Preload - Introduction
Sometimes it’s important to induce a steady state
preload before performing a transient dynamic analysis.
Rotating fan or turbine blades, rotating flywheels
Gravity
Pressure vessels or tires
Interference-fit assemblies
Stresses induced by a torqued bolt
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 1
Preloads in LS-DYNA
Preload Analysis Techniques (General)
Explicit dynamic relaxation
Normal transient analysis with mass damping
Implicit methods
2 s eparate runs: Implicit preload followed by Explicit
Implicit/Explicit s witching
Implicit by *CONTROL_DYNAMIC_RELAXATION
Specific ways to preload Bolts
Thermal load
Interference contact
Initialize stress in solid cross-section
Initialize force in beams
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 2
1
Explicit Dynamic Relaxation (DR)
Explicit Dynamic Relaxation is an optional transient analysis
that takes place in ‘pseudo-time’ (precedes normal transient
analysis).
DR is typically used to preload a model before onset of transient
loading. Preload stresses are typically elastic and displacements
are small.
In explicit DR, the computed nodal velocities are reduced each
time step by the dynamic relaxation factor (default = .995).
Thus the DR solution undergoes a form of damping.
The distortional kinetic energy is monitored. When this
energy has been sufficiently reduced, the DR phase terminates
and the solution automatically proceeds to the normal transient
analysis.
Alternatively, DR can be terminated at a preset termination
time.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 3
Explicit Dynamic Relaxation
DR is typically invoked by setting the variable SIDR to 1 or 2 in
*DEFINE_CURVE. This makes the load defined by the curve
applicable to the DR phase.
Curve guidelines
Ramp the load during the DR phase and then hold load constant until
s olution “converges”, i.e., until the dis tortional KE becomes s ufficiently
small.
Maintain the preload in s ubsequent trans ient analysis phase (us e
separate load curve without the ramp)
Time Time
Make sure DR converges
somewhere on plateau
SIDR = 1 (DR phase) SIDR = 0 (transient phase)
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 4
2
Explicit Dynamic Relaxation
*CONTROL_DYNAMIC_RELAXATION variables
NRCYCK: Iterations between convergence checks (default=250)
Als o affects output interval for binary output d3drlf
DRTOL: Convergence tolerance (default=0.001)
Ratio of dis tortional KE at convergence to peak distortional KE
Smaller value results in converged solution nearer to steady state
but DR s olution will take longer
DRFCTR: Dynamic relaxation factor (default=0.995)
Scaling factor for nodal velocities each time s tep
If value is too s mall, model may never reach s teady s tate due to
overdamping
DRTERM: Optional termination time for DR
DR will s top when time reaches DRTERM even if DR s olution is not
converged
TSSFDR: Time step scale factor used during DR; can be different
than TSSFAC used for normal transient phase.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 5
*CONTROL_DYNAMIC_RELAXATION parameters cont’d
IDRFLG
EQ.-999: DR not activated. Overrides SIDR in *DEFINE_CURVE.
EQ.0: Explicit DR not activated unless by SIDR in
*DEFINE_CURVE
EQ.1: Explicit dynamic relaxation activated with convergence test
based on distortional KE of all parts
EQ.3: Explicit dynamic relaxation activated with convergence test
based only on distortional KE of parts in part set DRPSET
EQ.2: Invokes a completely different and faster initialization
approach … Initialization by Prescribed Geometry (see next
slide)
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 6
3
*CONTROL_DYNAMIC_RELAXATION parameters cont’d
IDRFLG
EQ.2: Invokes a completely different and faster initialization
approach … Initialization by Prescribed Geometry
Requires a supplemental input file containing the nodal
displacements and rotations corresponding to the preloaded
state.
• Such a file [Link] is produced by LS-DYNA at the
conclusion of a standard DR run .
• LS-PrePost can also produce this file via Output button, but it
won’t include the nodal rotations required of beams and
shells.
• Must include “m=[Link]” on execution line.
When IDRFLG=2, LS-DYNA runs a precursor explicit analysis,
ramping linearly to the specified nodal displacements and
rotations in NC time steps (default=100 time steps).
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 7
Output Related to Dynamic Relaxation
ASCII output files are NOT written during DR phase, e.g., glstat,
matsum , rcforc , etc.
Time history data of specified nodes and elements
(*DATABASE_HISTORY_ option ) are written to binary d3thdt
(*DATABASE_BINARY_D3THDT) if IDRFLG=-1
Binary database d3drlf is written by including command
*DATABASE_BINARY_D3DRLF. If output interval set to 1, then
a plot state is written to d3drlf whenever convergence is
checked during explicit DR
d3drlf is to explicit DR phase what d3plot is to normal transient
phase
ASCII relax file, containing time histories of distortional KE and
convergence factor, is produced by default. Data can be plotted
using LS-PrePost.
At the conclusion of DR, d3dump01 and [Link] are
written. The latter contains nodal displacements and rotations.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 8
4
Output - Explicit Dynamic Relaxation
DR information is written to the screen. The
normal transient phase starts when the
convergence tolerance or a specified
termination time for DR is reached.
Convergence factor from relax file
Total Kinetic Energy from relax file
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 9
Loads during Dynamic Relaxation
Gravity loads and centrifugal loads (spinning bodies) are imposed
using *LOAD_BODY_option.
LCID and LCIDDR are separate curves for normal transient phase
and DR phas e, res pectively.
Temperatures are prescribed using *LOAD_THERMAL_option.
Parts , e.g., bolts , defined with a coefficient of thermal expans ion will
res pond to the temperature
LCID and LCIDDR in *LOAD_THERMAL_LOAD_CURVE are separate
curves of temperature for normal trans ient phase and DR phas e, res p.
Other load types or boundary conditions can also be applied during
DR if SIDR in corresponding *DEFINE_CURVE is set to 1 or 2.
Example: *LOAD_SEGMENT, *BOUNDARY_PRESCRIBED_MOTION_option.
*CONTACT_..._INTERFERENCE imposes forces associated with
geometric interference, as in a press-fit assembly.
*INITIAL_... (more on that later)
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 10
5
Explicit Dynamic Relaxation
Example – Gravity Loading on a Tire
Consider a tire with a constant gravity load. Without DR, the tire
bounces during the simulation as seen when plotting the Z-
displacement for a node on the tire rim. Now see the case with DR
on the next slide.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 11
Dynamic Relaxation
Example – Gravity Loading on a Tire (cont’d)
DR was then included with a ramped gravity curve for the DR phase, i.e., load
curve LCIDDR (*LOAD_BODY_Z) has SIDR (*DEFINE_CURVE) set to 1. The
ramp time covers approximately 2000 time steps. The
*CONTROL_DYNAMIC_RELAXATION parameters are all set to default. The
response during the normal transient phase following the DR phase is shown in
curve B below.
A No DR
B With DR
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 12
6
Dynamic Relaxation
Example – Gravity Loading on a Tire (con’td)
Three different settings of the energy convergence tolerance, DRTOL, were
then tried: 1e-3 (default), 1e-4 and 1e-6. This tolerance is the only change in
the model.
The value of DRTOL offers a tradeoff between run time and amplitude of
residual dynamic oscillation.
DRTOL 1e-3 1e-4 1e-6
Elapsed Time (sec) 3808 5032 13755
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 13
Preload during Normal Transient
Analysis
As an alternative to using DR, in some cases the preload
can be established in the early part of a normal
transient analysis.
Ramp up preload over a finite time and then hold load steady.
Use time-dependent mass damping (*DAMPING_GLOBAL with a
curve) to inhibit dynamic response during preloading.
Drop damping constant to zero after a near s teady s tate preload
s olution is es tablished and trans ient loading is ready to be applied.
Apply transient loads AFTER preload is established.
Us e *INITIAL_VELOCITY_GENERATION_START_TIME for problems
whos e trans ient res ponse is driven by initial velocityin order to
delay onset of “initial” velocity until after the preload is established.
Us e nonzero birthtime or nonzero arrival time for trans ient loads
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 14
7
Preload during Normal Transient
Analysis
Preload Transient Load
Load Load
t1 Time t2 Time
Mass Damping
Damping
Coefficient
t1 t2 Time
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 15
Preload via Implicit Analysis
In general, implicit static and quasi-static, implicit
dynamic analyses are well-suited to inducing preload. The
latter will tolerate rigid body modes and is less likely to
encounter difficulty in attaining convergence.
Implicit analysis requires the command
*CONTROL_IMPLICIT_GENERAL
Other implicit-related commands commonly used are:
*CONTROL_IMPLICIT_AUTO automatically adjusts step size based on
ease or difficulty in achieving convergence
*CONTROL_IMPLICIT_DYNAMICS is used to make the implicit solution
dynamic rather than static
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 16
8
Preload via Implicit Analysis
Approach 1: Make 2 separate runs; preload run followed
by transient run
Make an implicit run applying only the preload. Include
*INTERFACE_SPRINGBACK_LSDYNA in the input. This creates an
ASCII file called dynain when the simulation is finished. The
dynain file contains keyword commands describing the preloaded
state in terms of deformed geometry, stresses, and plastic strains.
Merge these commands into a copy of the original input deck,
deselect the implicit cards, incorporate the transient loading, and
use this deck to run a second simulation (explicit) that effectively
starts from the preloaded state.
The dynain file does not include contact forces nor does it contain
nodal velocities. Thus these quantities from the preload analysis do
not carry over to the second analysis.
Using data from the last state of the first run’s d3plot , LS-PrePost ®
can output a dynain file via Output > Format: Dynain Ascii > Write.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 17
Preload via Implicit Analysis
Approach 2: Single run employing implicit/explicit
switching
Use one input deck where switching between implicit and
explicit solvers is determined by a curve.
The abscissa of the curve is time and the ordinate is set to
1.0 for implicit and to 0.0 for explicit (curve is a step
function).
This switching is activated by setting IMFLAG in
*CONTROL_IMPLICIT_GENERAL to -|curve ID|. Switching
from one analysis to the other is seamless and has no CPU or
I/O overhead.
The objective is to apply the preload using the implicit solver
and then switch to explicit on-the-fly to begin the transient
solution.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 18
9
Preload via Implicit Analysis
Approach 3: Single run featuring implicit by
*CONTROL_DYNAMIC_RELAXATION
IDRFLG=5 or 6 activates precursor implicit solution to
achieve preloaded state.
Only part set DRPSET is active during implicit phase if
IDRFLG=6
Set DRTERM to termination time of implicit preload solution.
*CONTROL_IMPLICIT_... commands provide controls on
implicit preload solution.
Set implicit s tep s ize DT0 in *CONTROL_IMPLICIT_GENERAL
Leave IMFLAG=0 s o only the precurs or, preload s olution is
implicit
Other implicit controls are at the dis cretion of the analys t (s tatic
vs. implicit transient, automatic step adjustment, etc.)
Regular, explicit solution commences from t=0, starting from
the preloaded state.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 19
Preloading Bolts
Iterative Loading Types
Require multiple runs to tune load in order to give desired bolt
stress
*LOAD_THERMAL_LOAD_CURVE
*CONTACT_INTERFERENCE
Non-iterative Loading Types
Bolt stress is specified directly.
*INITIAL_STRESS_SECTION
Solid elements only
*INITIAL_AXIAL_FORCE_BEAM
Type 9 beams only
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 20
10
*LOAD_THERMAL_LOAD_CURVE
In this method, we shrink the bolt by cooling it. As the bolt
contracts during the DR phase, preload is induced.
Coefficient of thermal expansion (CTE) must be given for bolt
material, e.g., via *MAT_ADD_THERMAL_EXPANSION.
Negative temperature is prescribed using
*LOAD_THERMAL_LOAD_CURVE.
LCID = curve of temperature vs . time for trans ient phas e (cons tant T).
LCIDDR = curve of temperature vs . time for DR phas e.
SIDR=1 in *DEFINE_CURVE.
Ramp T and then hold cons tant.
Temperature T (or CTE) necessary to produce a target bolt stress s
can be estimated.
s = E * CTE * -T
Adjust T (or CTE) in subsequent run to fine tune the final bolt stress
Example: [Link]
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 21
*CONTACT_..._INTERFERENCE
Developed for modeling interference-fit assemblies.
Define the mesh to include finite initial penetration between parts.
The meshed geometry represents the unstressed state.
Initial penetrations are not removed during start up of analysis but
rather are allowed to generate contact forces.
To avoid sudden, large contact forces, the contact stiffness is scaled
with time using LCID1 (DR phase) and LCID2 (Transient phase).
Shell thickness offsets are considered.
Orientation of contact segments is important.
Keyword commands for interference contact:
*CONTACT_NODES_TO_SURFACE_INTERFERENCE
*CONTACT_ONE_WAY_SURFACE_TO_SURFACE_INTERFERENCE
*CONTACT_SURFACE_TO_SURFACE_INTERFERENCE
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 22
11
*CONTACT_..._INTERFERENCE
If DR phase is included,
Dynamic relaxation (LCID1) + Transient Phase (LCID2)
Contact 1.0 Contact 1.0
Stiffness Stiffness
Scale Scale
Factor Factor
Time Time
Or if no DR phase,
Transient Phase Only (LCID2) if LCID1=0
Contact 1.0
Stiffness
Scale
Factor
Time
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 23
*CONTACT_..._INTERFERENCE
Example:
[Link]
Four bolts clamp 1.0” thick, solid rings together.
Mesh is defined so each bolt head and each nut overlap (penetrate) the
solid ring surface by 0.003”.
Trial overlap based loosely on target bolt stress/(bolt length * E)
*CONTACT_SURFACE_TO_SURFACE_INTERFERENCE defined between
overlapping surfaces.
Contact stiffness is ramped up during DR phase.
Overlap can be adjusted in subsequent trials to fine tune the bolt stress.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 24
12
Preloading Bolts Modeled with Solid Elements
*INITIAL_STRESS_SECTION is yet another method for
preloading bolts. It acts by prescribing a stress value to
solid elements cut by a cross-section.
Stress (normal to the cross-section) is defined via
*DEFINE_CURVE (stress vs. time).
This curve is typically flagged with SIDR=1, so that dynamic
relaxation is invoked for applying the preload.
Curve should ramp stress from zero and then hold target stress value
long enough for a state of near equilibrium in the model to be
reached, i.e., long enough for DR to converge.
Physical location of cross-section is defined via
*DATABASE_CROSS_SECTION.
A part set, together with the cross-section, identify the elements
subject to the prescribed stress.
Contact damping (VDC) and/or *DAMPING_PART_STIFFNESS may
be necessary to attain convergence during the DR phase.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 25
Example:
*INITIAL_STRESS_SECTION
[Link]
Four bolts clamp 1.0” thick, solid rings together.
A cross -section is defined that cuts through the bolts somewhere in the
middle.
The bolt elements cut by the cross -section have longitudinal stress ramped
to 20,000 psi during DR phase using *INITIAL_STRESS_SECTION .
The direction of prestress is normal to the plane.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 26
13
*INITIAL_STRESS_SECTION
Example of preloaded solid bolts
Target bolt stress is
achieved directly without
the need for multiple trial
simulations.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 27
Preloading Bolts Modeled with Beam Elements
*INITIAL_AXIAL_FORCE_BEAM will preload beam
elements to a prescribed axial force.
The preload curve (axial force vs. time) is defined with
*DEFINE_CURVE.
The curve is typically flagged with SIDR=1 so preload is applied
during the DR phas e.
Curve should ramp force from zero and then hold target force
value long enough for a s tate of near equilibrium in the model to
be reached, i.e., long enough for DR to converge.
The beams to be loaded are given by *SET_BEAM.
Beam formulation (ELFORM) must be set to 9 (spot weld
beam) and the material model should be *MAT_SPOTWELD.
Contact damping (VDC=20) is recommended.
*DAMPING_PART_STIFFNESS (COEF=0.1) may speed up
convergence during DR phase.
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 28
14
Example:
*INITIAL_AXIAL_FORCE_BEAM
[Link]
The bolt is modeled with a type 9 beam and *MAT_100.
The bolt beam is attached to the plates being bolted by 4 rigid beams at
each end.
The bolt is preloaded by ramping the axial force to 0.05 during the DR
phase using *INITIAL_AXIAL_FORCE_BEAM.
No additional load is applied in subsequent transient phase.
Bolt
Deformable plates
Rigid beams Bolt beam
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 29
*INITIAL_AXIAL_FORCE_BEAM
Example of preloaded beam bolt (cont’d)
Stress at conclusion of DR phase due to bolt preload.
Axial force in bolt is
successfully
initialized
Copyright © 2003-2012 by LIVERMORE SOFTWARE TECHNOLOGY CORPORATION Preload 30
15