Stress Initialization in LS‐DYNA
A short overview
/SURI.BALA/LSTC
Introduction
Many problems today involve initialization of structures due to pre‐load
Some examples include the following (not to scale)
Bolt PreStress Gravity Initialization Interference
How can we simulate this in LS‐DYNA
LS‐DYNA is a general purpose finite element program and hence provides
several methods to perform stress‐initialization in any simulation
This document will review all available methods to perform stress‐initialization
followed by transient event
Available Methods to Simulate Pre‐Loads
1. Restart Analysis
In this method, the stress‐initialization is performed first and the results at the
end of this method is saved in a ‘restart’ file used later (using r=restart_file)
by the next simulation.
2. Sequential
In this method, all pre‐loads are quasi‐statically applied while the transient loads
are zero. After the quasi‐static loads reach their maximum values, the
transient loads are invoked
3. Dynamic Relaxation
In this method a “pseudo” analysis is performed quasi‐statically using either
IMLPICIT or Explicit method
4. Dependent Case Definitions
*CASE keyword allows to perform dependent or independent multi‐stage events.
Dependent simulations will use the results (stresses, geometry and velocity)
from earlier case using simple or full restart
1. Restart Analysis
In this method, the pre‐load is simulated by a “first” run.
Node and Element history variables (displacements, stresses) are
stored in a binary file (runrsf or d3dump) from the last state.
In the second run, the model is then mapped with the stored history
variables providing the initial pre‐stressed state.
The second run could contain the exact model as the first (simple
restart) or can be very different (full restart). Full restarts will require
the use of *STRESS_INITIALIZATION keyword in the second input file.
First Run Second Run
lsdyna i=first.k Lsdyna i=second.k r=runrsf
runrsf/d3dump
2. Sequential
In this method, the pre‐loads and the transient loads are applied to a
single model. During the pre‐load, all transient loads are set to zero.
After the pre‐loads reach their maximum value, the transient loads
are then ramped
Transient loads involving INITIAL_VELOCITY will require special
treatment such as using *BOUNDARY_PRESCRIBED_MOTION to
enforce zero velocity until the pre‐loads reach maximum value. This
boundary definition is then “killed” using a nonzero DEATH time
transient load
preload
End of pre‐load Simulation Time
(death time for prescribed motion)
3. Dynamic Relaxation (DR)
Dynamic relaxation (DR) is a solution method originally implemented to
perform quasi‐static simulation in a “psuedo” time prior to transient analysis.
Stresses and the geometry at the end of this “psuedo” phase are
automatically used in the transient analysis providing the desired “stress‐
initialization”
Stress Intialization Transient Analysis
Pseudo Phase Transient Phase
D3DRLF D3PLOT
Dynamic Relaxation Supplemental Information
DR can be run either in Explicit mode or in Implicit
Mode. Both methods will use independent
convergence criteria to determine if a quasi‐static
response has been attained. User can also terminate
the DR phase which is mandatory when using Implicit
solver
Distortional Energy During the explicit DR phase, the ratio of the current to
Current peak distortional energy (total energy minus kinetic
energy due to rigid body motion) is monitored . When
the ratio is less than a certain user‐defined value, the
Peak DR phase is assumed to be converged and is
terminated
Graphical data during the DR phase can be requested
using *DATABASE_BINARY_D3DRLF which will be
output to D3DRLF file. This file is written in the same
format as D3PLOT and can be easily read‐into any post‐
processing software
4. Case Controls
*CASE keywords allow the modeling of multiple independent or dependent
simulations in one single file. This option is ideal to simulate multi‐stage
problems. To use this functionality, it is required to use a separate code
named “LSCASEDRIVER” that comes in Python or C language.
Independent simulations are those which do not use information (history
variables) from the previous stage
Preloading information used during transient loading is a “dependent”
simulation. CASE keyword internally uses “restart method” (method 1) but
makes the invisible to the user. The user is required to use the “r=restart” in
the command line argument used by the LSCASEDRIVER.
runrsf
Stage 1 Stage 2 Stage 1 Stage 2
independent dependent
How do they compare
Setup and Supports Multiple Preload Results
Method
Post‐Process Solution Methods Reusability
Restart easy yes Yes
Sequential easy Yes No
DR easy Yes Yes
Case Controls easy Yes Yes
Getting Familiar with DR
Dynamic relaxation (DR) is activated by setting a non‐zero (1 or 2) value for
SIDR in the *DEFINE_CURVE keyword
The default value of SIDR (0) indicates that the curve is valid only for transient
analysis. SIDR = 1 indicates the curve is used in DR Phase. SIDR = 2 indicates
the curve is used in both DR and Transient Phases
There are several parameters that are available
to use DR feature
All the parameters are available in
*CONTROL_DYNAMIC_RELAXATION keyword
A good review of these parameters is essential
to use the feature as expected
*CONTROL_DYNAMIC_RELAXATION/NRCYCK
NRCYCK specifies the number of cycles to check for
distortional energy (non‐zero strain) ratio for
convergence
This parameter is not used if the DR phase is run in
IMLPICIT mode
*CONTROL_DYNAMIC_RELAXATION/DRTOL
DRTOL is the tolerance used to decide the
convergence. When ratio of current to peak
distortional energy is <= DRTOL, convergence is said
to be achieved. Distortional energy is total energy
minus kinetic energy associated with rigidbody
motion
This parameter is not used if the DR phase is run in
IMLPICIT mode. DR phase is also terminated If the
convergence is attained earlier than DRTERM
*CONTROL_DYNAMIC_RELAXATION/DRFCTR
DRFCT is the dynamic relaxation factor used to
damp the nodal velocities each timestep
This parameter is not used if the DR phase is run in
IMLPICIT mode
*CONTROL_DYNAMIC_RELAXATION/DRTERM
DRTERM is the termination time for the “pseudo”
phase. By default, the termination for DR is
determined by the convergence criteria DTOL. If
DRTERM is non‐zero, the DR phase is terminated
based on which criteria is reached first.
This parameter is used by both IMPLICIT and
EXPLICIT. It is mandatory to be non‐zero if IMLPICIT
is used in DR
*CONTROL_DYNAMIC_RELAXATION/TSSFDR
TSSFDR is the timestep scale factor applicable only
in the DR phase
This parameter is not used if IMLPICIT is used in DR
phase
*CONTROL_DYNAMIC_RELAXATION/IDRFLG
IDRFLG tells LS‐DYNA if the DR is active or not. Irrespective
of SIDR flag in DEFINE_CURVE, DR can be turned off by
setting IDRFLG=‐999. This is often a convenient way to
suppress DR if multiple curves are used in DR. If IDRFLG=‐
1, DR is active and all timehistories are output during the
DR phase.
This parameter is the only recommended way to invoke
IMPLICIT solver (=5) in the DR phase.
Few notes on Interference Fit
Interference fit between parts is best modeled
using_INTERFERENCE option to CONTACT_NODES_TO_SURFACE
or CONTACT_SURFACE_TO_SURFACE
The interference is removed gradually by scaling the contact
1
stiffness thickness ideally from zero to maximum value of unity. Scaling of
the stiffness can be performed using DR or in the transient phase
0
slave
It is best recommended to use one‐way and choosing the part
that is penetrating as the slave.
master
Interference keyword
For DR phase For Transient phase
Thank You !
suri@lstc.com
http://www.lstc.com
