results.

5. Are you solving a linear problem now? ___________________________________
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #1-2: Elastic-Plastic Tensile Test

_________________________________________________________________________________

Objectives
Learn how to observe convergence behavior of nonlinear equilibrium iterations.
Learn to use automatic time step control for nonlinear problems.
Learn the benefits of force vs. displacement controlled simulations.

Problem Description
A static tensile test is simulated using shell elements and a nonlinear, elastic-plastic material
model. One end of the specimen is constrained, while concentrated nodal loads are applied at the
other end. Uniform stresses develop in the narrowed center section.

Input Filename: tensile2.k

Procedure

6. Which material model is used? What is the yield stress? ______________________
7. How is load applied? __________________________________________________
8. How many steps are used to apply the load? ________________________________

Run the input deck. Does the job run to completion? ( ) normal termination ( ) error
termination

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
9. At what time does the solution begin to struggle? time = _____________________

Using the postprocessor, plot the X-displacement of an end node vs. time.

10. What is happening at the time shown above? _____________________________

11. max end displacement ______ max eff. stress ______ max eff. strain ____

Activate the nonlinear print flag to get more information about the nonlinear solution process.

12. What two methods are available for this? __________________________________

13. Repeat the simulation. What do you observe? ______________________________

Switch from load control to displacement control, and repeat the simulation (Hint: helpful
keywords are commented out in the original input deck). Using the postprocessor, again plot the
X-displacement of an end node.


15. Why is this problem easier to solve? _____________________________________

Return to the original input deck, and activate automatic time step control (IAUTO=1 on the
keyword *CONTROL_IMPLICIT_AUTO). Repeat the simulation.

16. What happens? ______________________________________________________


Using the ASCII menu, load the GLSTAT database and plot the step size vs. time.

18. When does the step size begin to decrease? _________________________________

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
implicit tensile test, elastic-plastic material
$
$ test coupon, 200 mm long, 20 mm width, 2.67 mm thickness, 50.8 mm gauge length
$
$ units; mm, s, ton, N
$
$ By A. Tabiei
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
*CONTROL_TERMINATION
1.0000
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
$ imflag dt0 iefs nstepsb igso
1 0.01 0 0 0
$
$ nlsolvr ilimit maxref dctol ectol rctol lstol
0 0 0 0.0 0.0 0 0
$ dnorm divflag inistif nlprint
0 0 0 0
$
$ lsolvr prntflg negeig
0 0 0
$
$ iauto iteopt itewin dtmin dtmax
0 0 0 0.0 0.0
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
*DATABASE_BINARY_D3PLOT
0.0100
*DATABASE_EXTENT_BINARY
$ neiph neips maxint strflg sigflg epsflg rltflg engflg
1 1
$ cmpflg ieverp beamip

$
$
*DATABASE_NODFOR
0.0001
*DATABASE_NODAL_FORCE_GROUP
2
*SET_NODE_LIST
2
29,30,31,32,33,38,39
$
$
*DATABASE_ELOUT
0.0001
*DATABASE_HISTORY_SHELL
71,72,73,74,75,76,77,78
79,80
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*CONTROL_SHELL
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
$ WRPANG ITRIST IRNXX ISTUPD THEORY BWC MITER PROJ
0 0 0 0 0 0 1 0
$
*SECTION_SHELL
$ ID elform
1 16
$ t1 t2 t3 t4
2.670E-00 2.670E-00 2.670E-00 2.670E-00
$
*PART
shell tensile strip
1 1 1
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
$ material 37, load curve gives hardening
*MAT_TRANSVERSELY_ANISOTROPIC_ELASTIC_PLASTIC
$ MID RO E PR SIGY ETAN R HLCID
1 0.780E-08 0.207E+06 0.280E+00 0.200E+03 0.572E+03 0.140E+01 13
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
$ material 37, constant hardening modulus
$*MAT_TRANSVERSELY_ANISOTROPIC_ELASTIC_PLASTIC
$ 1 0.780E-08 0.207E+06 0.280E+00 0.200E+03 0.572E+03 0.140E+01 0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
$ hardening curve: effective stress vs. effective plastic strain
*DEFINE_CURVE
13
0.00000000 200.0
0.00006657 200.5
0.00013650 201.1
0.00020990 201.6
0.00028690 202.2
0.00036780 202.7
0.00045280 203.2
0.00054200 203.8
0.00091670 205.8
0.00136600 207.9
0.00190600 210.0
0.00255300 212.1
0.00472900 217.2
0.00654700 220.3
0.00891000 223.4
0.03259000 250.8
0.04616000 270.5
0.06516000 291.5
0.09176000 313.8
0.13830000 342.4
0.20810001 373.2
0.31279999 406.2
0.46990001 441.6
0.70560002 479.7
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
*LOAD_NODE_POINT
204 1 3 1
205 1 3 1
206 1 3 1
207 1 3 1
212 1 3 0.5
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
213 1 3 0.5
*DEFINE_CURVE
3
0.00 0.00
1.00 2000.00
2.00 2000.00
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
$*BOUNDARY_PRESCRIBED_MOTION_NODE
$ 204 1 2 4 1.00
$ 205 1 2 4 1.00
$ 206 1 2 4 1.00
$ 207 1 2 4 1.00
$ 212 1 2 4 1.00
$ 213 1 2 4 1.00
$*DEFINE_CURVE
$ 4
$ 0.00 0.00
$ 1.00 10.00
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
*BOUNDARY_SPC_NODE
30 0 1 1 1 1 1 1
31 0 1 1 1 1 1 1
32 0 1 1 1 1 1 1
33 0 1 1 1 1 1 1
38 0 1 1 1 1 1 1
39 0 1 1 1 1 1 1
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
*NODE
14 1.000000000E+01 1.600000000E+01 0.000000000E+00
*ELEMENT_SHELL
1 1 38 30 14 40
*end

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #2: Ellipsoidal Dome

_________________________________________________________________________________

Objectives
Learn the behavior of different element formulations in static implicit simulations.
Learn how to minimize hourglass problems.

Problem Description
A static load is applied to the center of an ellipsoidal dome. Shell elements are used. Nodes at the
base of the dome are constrained, and included in a NODFOR output database.

Input Filename: ellipse.k

Procedure

1. What is the maximum value of the applied load?_____________________________

2. How many load steps are used? _________________________________________

3. What shell element formulation is used? __________________________________
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

4. What hourglass control type is used? _____________________________________

Run the simulation, and postprocess the results.

5. What happens? _______________________________________________________

6. Applied load __________ Center displacement _______

Repeat the simulation using the S/R Hughes-Liu shell (ELFORM=6), and postprocess the results.

7. ____________________ Center displacement _______

Repeat the simulation using the Fast Fully Integrated shell (ELFORM=16), and postprocess the
results.


9. What conclusion can you draw about the effect of hourglass deformation on overall structural
stiffness for this problem? ____________________________________________

Concentrated applied loads are often responsible for initiating hourglass deformation. Edit the
input deck and replace the single concentrated nodal load with pressure acting on the adjacent
segments (Hint: Helpful keywords are included in the comments inside the input deck). Repeat
the simulation using the default shell.

10. What happens? ______________________________________________________


WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
PINCHED ELLIPSE
$
$ A. Tabiei, March 99
$
1.0000
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
1 0.01 0 0 0
$
0 0 0 0.0 0.0 0 0
0 0 0 0
$
0 0 0
$
1 0 0 0.0 0.0
$
$ imass gamma beta
0 0.0 0.0
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
1 1

0.01
*DATABASE_GLSTAT
0.01
*DATABASE_NODFOR
0.01
2
*SET_NODE_LIST
2
1 7 13 19 25 31 37 48
54 60 66 72 78 84 91 97
103 109 115 121 127 206 212 218
224 230 236 242
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
*CONTROL_HOURGLASS
4
$
*MAT_elastic
$ MID RO E PR
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
1 7.890E-09 2.100E+05 3.000E-01
$
*SECTION_SHELL
$ SID ELFORM
1 0
3.000E-00 3.000E-00 3.000E-00 3.000E-00
$
*PART
SHELL
$ PID SID MID ADPOPT
1 1 1
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
*LOAD_NODE_POINT
$ NID DOF LCID SF
164 3 1 -1.0
$
*DEFINE_CURVE
1
0.00000000000000E+00 00.000000000000E+00
1.00000000000000E+00 500000.00000000E+00
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
$*LOAD_SEGMENT
$$ LCID SF N1 N2 N3 N4
$ 2 -1.0 154 155 164 163
$ 2 -1.0 155 156 165 164
$ 2 -1.0 163 164 173 172
$ 2 -1.0 164 165 174 173
$
$*DEFINE_CURVE
$ 2
$0.00000000000000E+00 00.000000000000E+00
$1.00000000000000E+00 7500.0000000000E+00
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
1 22.627417 -16.97056299999 0.0 7 7
*ELEMENT_SHELL
1 1 1 2 8 7
*END

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #3: Adaptive Ellipsoidal Dome

_________________________________________________________________________________

Objectives
Learn to activate mesh adaptivity in an implicit simulation.
Learn how to minimize hourglass problems.

Problem Description
A static load is applied to the center of an ellipsoidal dome. Shell elements are used. Nodes at the
base of the dome are constrained, and included in a NODFOR output database. Adaptivity is used
to automatically refine the mesh in areas of high curvature.

Input Filename: aellipse.k

Procedure

1. How frequently will the mesh be evaluated for refinement? ____________________

2. How many times can each element be subdivided? __________________________

3. How do you indicate which parts will be adapted? __________________________
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
4. Which element formulation is used? _____________________________________

5. How is load applied? _________________________________________________

Execute the simulation, and view the results with the postprocessor.


7. How many (auto) time steps were used? __________________________________

8. Does the adaptive mesh improve the hourglassing problem ___________________

Switch to pressure driven load application, and repeat the simulation. Postprocess the results.
Using the NODFOR database, verify that the load is applied correctly as the mesh is refined.


10. How many (auto) time steps were used? __________________________________

11. Does the pressure load improve nonlinear convergence? _____________________

Experiment with shell element formulation #16.

12. Does shell type #16 improve hourglassing?_________________________________

13. Does shell type #16 improve convergence behavior (number of auto steps)? ______

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
PINCHED ELLIPSE
$
$ A. Tabiei, March 99
$
1.0000
$
*CONTROL_ADAPTIVE
$ adpfreq adptol adpopt maxlvl tbirth tdeath lcadp ioflag
0.10 5.000 2 3 0.0 0.0 0 0
$ adpsize adpass ireflg adpene
0.0000000 1 0 1.0
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
1 0.01 0 0 0
$
0 0 0 0.0 0.0 0 0
0 0 0 0
$
0 0 0
$
1 0 0 0.0 0.0
$
$ imass gamma beta
0 0.0 0.0
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
1 1

0.01
*DATABASE_GLSTAT
0.01
*DATABASE_NODFOR
0.01
2
*SET_NODE_LIST
2
1 7 13 19 25 31 37 48
54 60 66 72 78 84 91 97
103 109 115 121 127 206 212 218
224 230 236 242
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
$
*CONTROL_HOURGLASS
4
$
*MAT_elastic
$ MID RO E PR
1 7.890E-09 2.100E+05 3.000E-01
$
*SECTION_SHELL
$ SID ELFORM
1 2
3.000E-00 3.000E-00 3.000E-00 3.000E-00
$
*PART
SHELL
$ PID SID MID ADPOPT
1 1 1 0
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
*LOAD_NODE_POINT
$ NID DOF LCID SF
164 3 1 -1.0
$
*DEFINE_CURVE
1
0.00000000000000E+00 00.000000000000E+00
1.00000000000000E+00 500000.00000000E+00
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
$*LOAD_SEGMENT
$$ LCID SF N1 N2 N3 N4
$ 2 -1.0 154 155 164 163
$ 2 -1.0 155 156 165 164
$ 2 -1.0 163 164 173 172
$ 2 -1.0 164 165 174 173
$
$*DEFINE_CURVE
$ 2
$0.00000000000000E+00 00.000000000000E+00
$1.00000000000000E+00 7500.0000000000E+00
$
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
1 22.627417 -16.97056299999 0.0 7 7
*ELEMENT_SHELL
1 1 1 2 8 7
*END

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #4: Truck Bumper

_________________________________________________________________________________

Objectives
Learn the behavior of different contact interfaces in static implicit simulations.
Learn how to set input parameters for implicit contact problems.

Problem Description
The rear bumper of a truck is modeled using shell elements. A solid, rigid bar is displaced into the
bumper, causing plastic buckling of the support. Post-buckling response is determined.

Input Filename: bumper-truck.k

Procedure

1. What type of contact interface is used? ____________________________________

2. Is the thickness of the shell elements considered? ___________________________

Run the simulation and postprocess the results.

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

3. Is there a convergence trouble? _________________________________________

Activate thickness offsets using *CONTROL_CONTACT, and repeat the simulation.
Convergence will be difficult, so temporarily increase the default displacement and energy
convergence tolerances to DCTOL=0.005 and ECTOL=0.05 using keyword
*CONTROL_IMPLICIT_SOLUTION. Watch the values of the convergence norms displayed on
the screen.

4. Which norm is decreasing most slowly? ( ) displacement ( ) energy

Switch to contact interface *CONTACT_AUTOMATIC_SURFACE_TO_SURFACE. The
AUTOMATIC contact types are superior when gaps are opening. Repeat the simulation, check
the contact results using the postprocessor.

5. Is the contact interface now separating properly? ___________________________

6. Does the force/displacement curve look reasonable?__________________________

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
truck rear bumper
$
$ units: in,lbf,sec
$
0.0500000
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
1 0.0001 0 0 0
$
0 0 0 0.0 0.0 0 0.0
$ 0 0 0 0.005 0.050 0 0.0
0 0 0 1
$ arcctl arcdir arclen arcmth arcdmp
0 0 0.0 0 0
$
0 0 2
$
1 100 0 0.0001 0.001
$
$ imass gamma beta
0 0.0 0.0
$
$ ias scale tstart tend
0 0.0 0 0
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
$ displace bumper 5 inches
$
*BOUNDARY_PRESCRIBED_MOTION_RIGID
4 2 2 1 1.0000000 0 0.0000000 0.0000000
*DEFINE_CURVE
1 0 1.0000000 1.0000000 0.0000000 0.0000000
0.00000000E+00 0.00000000E+00
0.50000000E-01 5.00000000E+00
2.00000000E-01 5.00000000E+00
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
*CONTROL_CONTACT
$ SLSFAC RWPNAL ISLCHK SHLTHK PENOPT THKCHG ORIEN
0 0 0 0 0 0 0
$ (second card blank)

$
*CONTROL_PARALLEL
$ NCPU NUMRHS CONST PARA
0 0 0 0
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
$
*CONTACT_SURFACE_TO_SURFACE
4 3 3 3
0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0 0.0000000 0.0000000
0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
$ optional card A (blank)

$ optional card B
$ PENMAX THKOPT SHLTHK SNLOG
0 0 0 0
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
*DATABASE_GLSTAT
1e-4
*DATABASE_RCFORC
1e-4
1e-4
$
$========1=========2=========3=========4=========5=========6=========7=========8
$
$ material properties
$
*MAT_PLASTIC_KINEMATIC
$ 1020 steel bilinear material model
1 7.30000-4 3.00000+7 0.3000000 30000.000 1.25000+5 0.0000000
0.0000000 0.0000000 0.0000000
*MAT_RIGID
2 7.30000-4 3.00000+7 0.3000000 0.0000000 0.0000000 0.0000000
$ constrain all dof of frame
1.0000000 7.0000000 7.0000000
0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
*MAT_PLASTIC_KINEMATIC
$ 1020 steel bilinear material model
3 7.30000-4 3.00000+7 0.3000000 30000.000 1.25000+5 0.0000000
0.0000000 0.0000000 0.0000000
*MAT_RIGID
4 7.30000-4 3.00000+7 0.3000000 0.0000000 0.0000000 0.0000000
$ allow only y dof of impactor
1.0000000 6.0000000 7.0000000
0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
*MAT_POWER_LAW_PLASTICITY
$ alternate 1020 steel material model
5 7.30000-4 3.00000+7 0.3000000 91353 .1911
0 0
*SECTION_SHELL
1 6 0.0000000 5.0000000 0.0000000 0.0000000 0
0.2500000 0.2500000 0.2500000 0.2500000
*SECTION_SHELL
2 6 0.0000000 5.0000000 0.0000000 0.0000000 0
0.2500000 0.2500000 0.2500000 0.2500000
*SECTION_SHELL
3 6 0.0000000 5.0000000 0.0000000 0.0000000 0
0.1875000 0.1875000 0.1875000 0.1875000
*SECTION_SOLID
4
$*CONTROL_SHELL
$0,0,0,0,0,0,2
$
*PART
support
1 1 1
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*PART
frame
2 2 2
*PART
bumper
3 3 3
*PART
impactor
4 4 4
$
$ ---------------------------------------------------------------
$
*NODE
1 2.400000000E+01 0.000000000E+00 0.000000000E+00 0 0
*ELEMENT_SOLID
1 4 1813 1849 1855 1819 1814 1850 1856 1820
*ELEMENT_SHELL
1 1 1 13 14 2
*END
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #5: Door Beam

_________________________________________________________________________________

Objective
Learn to perform a springback simulation using an entirely static implicit analysis.
Learn the benefits of Full Newton method for contact problems.
Learn about the available displacement convergence norm options.

Problem Description
A doorbeam subassembly is deformed by a rigid pole. Shell elements are used throughout, and
nodal rigid bodies are used to spotweld the components of the doorbeam. The pole is displaced to
deform the doorbeam, then retracted to evaluate springback.

Input Filename: doorbeam.k

Procedure

1. What type of contact interface is selected? Why? ____________________________

2. Why is the doorbeam chosen as the slave side? _____________________________

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Run the simulation and observe the convergence behavior.
Since this is a contact dominated problem, frequent stiffness reformations should help improve
convergence. Modify the nonlinear solution strategy to use more stiffness reformations and fewer
equilibrium iterations. Reform the stiffness matrix every five iterations, and allow twenty
reformations before giving up on the step. Also, relax the displacement convergence tolerance to
0.010.

Repeat the simulation. Postprocess the results and observe the force-deflection curve.

3. Why is the force-deflection behavior not smooth? ___________________________

Select the alternate displacement tolerance scheme using DNORM=1 on
*CONTROL_IMPLICIT_SOLUTION, and repeat the simulation.

4. Does the alternative displacement tolerance become ( ) more or ( ) less strict when total
displacements are large, as they are at the end of this problem?

5. Is the force-deflection curve more smooth? Why? __________________________

6. What is the springback deflection at the center of the doorbeam? _______________

7. run the problem using explicit simulation. Change the termination to a reasonable time for the
simulation. Compare the contact forces from the two runs.

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
static implicit door beam, 150 mm
$
1.10000
$
$ ---------------------------------------------------------------
$
1 0.010 0 0 1
$
0 0 0 0.000 0.00 0 0
$ dnorm divflag inistif nlprt
0 0 0 0
$
0 0 0
$
1 200 0 0.0 0.05
$
$ ---------------------------------------------------------------
$
0.0001000
$
*DATABASE_RCFORC
0.0001000
$
$ ---------------------------------------------------------------
$
2 7.83500-6 200.00000 0.2800000 0.2067000 0.0000000 5.00000+9 0.0000000
0.1700000 4.0000000 0 0
0.0000000 0.0230000 0.0940000 0.1380000 0.1510000 0.3010000 0.7010000 0.9010000
0.2070000 0.2580000 0.3330000 0.3610000 0.3910000 0.4390000 0.5060000 0.5280000
*MAT_RIGID
4 7.83500-6 200.00000 0.2800000 0.0000000 0.0000000 0.0000000
1.0000000 6.0000000 7.0000000
0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
*SECTION_SHELL
8 16 0.0000000 5.0000000 0.0000000 0.0000000 0
3.0000000 3.0000000 3.0000000 3.0000000 0.0000000
*SECTION_SHELL
4 16 0.0000000 5.0000000 0.0000000 0.0000000 0
3.0000000 3.0000000 3.0000000 3.0000000 0.0000000
*PART
PSHELL BBRKT
48 8 2 0 0 0 0 0
*PART
PSHELL BBRKT2
52 8 2 0 0 0 0 0
*PART
PSHELL DOORBEAM
56 8 2 0 0 0 0 0
*PART

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
4 4 4 0 0 0 0 0
$
$ ---------------------------------------------------------------
$
*BOUNDARY_PRESCRIBED_MOTION_RIGID
4 2 2 1 1.0000000 0 0.0000000 0.0000000
$
*DEFINE_CURVE
1
0.00000000E+00 0.00000000E+00
1.00000000E+00 1.50000000E+02
2.00000000E+00 0.00000000E+00
$
$ ---------------------------------------------------------------
$
*CONTROL_CONTACT
0.1000000 0.0000000 0 1 0 0 0
0 0 0 0 0.0000000 0 0 0
$
*CONTACT_NODES_TO_SURFACE
2 1 2 2 0 0 1 1
0.2000000 0.2000000 0.0000000 0.0000000 0.0000000 1 0.0000000 1.0100000
0.0000000 0.0000000 0.0000000 0.0000000 1.0000000 1.0000000
*SET_PART
1
4
*SET_PART
2
56
$
$ ---------------------------------------------------------------
$
*CONSTRAINED_NODAL_RIGID_BODY
3 0
*SET_NODE_LIST
3
77041 77045 77049 77230 77231 77232
4 0
*SET_NODE_LIST
4
77039 77043 77047 77204 77205 77207 77208 77213
5 0
*SET_NODE_LIST
5
76945 76946 76961 77261 77262 77263
6 0
*SET_NODE_LIST
6
76937 76940 76955 77243 77244 77245 77246 77247
$
$ ---------------------------------------------------------------
$
*NODE
1 3.745967041E+03-1.019032959E+03 5.000000000E+02 0 0
*ELEMENT_SHELL
1 4 1 6 7 2
*END
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #6: Seamless Springback

_________________________________________________________________________________

Objective
Learn to run a combined explicit/implicit simulation for dynamic analysis followed by static
springback.

Problem Description
A tensile strip of shell elements is slowly loaded using the explicit dynamic method. At the end of
the simulation, the analysis type is automatically switched to static implicit, the load is removed,
and a static unloaded solution is obtained.

Input Filename: springback.k

Procedure

1. How is load applied to the strip? _________________________________________

2. How is load released for springback? _____________________________________

Add the keywords *INTERFACE_SPRINGBACK_SEAMLESS and *SET_PART_LIST to
activate seamless springback. Run the simulation.

3. What is the termination time for the explicit solution? _______________________

4. How many implicit steps were taken during springback? _____________________
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

5. What was the implicit time step size for springback? _________________________

6. How was this step size chosen? Does its value matter? _____________________

Postprocess the results and verify the springback solution analytically by extracting the following
information:

Elastic strain calculation:
7. Maximum axial stress in tensile test = _____________________________________
8. Youngs Modulus for this material = ______________________________________
9. Elastic strain = sigma/E = ______________________________________________

Springback displacement:
10. Specimen length after loading = _________________________________________
11. Elastic springback = strain * L = _________________________________________

Validation:
12. Computed displacement of tip node after loading ____________________________
13. Computed displacement of tip node after springback _________________________
14. Computed displacement due to springback _________________________________

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
1D tensile test with seamless springback
$ This is a simple uniaxial tensile test with seamless springback.
$ Prescribed velocity is applied to the nodes at the end of the
$ strip, and removed using a death time of 0.005 to allow springback.
$ so springback (elastic) strain should be 2.98e-3
$ Length (loaded) = 44.38
$ so springback tip displacement should be 0.132 (at node 303)
$ units: mm, ton, N, N/mm^2
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ ENDTIM
0.005
*CONTROL_TIMESTEP
$ DTINIT SCFT ISDO TSLIMT DTMS
0.000 0.900 0 -3.0E-07
$ DT/CYCL
5.00E-04
*DATABASE_BINARY_RUNRSF
$ DT/CYCL
2000
*DATABASE_GLSTAT
1.0e-5
*DATABASE_ELOUT
1.0e-5
*DATABASE_NODOUT
1.0e-5
*DATABASE_HISTORY_SHELL
1,2,199,200
*DATABASE_history_node
303
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ NEIPH NEIPS MAXINT STRFLG SIGFLG EPSFLG RLTFLG ENGFLG
1 1
$ CMPFLG IEVERP BEAMIP DCOMP SHGE STSSZ
1 2
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ The following *CONTROL_IMPLICIT cards are not necessary, since
$ all of the default settings are used. They are included for
$ convenience in case you wish to try changing something.
$
$
$ imflag dt0 iefs
0 0.50 0 2
$ ias scale tstart tend
0 0.0 0 0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*INTERFACE_SPRINGBACK_SEAMLESS
982
6 7 0
1 6 0
303 5 0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SET_PART_LIST
982
1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*CONTROL_ADAPTIVE
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
$ ADPFREQ ADPTOL ADPOPT MAXLVL TBIRTH TDEATH LCADP IOFLAG
0.100E-01 0 2 1 4.900E-03 1.0000000 0 0
$ ADPSIZE ADPASS IREFLG ADPENE
0.0000000 0 0 5.0000
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*PART
PART PID = 1
$ PID SID MID EOSID HGID GRAVTY ADOPTY
1 1 1 1
*SECTION_SHELL
$ SID ELFORM SHRF NIP
1 2 1.0 5.0
$ T1 T2 T3 T4
1.00 1.00 1.00 1.00
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*MAT_TRANSVERSELY_ANISOTROPIC_ELASTIC_PLASTIC
1 0.787E-08 0.207E+06 0.33 1.00 4
$ stress-strain curve for the sheet metal
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$--------LOAD CURVE DATA FOR STRAIN STRESS ------*
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*DEFINE_CURVE
$ LCID SIDR SCLA SCLO
4 0
$ A1 O1
0.0000E+00 0.3282E+03
0.1000E-01 0.4245E+03
0.3000E-01 0.5061E+03
0.5000E-01 0.5492E+03
0.7000E-01 0.5796E+03
0.9000E-01 0.6034E+03
0.1100E+00 0.6231E+03
0.1300E+00 0.6400E+03
0.1500E+00 0.6548E+03
0.1700E+00 0.6680E+03
0.1900E+00 0.6800E+03
0.2100E+00 0.6910E+03
0.2300E+00 0.7011E+03
0.2500E+00 0.7105E+03
0.2700E+00 0.7194E+03
0.2900E+00 0.7276E+03
0.9990E+01 0.7316E+02
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$-----------LOAD CURVE DATA FOR BINDER TRAVEL------*
$--------1---------2---------3---------4---------5---------6---------7---------8
*DEFINE_CURVE
$ LCID SIDR SCLA SCLO
1 0
$ A1 O1
0.0 0.0
1.0E-03 1000.0
5.0E-03 1000.0
$--------1---------2---------3---------4---------5---------6---------7---------8
*BOUNDARY_PRESCRIBED_MOTION_NODE
$ NID DOF VAD LCID SF VID DEATH
301 1 0 1 1.0 0.0050
302 1 0 1 1.0 0.0050
303 1 0 1 1.0 0.0050
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
$ NODE X Y Z TC RC
1 0.976562500E-03 0.100019989E+02 0.000000000E+00 6
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*ELEMENT_SHELL
$ EID PID N1 N2 N3 N4
1 1 1 2 3 4
*END
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #7: Beam and GNL

Beam in bending

Objective:

Get familiar with some of the syntax associated with implicit analysis.
Learn about warning messages.
Learn to perform nonlinear geometrical analysis.
Learn to the effect of NLSOLVR=1 or 2 in *CONTROL_IMPLICIT_SOLUTION.

Introduction:

This workshop models a simple cantilever beam. The beam is 100 mm long and 2 mm by 2 mm
cross-section. The material is elastic with steel properties.

Input Filename: beam.k

Exercise:

Open the file beam.k with your favorite text editor. Understand the content of the file. Run the file
and observe the warning message. Fix the problem by placing SPC on the reference node (node 35)
for the beam. Generate the nodal force and displacement plots using Taurus. Record the end
displacement and the fixed end nodal forces. Run the problem with the given parameters below and
fill in the table (please note that the termination time is 1.0, however, the time in the *define curve
is 2.0).
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Tip deflection Nodal force at both end
NLSOLVR=2, tip load=10 N
Set dt=1.0, NLSOLVER=1, 25N
Exact solution for load=25 N

Q-1:
Why there is a difference between exact solution and LSDYNA/IMPLICIT?

Q-2:
Is equilibrium satisfied? What happens when NLSOLVR is equal to one?

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
GNL BEAM
$
$ This file is to demonstrate beam element and GNL
$ 100 mm long beam with 2x2 mm rectangular cross-section
$ A. Tabiei,
$
$ mm, s, N, ton
$
$ ENDTIM
1.000E+00
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*DATABASE_MATSUM
0.100E-05
*DATABASE_NODOUT
0.100E-05
*DATABASE_GLSTAT
0.100E-05
*DATABASE_NODFOR
0.100E-05
*DATABASE_BNDOUT
0.100E-05
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ DT/CYCL
0.100E-03
*DATABASE_HISTORY_NODE
$ id1
11
*DATABASE_nodal_force_group
5
*SET_NODE_LIST
5
1 11
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1 0.01 0 0 0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
2 0 0 0.000 0.00 0 0
$ dnorm divflag inistif nlprt
0 0 0 0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*PART
BEAM
$ PID SID MID
3 4 2
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*MAT_ELASTIC
$ MID RO E PR
2 7.830E-09 2.070E+05 2.800E-01
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SECTION_BEAM
$ SID ELFORM QR CST
4 1
$ TS1 TS2 TT1 TT2
2.0 2.0 2.0 2.0
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*LOAD_NODE_POINT
$ NODE DOF LCID SF
11 1 1 1.0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*DEFINE_CURVE
1
0.00000000000000E+00 0.0000000000000E+00
2.00000000000000E+00 20.000000000000E+00
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
$ NODE X Y Z
1 0.500000000E+02 0.500000000E+02 0.000000000E+00 7 7
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*ELEMENT_BEAM
$ EID PID N1 N2 N3
1 3 1 2 35
*END
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #8: Axial Buckling of Cylindrical Shell

Objective:

Get familiar with some of the syntax associated with implicit analysis.
Learn to perform stability analysis.
Bucking eigen-value analysis versus post-buckling analysis.

Introduction:

This workshop models an imperfect cylindrical shell. The initial geometric imperfections where
obtained from performing eigen-value analysis and superimposing fractions of the eigen- modes on
the perfect shell. The material is elastic. Stability problems in general can be modeled with implicit
and some times with explicit. This example will show how to do that.

Input Filename: cylind-buckl.k

Exercise:

Open the file cylind-buckl.k and understand its content. The problem is set up to perform an
implicit buckling analysis.

1. Is the problem load control or displacement control ?____________________
2. Will the problem perform post-buckling or eigen-value buckling analysis?____________

Run the problem and post process. Look at the resultant nodal forces using the ASCII file
NODFOR.

3. What is the limit point load? ___________

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Now change the file to perform eigen-value buckling analysis. You can do this by *CONTROL
IMPLICIT BUCKL. Need to set the time step to zero. Also need to change to load control by
putting $ in front the *BOUNDARY_PRE card. Run the file and compare the resultant
nodal forces. Load D3EIGV and animate the eigenvectors.
Now add *CONSTRAINED_NODE_SET to set 5 in the x-direction. Run the file and compare the
resultant nodal forces. Load D3EIGV and animate the eigenvectors.

Q-1: why the eigen-value buckling load is different than the limit point from the post buckling
analysis?

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

*KEYWORD
*TITLE
Post Buckling Analysis
$ Imperfect shell
$ thickness=2.0 mm, length=101 mm, diamter=202 mm.
$ By A. Tabiei, March 1, 1999
$ units: mm, s, ton, N
0.10000
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
0.01
*DATABASE_BINARY_D3THDT
0.01
*DATABASE_GLSTAT
1.0e-5
*DATABASE_ELOUT
1.0e-5
*DATABASE_NODFOR
1.0e-5
*DATABASE_NODOUT
1.0e-5
*DATABASE_SECFORC
1.0e-5
1
21 356 347
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*control_implicit_general
$ IMFLAG DT0
1 0.01
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*MAT_elastic
$ MID RO E PR
1 7.890E-09 2.100E+05 3.000E-01
*SECTION_SHELL
1 6
2.0000000 2.0000000 2.0000000 2.0000000
*PART
MATERIAL FOR PART : P1
1 1 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SET_NODE_LIST
3
1 2 3 4 5 6 7 8
9 10 11 12 13 14 15 16
17 18 19 20 21 233 234 235
236 237 238 239 240 241 242 243
244 245 246 247 248 249 250 251
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*DATABASE_CROSS_SECTION_SET
1 0 0 2 0 0 0 0
*SET_NODES_LIST
1
211 212 213 214 215 216 217 218
219 220 221 222 223 224 225 226
227 228 229 230 231 443 444 445
446 447 448 449 450 451 452 453
454 455 456 457 458 459 460 461
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*SET_SHELL_LIST
2
181 182 183 184 185 186 187 188
189 190 191 192 193 194 195 196
197 198 199 200 381 382 383 384
385 386 387 388 389 390 391 392
393 394 395 396 397 398 399 400
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*LOAD_NODE_POINT
5,1,1
*DEFINE_CURVE
1
0.00000000E+00 0.00000000E+00
0.10000000E+00 5.00000000E+00
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*BOUNDARY_PRESCRIBED_MOTION_SET
3 1 2 1 -1.00
*CONSTRAINED_NODE_SET
5 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
1 1.015973969E+02-8.854999542E+00-1.012129974E+02 5 0
*ELEMENT_SHELL
1 1 1 22 23 2
*END
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________


Problem #9: Implicit/Explicit Analysis
_________________________________________________________________________________

Objective:
Learn to run a combined explicit/implicit simulation.

Problem Description:
A steel plate is modeled with solid elements. The plate is constrained on all edges and loaded by
uniform pressure. The uniform pressure loading is to be analyzed using an implicit step. Then the
loaded plate is to be impacted by half a sphere. The impact scenario is to be analyzed using an
explicit step.

Input Filename: im-ex-plate.k
im-ex-ball.k
Procedure:
Copy the input files to your local directory. Using an editor, view the input file and answer the

1. How is load applied to the plate? _________________________________________
2. Is the problem running implicit step? If not fix it to run an implicit step.

3. What kind of element formulation is used for the solid element? ________________

Add the keywords *INTERFACE_SPRINGBACK_DYNA3D for the plate part (you do not need
to add constrains). This control card will force LSDYNA to generate an ascii file called dynain at
the end of the simulation. This file contains the internal stresses at the element integration points
and the deformed geometry at the termination time.
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Run the simulation.

4. What is the maximum effective stress (Von Misses) at the last state? ____________

5. Now examine the dynain. Can you explain what is in it?_____________________

Open the file im-ex-ball.k with your favorite text editor. This file contains both the nodal point for
the plate and half sphere. Need to cut the initial stresses and the new deformed nodal coordinate
from dynain file and put in this file (in-ex-ball.k).

6. What kind of load is applied in this file, and what is the load curve? _____________

7. What is the given initial velocity for the sphere? _____________________________

Comment the initial velocity card and run the simulation. Do an animation of the effect stress.

8. How is the effective stress change as a function of time, is it constant? ___________

Now you can apply the initial stress and run the problem. Post processes.

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
PLATE UNDER PRESSURE IMPLICIT
$ Units: m, s, kg, N
$ A. Tabiei
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*control_implicit_dynamics
$ IMASS GAMA BETA
1 0.5 0.25
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ IMFLAG DT0
1 0.0001
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*control_implicit_solution
$ NSOLVR ILIMIT MAXREF DCTOL ECTOL LSTOL
2 0.001 0.01 0.9
$ DNORM DIVERG ISTIF LPRINT
2 2 1 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*control_implicit_auto
$ IAUTO ITEOPT ITEWIN DTMIN DTMAX
1 100 1.0E-6 1.0e-3
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*INTERFACE_SPRINGBACK_DYNA3D
123
*SET_PART_LIST
123
1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ ENDTIM
1.000E+00
*DATABASE_GLSTAT
0.001E+00
$ DT/CYCL
0.100E+00
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*CONTROL_ACCURACY
1,1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*PART
PART PID = 1
$ PID SID MID
1 1 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ MID RO E PR SIG ETAN FAIL
1 7.890E+03 2.100E+11 3.000E-01 3.150E+08 0.000E+00 0.000
$ C P LCSS
0.0 0 0
$ (EPS) effective plastic strain vs efftive stress (ES)
0.0 0.019 0.05 0.165 0.33 0.495 0.625 0.0
315E+06 315E+06 427.8E+06 500.8E+06 504.3E+06 506.5E+06 400E+06 0.0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SECTION_SOLID
$ SID ELFORM
1 2
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
1-2.500000000E+00-2.500000000E+00 1.000000000E+00 7 7
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*ELEMENT_SOLID
$ EID PID N1 N2 N3 N4 N5 N6 N7 N8
1 1 1 33 35 3 2 34 36 4
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*DEFINE_CURVE
1
0.00000000E+00 0.00000000E+00
1.00000000E+00 1.00000000E+08
2.00000000E+00 1.00000000E+08
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*LOAD_SEGMENT
1 1 33 35 3
*END

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
half ball impact problem using explicit/implicit formulation
$ 1.414 m radius half ball impacting plate 5x5x1 m (Steel).
$ Units: m, s, kg, N
$ A. Tabiei
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
0.01
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*CONTROL_ENERGY
$ HGEN RWEN SLNTEN RYLEN
2 1 2 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1.0000E-02
*DATABASE_RCFORC
1.0000E-06
*DATABASE_GLSTAT
1.0000E-06
*DATABASE_NODOUT
0.100E-05
*DATABASE_ELOUT
0.100E-05
*DATABASE_HISTORY_NODE_SET
$ id1
10
*SET_NODE
10
240,242,272,274
*DATABASE_HISTORY_SOLID
$ id1
113
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ DT/CYCL
0.100E-02
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*CONTROL_ACCURACY
1,1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1 7.890E+03 2.100E+11 3.000E-01 3.150E+08 0.000E+00 0.000
$ C P LCSS
0.0 0 0
0.0 0.019 0.05 0.165 0.33 0.495 0.625 0.0
315E+06 315E+06 427.8E+06 500.8E+06 504.3E+06 506.5E+06 400E+06 0.0
2 7.890E+03 2.100E+11 3.000E-01 3.150E+08 0.000E+00 0.000
$ C P LCSS
0.0 0 0
0.0 0.019 0.05 0.165 0.33 0.495 0.625 0.0
315E+06 315E+06 427.8E+06 500.8E+06 504.3E+06 506.5E+06 400E+06 0.0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*PART
elastic plate
1 1 1
*PART
elastic ball
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
2 2 2
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SECTION_SOLID
1 2
*SECTION_SOLID
2 2
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*INITIAL_VELOCITY
1
0,0,250
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SET_NODE_LIST
1
515 518 521 524 527 530 533 536
539 540 541 542 543 544 545 546
547 548 549 550 551 552 553 554
555 556 557 560 563 566 569 572
575 576 577 578 579 580 581 582
583 584 585 586 587 590 593 596
599 602 605 606 607 608 609 610
611 612 613 614 615 616 617 631
633 635 637 638 639 640 641 642
643 644 645 646 647 648 649 650
651 652 653 667 669 671 673 674
675 676 677 678 679 680 681 682
683 684 685 686 687 688 689 691
693 695 697 698 699 700 701 704
707 710 713 716 719 720 721 722
723 724 725 726 727 728 729 730
731 745 747 749 751 752 753 754
755 756 757 758 759 760 761 762
763 764 765 766 767 781 783 785
787 788 789 790 791 792 793 794
795 796 797 798 799 800 801 802
803 805 807 809 811 812 813 814
815 817 819 821 823 824 825 826
827 829 831 832 833
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
$ NODE X Y Z TC RC
515-0.707106829E+00-0.707106829E+00-0.460000396E+00
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*ELEMENT_SOLID
1 1 1 33 35 3 2 34 36 4
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*CONTACT_SURFACE_TO_SURFACE_TITLE
$ CID IF. NAME
1 IF000001
$ SSID MSID SSTYP MSTYP SBOXID MBOXID SPR MPR
1 2 0 0 0 0 0 0
$ FS FD DC V VDC PENCHK BT DT
0.000E+00 0.000E+00 0.000E+00 0.000E+00 0.000E+00 0 0.000E+00 0.000E+00
$ SFS SFM SST MST SFST SFMT FSF VSF
0.000E+00 0.000E+00 0.000E+00 0.000E+00 0.000E+00 0.000E+00 0.100E+01 0.100E+01
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SET_SEGMENT
1
515 540 542 518
833 815 803 801
*SET_SEGMENT
2
1 3 35 33
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
1-0.250000000E+01-0.250000000E+01 0.100000000E+01 7 7
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*INITIAL_STRESS_SOLID
1 8
1.264E+07 1.264E+07 1.545E+08-1.433E+08-5.396E+07-5.396E+07 1.085E-04
9.326E+07-5.625E+07 1.428E+08-1.291E+08-7.521E+07-4.939E+04 2.427E-03
1.194E+07 1.194E+07 1.559E+08-1.617E+08 3.029E+05 3.029E+05 3.449E-03
-5.625E+07 9.326E+07 1.428E+08-1.291E+08-4.939E+04-7.521E+07 2.427E-03
1.083E+08 1.083E+08-3.672E+07 1.437E+08-5.201E+07-5.201E+07 3.402E-04
3.096E+07 1.755E+08-2.662E+07 1.255E+08-8.016E+07-5.402E+06 3.038E-03
1.156E+08 1.156E+08-5.140E+07 1.519E+08-1.755E+07-1.755E+07 4.352E-03
1.755E+08 3.096E+07-2.662E+07 1.255E+08-5.402E+06-8.016E+07 3.038E-03
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*DEFINE_CURVE
1
0.00000000E+00 1.00000000E+08
2.00000000E+00 1.00000000E+08
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*LOAD_SEGMENT
1 1 33 35 3
*END
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________


Problem #10: Bolt Analysis
_________________________________________________________________________________

Objective:
Learn to rigid body, prescribed motion, and equilibrium check.

A steel beam section is bolted as shown. The bolt in modeled by both rigid and deformable
material. The bolt is given prescribed displacement so that the bolt is pulled out of the slotted hole.

Input Filename: bolt.k

Procedure:
following question:

1. What is the allowed direction for motion of the rigid material (*mat_rigid)? _______
(hint: look at the third card of *mat_rigid).

2. Put the *boundary_prescribed_motion_rigid with the approperiate parameters. Use the defined
curve for the motion.
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
Run the simulation and look at the contact forces.

3. Switch the material of the bolt to deformable. You can do that by changing the material ID of
the deformable material. Put the *boundary_prescribed_motion_set with the appropriate
parameters. Use the defined curve and the defined node set for the motion (do not forget to
remove the *boundary_prescribed_motion_rigid from the file).

Run the simulation and look at the contact forces.

4. Plot the x-component of contact force. Plot sum of the nodal internal forces of the nodal set
specified (hint: look at the NODFOR).

5. Is equilibrium satisfied? ________________________________________________

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
Bolt Analysis By A. Tabiei
0.0100000
*CONTROL_CONTACT
$ SLSFAC RWPNAL ISLCHK SHLTHK
0.100 2
$ USRSTR USRFAC NSBCS INTERM

*CONTROL_ENERGY
2
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ IMASS GAMA BETA
0 0.5 0.25
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ IMFLAG DT0
1 0.00001
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
2 20 5 0.001 0.01 0.9
2 2 1 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1 100 1.0E-7 1.0e-4
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1.00000-3
*DATABASE_NODFOR
1.00000-5
*DATABASE_SLEOUT
1.00000-3
*DATABASE_RCFORC
1.00000-5
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1 7.83000-9 2.00000+5 0.3300000 345.00000 0.0000000 0.0000000 0.0000000
0.0000000 0.0000000 0 0
0.0000000 0.0200000 0.0300000 0.1500000 0.3000000 0.4000000 10.000000 0.0000000
345.00000 515.00000 600.00000 690.00000 740.00000 760.00000 700.00000 0.0000000
*MAT_RIGID
22 7.85000-9 2.10000+5 0.3000000 0.0000000 0.0000000 0.0000000
1.0000000 6.0000000 7.0000000
0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000
2 7.83000-9 2.20000+5 0.3300000 445.00000 0.0000000 0.0000000 0.0000000
0.0000000 0.0000000 0 0
0.0000000 0.0200000 0.0300000 0.1500000 0.3000000 0.4000000 10.000000 0.0000000
445.00000 615.00000 700.00000 790.00000 840.00000 860.00000 800.00000 0.0000000
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SECTION_SHELL
1 16 0.0000000 0.0000000 0.0000000 0.0000000 0
2.0000000 2.0000000 2.0000000 2.0000000 0.0000000
*SECTION_SHELL
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
2 16 0.0000000 0.0000000 0.0000000 0.0000000 0
2.0000000 2.0000000 2.0000000 2.0000000 0.0000000
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*PART
Mat.-Type: 24
27 1 1
*PART
Mat.-Type: 20 ** Rigid **
28 2 2
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*DEFINE_CURVE
1 0
0.00000000E+00 0.00000000E+00
0.01000000E+00 1.50000000E+01
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$*BOUNDARY_PRESCRIBED_MOTION_rigid
$ 28 2 2 1 1.0000000 0 0.0000000 0.0000000
1 2 2 1 1.0000000 0 0.0000000 0.0000000
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1
*SET_NODE_LIST
1
810 811 812 813 814 815 816 817
818 819 820 821 822 823 824 825
826 827 828
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
$ NODE X Y Z TC RC
1 0.406250000E+02-0.762940005E-07 0.240000000E+02
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*ELEMENT_SHELL
1 27 220 278 372 280
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
2 3 0 0 0 0 0 0
0.2200000 0.2000000 0.0000000 0.0000000 0.0000000 0 0.0000000 0.0000000
1.0000000 1.0000000 0.0000000 0.0000000 0.0000000 0.0000000
$ optional card A
$ soft
0
$ optional card B
$ penmax thkopt shlthk snlog
0.0 0 0 1
*SET_SEGMENT
$ Slave segments, surface: 1
$
2
220 278 372 280
280 372 284 221
*SET_SEGMENT
$ Master segments, surface: 1
$
3
297 83 424 432
291 297 432 434
449 448 749 750
*END

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________


Problem #11: Crash Analysis
_________________________________________________________________________________

Objective:
Learn to difference between explicit and dynamic implicit.

A steel front rail is constrained on one end and displacement control is applied on the other end.
The analysis to be run using explicit and implicit formulation.

Input Filename: crash-implicit.k

Procedure:

following question:

1. There are two boundary conditions specified.
What are these boundary conditions ? ________________________________________

2. How is motion prescribed and what out put is requested for post processing? ______

Run the simulation and post processes.

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
3. Plot reaction forces.

Change the simulation to an explicit simulation and run it.

4. Plot reaction forces. Is there a difference in the forces between the two runs._______

5. Comment on the simulations. ____________________________________________

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
*TITLE
Symmetric Short Crush Tube
$
$ Symmetric model - 1/4 of the tube
$
$ Units: mm, kg, ms, kN, GPa, kN-mm
$
$
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$
$$$$ Control Ouput
$
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$
$
$...>....1....>....2....>....3....>....4....>....5....>....6....>....7....>....8
$ f i f f f
$ endtim endcyc dtmin endneg endmas
1.00 0 0.0 0.0 0.0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ IMASS GAMA BETA
1 0.5 0.25
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ IMFLAG DT0
1 0.0001
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
2 20 5 0.005 0.01 0.9
2 2 1 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1 10 1.0E-6 1.0e-4
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$
*SET_NODE_LIST
$ SID DA1 DA2 DA3 DA4
1
$ NID1 NID2 NID3 NID4 NID5 NID6 NID7 NID8
99999 253 254 255 256 257 258 259
486 487 712 713 714 715 716
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SET_NODE_LIST
2
1 2 3 4 5 6 7 414
415 532 533 534 535 536
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*BOUNDARY_SPC_SET
1,0,1,1,0,0,0,0
1,3,2,99,-1.0
*DEFINE_CURVE
99
0.0,0.0
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
1.00,50.0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*CONTROL_ENERGY
$ i i i i
$ hgen rwen slnten rylen
2 2
$
$ f i
$ DT/CYCL LCDT - time interval between state dumps (D3PLOT)
0.1
$
$ i i i i i i i i

$ i i i
1
$
*DATABASE_BINARY_D3THDT
$ DT/CYCL LCDT - time interval between time history data dumps (D3THDT)
999999
$
2
*DATABASE_NODFOR
$ DT/CYCL
0.0001
$
$ Define nodes that output into NODOUT
$...>....1....>....2....>....3....>....4....>....5....>....6....>....7....>....8
$ id1 id2 id4 id4 id5 id6 id7 id8
99999 414 486
$
*DATABASE_NODOUT
$ DT/CYCL
0.001
*DATABASE_GLSTAT
$ DT/CYCL
0.001
*DATABASE_MATSUM
$ DT/CYCL
0.001
*DATABASE_SLEOUT
$ DT/CYCL
0.001
$
$
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$
$$$$ Define Contacts - Sliding Interfaces
$
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$
$
*CONTACT_AUTOMATIC_SINGLE_SURFACE
$...>....1....>....2....>....3....>....4....>....5....>....6....>....7....>....8
$ ssid msid sstyp mstyp sboxid mboxid spr mpr
0
$ Equating ssid to zero means that all segments are included in the contact
$
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
$ fs fd dc vc vdc penchk bt dt
0.08 0.08
$
$ sfs sfm sst mst sfst sfmt fsf vsf

$
$
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$
$$$$ Define Materials and Parts
$
$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
$
$
$...>....1....>....2....>....3....>....4....>....5....>....6....>....7....>....8
$ mid ro e pr sigy etan eppf tdel
1 7.830E-06 200.0 0.3 0.207 10.0
$
$ c p lcss lcsr
0 0
$ PLASTIC STRESS/STRAIN CURVES
$ eps1 eps2 eps3 eps4 eps5 eps6 eps7 eps8
$ es1 es2 es3 es4 es5 es6 es7 es8
0.000E+00 8.000E-02 1.600E-01 4.000E-01 9.900E+01
2.070E-01 2.500E-01 2.750E-01 2.899E-01 3.000E-01
$
*SECTION_SHELL
$ SHELL ELEMENT CROSS-SECTIONAL PROPERTIES
$...>....1....>....2....>....3....>....4....>....5....>....6....>....7....>....8
$ sid elform shrf nip propt qr/irid icomp
1 16 3.0000
$ t1 t2 t3 t4 nloc
2.000E+00 2.000E+00 2.000E+00 2.000E+00
$
*PART
$ HEADING
corner1
$...>....1....>....2....>....3....>....4....>....5....>....6....>....7....>....8
$ pid sid mid eosid hgid igrav adpopt
1 1 1
$
*NODE
$
$...>....1....>....2....>....3....>....4....>....5....>....6....>....7....>....8
$ NID X Y Z TC RC
99999 0.0 0.0 274.0 0 0
*ELEMENT_SHELL
$...>....1....>....2....>....3....>....4....>....5....>....6....>....7....>....8
752 1 547 552 553 553
*BOUNDARY_SPC_NODE
$...>....1....>....2....>....3....>....4....>....5....>....6....>....7....>....8
$ NID CID X Y Z RX RY RZ
1, 0,1,1,1, 1, 1, 1
259, 0,1,0,0, 0, 1, 1
$
$
*END
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________


Problem #12: Bumper Analysis
_________________________________________________________________________________

Objective:
Learn rigid body, prescribed motion, difference in implicit contact (auto vs non-auto).

A bumper is given prescribed motion to impact a rigid plate as shown. Two contact surfaces can be
defined. The non-automatic and the automatic contact can be defined. The non-automatic is
defined by segments sets and the automatic by part set.

Input Filename: bumper-1.k

Procedure:


1. What contact surface is used and how slave and master is defined. _______________

2. What is the penalty scale factor for the contact. ______________________________

3. Run the simulation save the contact forces.
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

4. Reduce the scale factor to 0.001 and run the problem. Save the contact force.

5. Change the contact to AUTOMATIC_SINGLE_SURFACE and define the slave and master by
parts (part 1 and part 4). You can find it at the bottom of the input file. Run it and post process.
Save the contact forces.

6. Now reduce the initial time step to 0.001 and run the problem. Save the contact force.

7. To make contact forces the most accurate set DNORM=1 on the IMPLICIT_SOLUTION card.
Run the problem and post process.
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
*KEYWORD
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*TITLE
BARRIER-BUMPER IMPACT
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ ENDTIM
1.000E+00
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ IMASS GAMA BETA
0 0.5 0.25
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ IMFLAG DT0
1 0.0001
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
2 20 5 0.005 0.01 0.9
2 2 1 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1 100 1.0E-7 1.0e-3
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*CONTROL_CONTACT
$ SLSFAC RWPNAL ISLCHK SHLTHK THKCHG ORIEN
0.100 2 2

$
*CONTROL_ENERGY
$ HGEN RWEN SLNTEN RYLEN
2 2 1 1
*DATABASE_GLSTAT
.100E-02
*DATABASE_RCFORC
.100E-02
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ DT/CYCL
.100E-00
0 0 3 0 1 1 1 1
0 0 0 0 0 0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*PART
$HEADING
PART PID = 1 PART NAME :BUMP.S
$ PID SID MID
1 1 2
*PART
$HEADING
PART PID = 2 PART NAME :RAILS.H
$ PID SID MID
2 1 2
*PART
$HEADING
PART PID = 3 PART NAME :BRACKS.H
$ PID SID MID
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
3 3 2
*PART
$HEADING
PART PID = 4 PART NAME :WALL.LIN
$ PID SID MID
4 1 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
$ MID RO E PR SIGY ETAN EPPF TDEL
2 7.830E-09 2.070E+05 2.800E-01 2.100E+02 0.000E+00 0.000E+00 0.000E+00
$ C P LCSS LCSR
.000E+00 .000E+00 .000E+00 .000E+00
$ EPS1 EPS2 EPS3 EPS4 EPS5 EPS6 EPS7 EPS8
0.000E+00 3.090E-02 4.090E-02 5.000E-02 1.510E-01 3.010E-01 7.010E-01 9.100E-01
$ ES1 ES2 ES3 ES4 ES5 ES6 ES7 ES8
2.100E+02 3.000E+02 3.140E+02 3.250E+02 3.900E+02 4.380E+02 5.050E+02 5.270E+02
*MAT_RIGID
$ MID RO E PR N COUPLE M ALIAS
1 7.830E-09 2.070E+05 2.800E-01 0.000E+00 0.000E+00 0.000E+00
$ CMO CON1 CON2
1.0 7.0 7.0
$LCO or A1 A2 A3 V1 V2 V3

$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SECTION_SHELL
$ SID ELFORM SHRF NIP PROPT QR/IRID ICOMP
1 16 .100E+01 3.0 1.0 .0
$ T1 T2 T3 T4 NLOC
1.000E+00 1.000E+00 1.000E+00 1.000E+00
*SECTION_SHELL
$ SID ELFORM SHRF NIP PROPT QR/IRID ICOMP
3 16 .100E+01 3.0 1.0 .0
$ T1 T2 T3 T4 NLOC
1.000E+00 1.000E+00 1.000E+00 1.000E+00
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*SET_NODE_LIST
$ SID
1
253 254 255 267 289 290 294 457
458 472 490 500
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*BOUNDARY_SPC_SET
1,0,1,0,1,0,0,0
1,2,2,99,-1.0
*DEFINE_CURVE
99
0.0,0.0
1.00,50.0
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*NODE
$ NODE X Y Z TC RC
1-0.500000000E+03-0.152500000E+03 0.313752014E+03
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*ELEMENT_SHELL
1 4 1 10 11 2
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
1 2 0 0 0 0
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
.100E+00 .000E+00 .000E+00 .000E+00 .000E+00 0 .000E+00 .000E+00
.000E+00 .000E+00 .000E+00 .000E+00 .000E+00 .000E+00 .100E+01 .100E+01
$ optional card A
$ soft
0
$ optional card B
0.0 0 0 1
*SET_SEGMENT
$ SID DA1 DA2 DA3
1
$ N1 N2 N3 N4
100 101 344 115
337 334 329 342
*SET_SEGMENT
2
$ N1 N2 N3 N4 A1 A2 A3 A4
1 2 11 10
89 90 99 98
$*CONTACT_AUTOMATIC_SINGLE_SURFACE
$ CID CONTACT INTERFACE TITLE
$ 2 IF02
$ 1 2 3 3 0 0
$ .100E+00 .000E+00 .000E+00 .000E+00 .000E+00 0 .000E+00 .000E+00
$ .000E+00 .000E+00 .000E+00 .000E+00 .000E+00 .000E+00 .100E+01 .100E+01
$ optional card A
$ soft
$ 0
$ optional card B
$ 0.0 0 0 1
$---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8
*END

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #13: Eigen-Value Analysis
_________________________________________________________________________________

Objective:
Learn to how to perform eigenvalue analysis.

A spring- mass system as shown. There are five springs and four concentrated masses. Closed form
solution is available.

Input Filename: springs.k

Procedure:


1. What material is specified for the sprint, what is the stiffness of the springs. ____
2. What if the concentrated masses at the end of each spring __________________
3. Need to perform eigen value analysis, how many eigen values this system has __
4. Perform the eigen value analysis and record the eigen values.

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #14: Eigen-Value Analysis
_________________________________________________________________________________

Objective:
Learn to how to perform eigen-value analysis.
Learn how to post process eigen-values and eigen-vectors.

A shallow cap problem as shown. The objective is to extract 10 frequencies and the corresponding
vibration modes..

Input Filename: shalowcap-eigen.k

Procedure:


1. What is the element formulation used in the model. _______________________
2. Need to perform eigen value analysis, add the necessary cards to extract 10 eigen values
this system has. Run and examine eigout. What many eigen values are near zero? _____
3. Now add constrains (SPC) to node set 1. Run and edit eigout. How many eigen values are
zeros? ______________.
4. Post process d3eigv using lspost. Animate. Now change the No. of div. to 2 near the +
button. Animate. What is lspost showing now?_______________________________.
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #15: Implicit-Explicit Switching
_________________________________________________________________________________

Objective:
Learn to use the implicit method for static initialization of an explicit simulation, learn to
switch formulation during a simulation, and learn to perform a simple eigenvalue analysis..

A cantilever strip of shell elements is loaded using the static implicit method. The analysis type is
then switched to explicit, the load is removed, and the dynamic response is simulated. The first
fundamental frequency is verified by eigenvalue analysis.

Input Filename: ie_shell.k

Procedure:

1. How is the implicit-explicit switching activated. __________________________
2. When dose the simulation run implicit?, explicit? _________________________
3. How many implicit steps will be taken. _________________________________
4. How is load applied? When it is removed? ______________________________
5. Run the simulation, and post process the results. Plot the y-displacement of a tip node vs.
time. Estimate the period of response for the first fundamental mode:. _________
WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________
6. Modify the input deck to conduct an eigen value analysis. Solve for the lowest five
eigenvalues. Which new output files are created? _________________________
7. View the eigout file, what is the period of the first fundamental mode? ________
8. Postprocess d3eigv, view the mode shapes. What is the meaning of the time associated
with each mode shape? _____________________________________________ .
9. Now learn to do many switches. Modify the original input deck to switch 4 times from
implicit to explicit.

WORKSHOP
_______________________________________________________________________________________________
_______________________________________________________________________________________________

Problem #16: Arc-Length Method
_________________________________________________________________________________

Objective:
Learn to how to perform post-buckling analysis using arclength method.
Learn how to perform snap-through post buckling.

A shallow cap problem as shown. The objective is to simulate the snap through phenomenon.

Input Filename: shallow-cap.k

Procedure:


1. What is the nonlinear solver used in the model.___________________________
2. Add the necessary cards to perform arc length method.
3. Look at the center node deflection.

Implicit Newer

Caricato da

Informazioni sul documento

Descrizione originale:

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Implicit Newer

Caricato da

Copyright:

Formati disponibili

LSDYNA IMPLICIT NOTES

is the internal forces vector

& & &

is the tangent damping matrix

is the tangent stiffness matrix.

Potrebbero piacerti anche