Technical Guide

Reference models for explicit

analyses using LS-DYNA

2019-03-21 v1.0 2019 © DYNAmore Nordic AB LS-DYNA

Introduction

Contents
Introduction ........................................................................................................................2
Control cards .....................................................................................................................3
Contact settings .................................................................................................................6
Example 1: Droptest ..........................................................................................................8
Example 2: Crash box......................................................................................................11
Example 3: Deep drawing ................................................................................................13
Example 4: Post-buckling strength ...................................................................................16
Example 5: Bolted connection..........................................................................................20
Example 6: Interference fit ...............................................................................................23
Example 7: Filled weld connection ...................................................................................26
Example 8: Rubber seal...................................................................................................29
Record of revisions ..........................................................................................................32
Copyright and Trademark Notice .....................................................................................32
Disclaimer ........................................................................................................................32

2019 © DYNAmore Nordic AB 1 LS-DYNA

Introduction

Introduction
This document includes examples of typical applications for explicit simulation using
LS-DYNA® from LSTC. The document is to be used in conjunction with running/studying
the accompanied FE-models. The intention from the authors is to provide hands-on
guidelines, by providing LS-DYNA models of typical applications, on how to make well-
conditioned LS-DYNA models for explicit simulation. The modeling is based on the
experience from the authors and an effort has been made to use generally accepted best
methods and settings. The idea is that the methods and settings described in this document
can serve as a good starting point when building similar models for explicit simulation.
The models have been built using LS-PrePost® from LSTC and ANSA from BETA CAE
Systems. The post-processing has been conducted using LS-PrePost and META from
BETA CAE Systems. Many figures in this document has been rendered using META. The
examples are developed with mpp/LS-DYNA in mind and tested with version R11.0.0 in
single precision. To provide a relevant measurement of how computationally demanding
each example is, the total elapsed time of each simulation/example is noted. For this
purpose, all examples have been benchmarked on 4 cores on a Linux machine equipped
with dual Xeon SP 6148 CPU`s (2.40 GHz). Output files such as D3PLOT and BINOUT can
be studied in any suitable LS-DYNA post-processor, such as for instance LS-PrePost.
There is an output file called “decomp_parts.lsprepost” that can be read into LS-PrePost
under the “Model and Part – Views - MPP” in combination with the D3PLOT which allows
the decomposition to be viewed.
Although suitable material models have been used in the examples, this document shall not
be considered as a material modeling guideline.
Support and training
For thoroughly details regarding LS-DYNA keywords and material models, the reader is
referred to the LS-DYNA User’s Manuals1. For help with using LS-DYNA and LS-PrePost
contact your local LS-DYNA distributor, see also www.lstc.com for a list of distributors.
Useful web resources include www.dynasupport.com for general support and
www.dynalook.com for conference papers. Example keyword files can be found at
www.dynaexamples.com. If you find errors in this document, you are welcome to contact
support@dynamore.se.
The document is under continuous development and future improved revisions are
expected.
System of units
The system of units used in the models is mm, ms, kg, kN, GPa.

1
Livermore Software Technology Corporation, LS-DYNA keyword user’s manual Volume 1, 2 & 3, Livermore
2019. (see also http://www.lstc.com/download/manuals).

2019 © DYNAmore Nordic AB 2 LS-DYNA

Control cards

Control cards
The control card settings used in the example models, see the file cards.inc in each example
model, is presented in this chapter. Most of the input parameters are left to their default
settings. The parameters that differ from default are here highlighted with blue colour. The
control settings have, for some example models, been slightly adjusted to comply better
with the specific problem at hand. See individual example models for details.
The settings, which are based on our experience, are shown in the cards below.

Comment: Objective stress update and invariant node numbering is switched on.

Comment: Bulk viscosity is activated for some shell element formulations in addition to the
standard bulk viscosity for solids.

Comment: Update of the reference node coordinate is turned on for beams for visualization
purposes, see NREFUP. This requires a unique third node do be defined for each beam
element. Output eroded internal and kinetic energy in the matsum file, see IERODE.

2019 © DYNAmore Nordic AB 3 LS-DYNA

Control cards

The following is activated: Shell thickness changes is considered in single surface contact,
see THKCHG above and ISTUPD in *CONTROL_SHELL. Actual shell thickness is used in
single surface contact, see SSTHK. SPOTDEL is activated so that the spot weld is deleted
when an attached shell fails.

Comment: Energy calculation is switched on for all elements.

The following is activated: Automatic sorting of degenerated quadrilateral shell elements,

see ESORT. Shell thickness change due to membrane straining is turned on, see ISTUPD.
Shell thickness offset, e.g. by NLOC, affects the contact reference plane, see CNTCO.

2019 © DYNAmore Nordic AB 4 LS-DYNA

Control cards

Comment: Automatic sorting of degenerated tetrahedron and pentahedral solid elements is

switched on.
Additional settings
In addition to the cards above, *CONTROL_MPP_IO_LSTC_REDUCED and *CONTROL_-
MPP_IO_NODUMP are used. There are no input parameters for these cards.
Note that *CONTROL_TERMINATION and *CONTROL_TIMESTEP can be found in the
main file run.key and not in the file cards.inc.

2019 © DYNAmore Nordic AB 5 LS-DYNA

Contact settings

Contact settings
Four main contact definitions are used in the examples, i.e. *CONTACT_AUTO-
MATIC_SINGLE_SURFACE with SOFT=1 or SOFT=2 for shells and solids and
*CONTACT_GENERAL for beams. In addition, *CONTACT_ERODING_SINGLE_SUR-
FACE is used for the cases where erosion of shells or solid occur in the model. The contact
settings are mostly left to their default values. Common for these contacts is that friction
coefficients are set with FS/FD. Contact damping is added with VDC. The soft constraint
option is chosen with SOFT. DTSTIF, which can be used to set the time step used to
calculate the contact stiffness is activated and set to the initial time step of the simulation at
hand.
The settings, which are based on our experience, is shown in the cards below.

Comment: SOFT=1. IGNORE is activated so that initial penetrations are tracked, i.e. nodes
are not moved.

Comment: SOFT=2 with SBOPT=3 and DEPTH=35. PSTIFF=1 and SFNBR=-1.0.

SHLEDG=1 to get square shell edges flush with the nodes.

2019 © DYNAmore Nordic AB 6 LS-DYNA

Contact settings

Comment: The eroding contact has the same settings as the ordinary SOFT=2 contact
presented above except that one extra card has been added specific for the ERODING
functionality.

Comment: *CONTACT_AUTOMATIC_GENERAL is only used for beam-to-beam contact in

this guideline.
Other contacts
In addition to the main contacts described in this section other contacts are occasionally
used such as for instance TIED-contacts, FORMING-contacts, MORTAR-contacts and
*FORCE_TRANSDUCER. Settings for these contacts can be found in the models where
they are used.

2019 © DYNAmore Nordic AB 7 LS-DYNA

Example 1: Droptest

Example 1: Droptest
This is an example of a drop test simulation using LS-DYNA. A ceramic mug is dropped
from a given height, i.e. 1.5 m, allowing it to freefall until it hits a relatively hard surface. The
model exemplifies, among others, a parameter definition, an accelerometer, a strain gauge
load cell, measuring of cross section forces and the contact treatment of eroding elements.

Simulation data
# nodes 185k
# elements 147k
Timestep size (ms) 0.1e-3 (constant)
Termination time (ms) 15.0
Solution time (minutes) 62
Elements and material
The mug and the surface are meshed with four solid ELFORM 1 hexahedral elements
through the thickness. The element size is 1 mm. Stiffness-based hourglass control type 6
is applied (QM=0.03). Material type MAT_1102 is used for the mug. Elements erode when
the negative pressure reaches 0.05 GPa. This can be confirmed by, for instance, plotting
the pressure for solid EID 817 or 19649 (depending on which element erode) in ELOUT.
The fracture is brittle.

2
Parameters collected from the article “Implementation and Validation of the Johnson-Holmquist Ceramic
Material Model in LS-DYNA”, by D. S. Cronin et al, and modified to fit the purpose of the simulation.

2019 © DYNAmore Nordic AB 8 LS-DYNA

Example 1: Droptest

Contact definitions
Contact type *ERODING_SINGLE_SURFACE with SOFT=2 is used for the internal contact
of the mug so that contact is treated correctly, i.e. with updated contact surface, also after
elements have been eroded. An ERODING contact automatically invokes a negative
volume failure criterion for all solids in the model. The use of PSFAIL on *CONTROL_SOLID
limits this criterion to the partset that is referenced by PSFAIL. Contact type
*AUTOMATIC_NODES_TO_SURFACE with SOFT=1 is used for the contact between the
mug and the surface. Additional tied contacts attach the handle and logo to the mug. The
contact force between the mug and the surface can be plotted from output file RCFORC,
see CID 2.
Control cards
No mass-scaling or very little mass scaling is recommended when performing drop test
simulations so that the kinetics is not significantly affected due to additional artificial added
mass. In this case mass scaling is applied, see DT2MS on *CONTROL_TIMESTEP, it adds
4 extra grams to the mug. This is an increase of 2% of the total weight of the mug. This may
be acceptable given that the time stepsize was increased by a factor 2 and therefore
shortening the simulation time considerably. Parameter MSSCL is set to 1 on
*DATABASE_EXTENT_BINARY allowing fringe ploting of the added mass from D3PLOT.
Loads and constraints
Gravity is applied with *LOAD_BODY_Z. The bottom of the impacted surface is fully
constrained with SPC.
Initial conditions
The impact velocity is set with a combination of *INITIAL_VELOCITY, *PARAMETER and
*PARAMETER_EXPRESSION so that only the drop height is required as input. Thereby
preparing the model for quick configuration when performing, for example, parameter
studies.
Output
An accelerometer is installed on the handle to output accelerations in a local coordinate
system in NODOUT (NID 185241). An additional node (NID 185243) on the same
accelerometer outputs accelerations in the global coordinate system. The global node can
be used to check that the output frequency of the accelerometer is sufficiently high. I.e. by
integrating each acceleration component (x, y, z) and compare the results to the
corresponding velocities. These curves should match when offsets are removed. A relevant
low-pass filter is needed to remove noise when plotting the NODOUT data. In this case a
Butterworth filter with a cut-off frequency of 3000 Hz produces good plots for velocity and
acceleration.

Two strain gauge load cells (shell EID 200001 and EID 200002) are installed on the inner
and outer surface of the mug, respectively. The strain gauge results, given in the local
coordinate system of the shell element, can be plotted from ELOUT. Again, a Butterworth
filter with a cut-off frequency 3000 Hz produce good plots. The results show tension on the

2019 © DYNAmore Nordic AB 9 LS-DYNA

Example 1: Droptest

outer surface (positive x-strain) and compression on the inner surface (negative x-strain) as
expected. Note that STRFLG=1 must be set on *DATABASE_EXTENT_BINARY to get
strain data in ELOUT and D3PLOT.
*DATABASE_CROSS_SECTION, and its output file SECFORC, is used for measuring the
force in the attachment between handle and mug. In this case the same results can be
plotted from RCFORC CID 1. Use the same filter as mentioned before to remove noise.
Simulation check
A scan of the d3hsp and message files shows no errors or alarming warnings. The energy
ratio is rather close to 1.0, as expected. The amount of added mass, that is checked by
plotting GLSTAT, is low. Animating the simulation shows that the contact function as
expected.

2019 © DYNAmore Nordic AB 10 LS-DYNA

Example 2: Crash box

Example 2: Crash box

This is an example of a crash box buckling analysis using LS-DYNA. Three identical, i.e.
except for the mesh size, crash boxes are impacted by rigid cylinders. Each cylinder has a
weight of 700 kg. The impact speed is 30 km/h. This example demonstrates, among others,
*MAT_24 with strain rate effects, history variables, a force transducer and the concept of
mesh convergence.

Simulation data
# nodes 81k
# elements 72k
Timestep size (ms) 0.3e-3 (constant)
Termination time (ms) 100.0
Solution time (minutes) 95
Elements and material
The three crash boxes (100x100x500 mm, thickness 2.0 mm) are modelled with shell
elements ELFORM -16, which is a fully integrated element type. Hourglass control type 8 is
applied. The mesh size is for the three boxes 10 mm, 5 mm and 2.5 mm, respectively. The
material model used for the steel is *MAT_24 where strain rate effects are accounted for by
adding a strain rate table definition. The table includes two hardening curves corresponding
to static load (strain rate=0) and strain rate 1000/s. Intermediate strain rate values in LS-
DYNA are interpolated between the curves. By setting VP=1 on the material card, the
numerical noise caused by strain rate effects can be minimized.
Contact definitions
The contact, for the complete model, is treated by one *AUTOMATIC_SINGLE_SURFACE
contact with SOFT=1. A key factor when evaluating the performance of a crash box is the

2019 © DYNAmore Nordic AB 11 LS-DYNA

Example 2: Crash box

force vs. displacement curve. By adding *FORCE_TRANSDUCER, i.e. one for each
impactor, the force between each impactor and its corresponding crash box can be
measured.
Control cards
By setting NEIPS=1 on *DATABASE_EXTENT_BINARY the effective plastic strain for
*MAT_24 is output to the D3PLOT as history variable #1, which can be fringe plotted in a
post-processor. In LS-PrePost this can be done under “FComp-Misc-History var#1”.
Loads and constraints
The rear end of the crash box is fixed with *BOUNDARY_SPC_SET.
Initial conditions
The initial velocity of the impactors is applied with *INITIAL_VELOCITY_RIGID_BODY.
Output
By specifying the same out frequency for RCFORC and NODOUT the force versus
displacement curve can easily and accurately be plotted in a post-processor. The force
curve for the tube with 5 mm mesh and 2.5 mm are similar whereas the 10 mm mesh
produce a significantly stiffer response. This shows that the 10 mm mesh is too coarse to
solve the buckling accurately, i.e. the 10 mm mesh is not converged.
Simulation check
A search in the d3hsp and message files shows no errors or alarming warnings. Note, there
are warnings about 16 elements with warpage angle > 30 degrees. These elements can be
found in the area of the triggers and can here be accepted. The total amount of added mass
due to mass-scaling, which is 16 grams is acceptable. The energy ratio is about 1.0
throughout the analysis and the overall energy balance is reasonable.

2019 © DYNAmore Nordic AB 12 LS-DYNA

Example 3: Deep drawing

Example 3: Deep drawing

This is an example of a deep drawing simulation using LS-DYNA. A half-symmetry cross
member is formed with a single action die process. This example shows a typical setup for
a deep drawing simulation with blank and tool definition, process boundary conditions,
material model and some recommended settings for metal forming simulations. A spring-
back analysis is performed after the forming simulation.
The model uses *PARAMETER and *PARAMETER_EXPRESSION to define the process.
The user needs to define the parts IDs for the blank and tools, distances between the tools,
blank thickness, tool velocities and binder force. The load curves for the motions of the tools
and the end time are calculated from the user input.

Simulation data
# nodes 57k (initial)
# elements 60k (initial)
Timestep size (ms) 0.54e-3 (constant)
Termination time (ms) 33.32
Solution time (minutes) 18
Elements and material
The blank and tools are modelled with shell elements. The tools are rigid and meshed with
deviation mesh mode which allows the elements to follow the correct tool geometry. The
blank is meshed with an initial element length of 10 mm but has also adaptive mesh applied

2019 © DYNAmore Nordic AB 13 LS-DYNA

Example 3: Deep drawing

which enables the elements to be refined in three steps. A stiffness based hourglass control
type 4 is applied (QM=0.1). *MAT_BARLAT_YLD2000 is used for the blank with material
parameters defining the anisotropic behaviour of the material. The hardening is defined with
a load curve and the material direction is defined by the AOPT flag.
Contact definitions
Contact type *FORMING_ONE_WAY_SURFACE_TO_SURFACE is used between the
blank and all the tools. This contact type is recommended for sheet metal forming
simulations with some parameters setting defined for this purpose, e.g. the tool thickness is
not considered. Note that this a non-symmetric contact type and the blank should be set as
slave. The contact is also sensitive for orientation of the mesh, and the normal of the tool
elements should be directed toward the blank.
Control cards
The mesh adaptivity of the blank is defined in *CONTROL_ADAPTIVE. Here the adaptivity
frequency, tolerance, refinement levels etc. are defined. Also set ADPOPT to 1 for the blank
in PART to activate the adaptivity for the part. In *CONTROL_CONTACT it is recommended
to set PENOPT=4 for sheet metal forming simulations as the mesh density is usually
different between contact slave and master sides.
Deep drawing simulations when the blank is clamped between the die and binder have
usually low dynamic effects and can therefore be mass scaled to a quite high ratio. The
timestep used for this simulation is 6e-4 ms, see DT2MS on *CONTROL_TIMESTEP, which
increases the blank mass by approximately 700 %.
Loads and constraints
The tool motion is defined with a trapezoidal load curve defining the velocity with a ramping
time of 1 ms. The process is divided into two steps; closing and forming. In the closing step
the die is translated with the closing velocity, by *BOUNDARY_PRESCRIB-
ED_MOTION_RIGID, and the binder is fixed. In the forming step the die is moved with the
forming velocity, and the binder has a load applied to it with *LOAD_RIGID_BODY. The
binder also has a vertical constraint (*CONSTRAINED_RIGID_BODY_STOPPERS)
applied to it during the forming step. The punch is fixed during the whole process.
Constraints on the blank symmetry edge is defined with *BOUNDARY_SPC_NODE with
nodal constraints to simulate a symmetry condition.
Output
The ASCII files usually needed for a sheet metal forming simulation is BNDOUT, GLSTAT,
MATSUM, RCFORC and SLEOUT. In *DATABASE_EXTENT_BINARY the number of
integration points written out is increased from three to five by MAXINT. The STRFLG is set
to 1 to get the strain data. The strain output is needed when doing post-processing with
Forming Limit Diagram (FLD).
*INTERFACE_SPRINGBACK_LSDYNA is used to extract a DYNAIN file at the end of the
simulation containing the thickness, strain and stress in the blank.

2019 © DYNAmore Nordic AB 14 LS-DYNA

Example 3: Deep drawing

Simulation check
A search in the d3hsp and message files shows no errors or significant warnings. There are
no hot spots of plastic strain in the blank that could indicate tool geometry errors. The
contact forces show some oscillations due to the mesh adaptivity, but apart from that the
curves looks smooth.
Springback
A springback analysis gives the resultant deformation of the blank when it is removed from
the forming tools. The analysis is performed using static implicit. The element formulation
is switched from ELFORM 2 to ELFORM 16. The number of integration points (NIP) is
increased from five to seven to improve the accuracy. Some constraint points need to be
added to prevent the rigid body motions. When adaptivity is used for the blank it is
convenient to use *CONSTRAINED_COORDINATE rather than SPCs as this keyword is
not dependant on the node numbering.

2019 © DYNAmore Nordic AB 15 LS-DYNA

Example 4: Post-buckling strength

Example 4: Post-buckling strength

This is an example of a quasi-static buckling analysis using LS-DYNA. The target of the
simulation is not only to find the ultimate strength of the profile but also to find the residual
post-buckling strength. The example demonstrates, among others, a testing rig with an
applied uniform pressure load, the use of a PI-regulator as well as the inclusion of geometry
perturbations. The use of selective mass-scaling is briefly discussed.

Simulation data
# nodes 136k
# elements 136k
Timestep size (ms) 0.664e-3 (constant)
Termination time (ms) 150.0
Solution time (minutes) 1573
Elements and material
The steel profile (750x45x2000 mm, thickness 0.5 mm) are modelled with shell elements
ELFORM -16, which is a fully integrated element type. Hourglass control type 8 is applied.
A uniform mesh with element size 4-5 mm is applied. Element thickness perturbation
(±0.003 mm) is introduced with *PERTURBATION_SHELL_THICKNESS. The material
model used for the steel is *MAT_24. Strain-rate effects are not considered since the
simulation is to be quasi-static. It is important to have a realistic description of the hardening
curve, also in the necking region, when simulating post-buckling behaviour. The material
curve in this example allows strain hardening up to about 20% plastic strain when necking
kicks in.

3
The solution time was reduced to 83 minutes with Selective Mass Scaling (SMS). SMS is not included in this
model but can easily be introduced as discussed in the document.

2019 © DYNAmore Nordic AB 16 LS-DYNA

Example 4: Post-buckling strength

Contact definitions
The contact between the profile and the supports is handled by an
*AUTOMATIC_SINGLE_SURFACE contact with SOFT=1, i.e. the global contact of the
model.
Control cards
The control cards settings in this example does not deviate from the settings as described
in chapter “Control cards”. What has been added for this example is *DATA-
BASE_CURVOUT that output the curve from *DEFINE_CURVE_FUNCTION, i.e. in this
example the steer signal for the pressure as applied by the PI-controller.
Loads and constraints
Boundary symmetry conditions are set in the transverse direction by *BOUND-
ARY_SPC_SET. Gravity is accounted for by *LOAD_BODY_Z. The profile is loaded with a
uniform pressure. The pressure can be applied in different ways where a relatively slow
ramp load is the most straight forward approach. This works fine up till ultimate strength is
reached. After that the kinetic energy will increase rapidly as the profile collapses due to
buckling. A ramp load cannot be used to evaluate the post buckling strength of this profile.
For this reason, a PI-controller is used as defined by *DEFINE_CURVE_FUNCTION with
function PIDCTL. The controller is set to keep a steady load velocity of 0.4 m/s by controlling
the applied pressure. As with all PI controllers adjusting the parameters can be a bit tricky.
In this case the initial settings of parameters, i.e. the proportional gain kp and the integral
gain ki, gave a fast prediction of ultimate load as well as post-buckling strength. However,
the elastic responds before buckling was oscillating too much. This was improved by
decreasing kp and increasing ki.
Output
The steer signal to the applied pressure is output in CURVOUT. By scaling CURVOUT by
a factor of 1.0e-5 the applied pressure (GPa) can be plotted. BNDOUT gives the reaction
forces in the supports. In addition, the z-displacement can be plotted from NODOUT. The
node at hand, i.e. NID 108355, is used by the PI-regulator to measure the velocity of the
applied load. This node is the dependent node in a *CONSTRAINED_INTERPOLATION
definition and follows the motion of the profile. By this arrangement a more robust setpoint
(in this case the velocity) can be monitored.

2019 © DYNAmore Nordic AB 17 LS-DYNA

Example 4: Post-buckling strength

Comment: Pressure vs. displacement.

Simulation check
The energy curves are reasonable. Most importantly the kinetic energy is held to a low level
throughout the simulation, i.e. the peak kinetic energy is about 7% of the total energy just
after buckling. It is then rapidly decreased to a level <1%. The figure below shows the kinetic
energy in the simulations with a ramp load compared to a PI-controller. It is evident that a
ramp cannot be used to evaluate post-buckling strength. Note that in the figure below the
time for the RAMP-curve has been scaled down a factor 10 for a clearer comparison.

Comment: Kinetic energy with a ramp load (red curve) and with a PI-controller (green
curve).

2019 © DYNAmore Nordic AB 18 LS-DYNA

Example 4: Post-buckling strength

The added mass due to conventional mass scaling is about 15 grams at the end of the
simulation, which is acceptable.
Some further testing showed that it is possible to decrease the simulation time about 40%
by using selective mass scaling. This can be an efficient way to speed up for instance an
optimization loop. Though, care must be taken so that the mass scaling does not
significantly affect the simulation results. To activate selective mass scaling set IMSCL=1,
or preferably refer to a *SET_PART, on *CONTROL_TIMESTEP. DT2MS=-1.4856e-3 was
found to give a good speed-up without compromising the results.

2019 © DYNAmore Nordic AB 19 LS-DYNA

Example 5: Bolted connection

Example 5: Bolted connection

This is an example of a bolted connection simulation in LS-DYNA. Bolted connections can
be modelled in many ways and with different levels of accuracy depending on the purpose.
The model presented here can probably be classified as (upper) mid-range in terms of
accuracy, i.e. the connection includes friction between bolt head/washer and underlying
surface, it allows slip due to play between the bolt shank and the hole edge, it allows tear-
out of the hole (providing the hole is modelled fine enough), it allows the bolt to deform with
a certain degree of freedom, it allows the bolt to fail (by for instance shear fracture) and it
includes the possibility of bolt pre-tensioning. The modelling technique presented here is
supposed to provide an affordable accuracy for structural and crash simulations where
loading above service loads are expected such as slip, bolt fracture and possibly tear-out.
The same bolted connection is in this example represented in two ways, i.e. by beam
elements (ELFORM 9) for the bolt shank as well as with solid elements for the bolt shank.
Some of the pros and cons of these two variations are briefly discussed.
This example demonstrates, among others, the use of *INITIAL_AXIAL_FORCE_BEAM,
*INITIAL_STRESS_SECTION, *CONTACT_AUTOMATIC_GENERAL and *CONTROL-
_DYNAMIC_RELAXATION.

Simulation data
# nodes 19k
# elements 18k
Timestep size (ms) 0.4e-3 (constant)
Termination time (ms) 40.0
Solution time (minutes) 13
Elements and material
One of the two bolts are modelled with beam elements ELFORM 9 (spotweld beam) for the
shank and shell elements ELFORM -16 for the bolt head/washer. The other bolt is modelled
completely with solid elements ELFORM 1. Hourglass control type 6 and type 8 is applied
to the solids and shells, respectively. The bolt shank is discretized with 7 elements along
the bolt shank in both cases resulting in a minimum element size of about 2 mm. Material

2019 © DYNAmore Nordic AB 20 LS-DYNA

Example 5: Bolted connection

model *MAT_100 (*MAT_SPOTWELD), which follows a bilinear plastic hardening curve, is

used for the ELFORM 9 beams since *MAT_100 is the required material model for spot
weld beams. *MAT_24 is used for the solid ELFORM 1 bolts. In this example the same
bilinear hardening curve is used for both the beam bolt and the solid bolt for simplicity
reasons. However, using solid elements opens the possibility to use a piecewise linear
hardening curve for the bolt. The bolt elements are set to fail at 20% effective plastic strain.
It can be mentioned that the main reason for using ELFORM 9 beams and not, for instance,
ELFORM 1 beams is to be able to use *INITIAL_AXIAL_FORCE_BEAM for the pre-stress.
Contact definitions
*CONTACT_ERODING_SINGLE_SURFACE with SOFT=2 (SHLEDG=1) is the main
contact used in this example. The reason for using ERODING is to maintain proper contact
also after solid elements have been eroded. The parameter PSFAIL on *CONTROL_SOLID
is used to limit the invoked negative volume failure criterion to only the solid bolt.
*CONTACT_AUTOMATIC_GENERAL_MPP is used between ELFORM 9 beams and the
null-beams that surrounds the hole.
CPARM8=2 activates contact for ELFORM 9 beams. Note that the diameter of the null-
beams is chosen to be very small (0.001 mm) so that the true diameter of the hole remains,
for contact considerations, unaffected. This is reasonable since the diameter of the
ELFORM 9 bolt is, in this context, large.
Control cards
The control cards basically follow the standard setup of this document, see the chapter “
Control cards”. Though, DRCPSID on *CONTROL_SHELL is set to activate a drilling
rotation constraint for the shell elements of the ELFORM 9 beam bolt head/washer. Without
this the beams of the bolt shank would be unconstrained in torsion/drilling.
Loads and constraints
The tubes are fully constrained at one end and pulled in the other end using rigid grips in
combination with *BOUNDARY_PRESCRIBED_MOTION_RIGID. The bolts are
prestressed using dynamic relaxation, i.e. before the transient analysis takes place.
Dynamic relaxation is activated by SIDR=1 on the load curves for *INITI-AL_FORCE_BEAM
and *INITIAL_STRESS_SECTION, respectively. Default parameters are used on
*CONTROL_DYNAMIC_RELAXATION except for DRTERM=10.0 that is used to limit the
relaxation phase to maximum of 10 ms. *DATABASE_BINARY_D3DRLF is set so that the
dynamic phase can be studied in more detail using a post-processor (see D3DRLF). Note
that KBEND and IZSHEAR is activated on *INITIAL_FORCE_BEAM and
*INITIAL_STRESS_SECTION to allow bending stiffness/shear stresses in the bolts during
the pre-stress/relaxation phase. In this case a constant loading curve was chosen for the
pre-stress, which was found to give fast convergence, see database file RELAX. It can be
mentioned that in many cases a gradual ramp up the load may be preferred.
Output
Spot weld forces are output to SWFORC for all ELFORM 9 spot weld beams. The elements
(beams and solids) that shear off during the loading are specially studied in this example

2019 © DYNAmore Nordic AB 21 LS-DYNA

Example 5: Bolted connection

using ELOUT. For instance, it can be shown that the beam element is deleted when all 4
integration points have reached 20% effective plastic strain. The solid element is deleted
when the single integration point reaches 20% effective plastic strain. In addition, cross
section definitions are used to study the shear failure of the solid bolt (see SECFORC). In
BNDOUT one can find the output reaction forces for the grip, which makes it easy to
determine the failure strength of the bolted connection. The results show that the beam bolt
variant gives a significantly stronger connection compared to the solid bolt variant. The main
reason for this is probably that beam elements lack important degrees of freedom compared
to the solid element variant. The offset in time of the reaction forces is related to the
geometry representation of the bolt: the beam bolt is cylindrical whereas the solid bolt is
faceted. Differences in contact stiffness influence the reaction forces as well.
Simulation check
The mass scaling is kept at a low level, i.e. about 80 grams, which is acceptable. The energy
components, i.e. hour glass energies, are reasonable. Regarding mass scaling it can be
mentioned that selective mass scaling can be a good option when the mesh locally, at the
bolted connection, needs to be refined compared to the rest of the model. This has not been
studied in this example though.

2019 © DYNAmore Nordic AB 22 LS-DYNA

Example 6: Interference fit

Example 6: Interference fit

This is an example of an interference fit analysis in LS-DYNA. The model consists of the
inner shaft and the outer hub. In the simulation, the hub is heated 250°C, which introduces
the interference pressure. The applied heating corresponds to an allowance of about 0.1
mm (0.25% of the diameter). Firstly, the shaft is loaded with a prescribed rotation and,
secondly, the shaft is pulled thus testing the performance of the friction fit for torque and
axial load. In this case the simulation results can be compared to an analytical solution. This
example demonstrates, among others, the use of *LOAD_THERMAL_VARIABLE,
*CONTACT_AUTOMATIC_SURFACE_TO_SURFACE_SMOOTH and the INTFOR file.

Simulation data
# nodes 7k
# elements 17k
Timestep size (ms) 0.3e-3 (constant)
Termination time (ms) 100.0
Solution time (minutes) 12
Elements and material
The shaft is modelled with ELFORM 1 hexahedron and ELFORM 15 pentahedron elements
(through automatic sorting). The hub is modelled with ELFORM 13 tetrahedron elements.
Hourglass control type 6 is applied. Both the shaft and the hub are made of steel, *MAT_24,
where the yield stress is set high to ensure only elastic deformation.
Contact definitions
The contact between shaft and hub is treated by *CONTACT_AUTOMATIC_SUR-
FACE_TO_SURFACE_SMOOTH. The SMOOTH-option is necessary here in order to
transfer the torque correctly, i.e. the faceted surface that would be the case for a non-

2019 © DYNAmore Nordic AB 23 LS-DYNA

Example 6: Interference fit

SMOOTH contact would yield an incorrect torque. The interface force file is requested by
setting SPR=SPR=1 on contact card 1 in combination with using S=iff on the execution
command line. The output frequency is set by DATABASE_BINARY_INTFOR. The static
and dynamic friction coefficient is 0.1 in the model.
Control cards
The control cards follow the setup as described in chapter “Control cards”. What is added
is *CONTROL_DYNAMIC_RELAXATION for the temperature load.
Loads and constraints
The outer surface of the hub is fixed by *CONSTRAINED_INTERPOLATION, which allows
radial movement. The beam element is added for the purpose of fixing the dependent node.
This boundary condition is chosen in order to be able to compare the results with a known
analytical solution. The loading, i.e. x-moment and x-force, is applied to the rigid end of the
shaft with *BOUNDARY_PRESCRIBED_MOTION_RIGID.
Output
The main results from this simulation is the interference pressure, which can be fringe
plotted through the interface force file, as well as the axial friction force and transmission
torque. The latter two can be plotted from BNDOUT, i.e. x-force and x-moment respectively,
see figure below.

Comment: Transmission torque (red curve) and transmission axial force (green curve).

2019 © DYNAmore Nordic AB 24 LS-DYNA

Example 6: Interference fit

Simulation check
The mass scaling is kept at a low level, i.e. about 0.2 grams and the energy curves make
sense. No alarming warnings in d3hsp and message files. The measured allowance, i.e.
after compensation for the contact penetration, is about 0.088 mm. Analytically this would
yield an interference pressure of 194 MPa. The simulation is within 5% of this value (fringe
plot the pressure on the shaft). The maximum/ultimate analytical transmission torque is
2438 Nm. The simulation is within 3% of this value. The analytical transmission axial force
is 121.9 kN. The simulation is within 6% of this value. The analytical estimate is sensitive to
the input allowance. Hence, the accuracy in the comparison is estimated to be within a few
percent.

2019 © DYNAmore Nordic AB 25 LS-DYNA

Example 7: Filled weld connection

Example 7: Filled weld connection

This example exemplifies the modelling of a construction that is joined together with a filet
weld. The example is collected from a Master Thesis by Rikard Svärd at the Division of
Solid Mechanics Linköping University4. The Master Thesis demonstrates three different
methods of modelling the fillet weld, which are rigid connections using
*CONSTRAINED_NODAL_RIGID_BODIES and deformable connections using 3-noded
shells and 4-noded shells, respectively. It also demonstrates the modelling of Heat Affected
Zones (HAZ). The modelling method chosen in this technical guide is a variant of the
deformable fillet weld proposed by the Master Thesis combined with directions from the
handbook “Svetsutvärdering med FEM” by Åsa Eriksson and Anna-Maria Lignell et al 5.

Simulation data
# nodes 13k
# elements 13k
Timestep size (ms) 0.35e-3 (constant)
Termination time (ms) 160.0
Solution time (minutes) 44

4
R. Svärd, ”Factors Affecting the Finite Element Simulation of a ROPS Test of a Volvo Cab”, Linköping
University, 2007.
5
Å. Eriksson; A-M. Lignell; C. Olsson; H. Spennare, “Svetsutvärdering med FEM. Handbok för
utmattningsbelastade konstruktioner”, Sweden, 2002.

2019 © DYNAmore Nordic AB 26 LS-DYNA

Example 7: Filled weld connection

Modelling of the weld

The two steel beams were meshed together using 4-node and 3-node shells. The meshing
was made automatically using the connection assistant in Ansa. The fillet weld was then
identified as the two rows of elements (5 mm elements) that connects the two parts (green
in figure below). These elements were assigned a unique part id, which was given an
increased sheet metal thickness to account for the added weld material. I.e. the weld was
given the nominal thickness of the beams (6 mm) plus half the throat thickness (a/2). The
throat thickness was estimated to 6 mm (not given in report). Therefore, the thickness of
the weld was set to 9 mm. In addition, three Heat Affected Zones (about 10 mm each) were
identified as proposed by the Master Thesis where the first zone includes the weld itself
(green and red in figure). These zones were given unique part id’s as well. The nominal
hardening curve used for the steel was scaled in the Heat Affected Zones, according to the
Master Thesis, to account for the increased yield strength.

Elements and material

The beams (80x80 mm, thickness 6.0 mm) and weld are modelled with shell elements
ELFORM -16 and Hourglass control type 8. The mesh size is 5 mm. The material model
used for the steel is *MAT_24 with no strain rate effects included. Material failure is not
considered in this simulation. Material data for the steel is retrieved from material testing4.
Contact definitions
The contact is treated by a *CONTACT_AUTOMATIC_SINGLE_SURFACE contact with
SOFT=2.
Control cards
The control cards follow the setup as described in chapter “Control cards”.

2019 © DYNAmore Nordic AB 27 LS-DYNA

Example 7: Filled weld connection

Loads and constraints

The loading replicates the physical test that was conducted in the Master Thesis. The
loading rate in the physical test was 5 mm/min, which can be considered as a static load.
The corresponding loading in the simulation is conducted over 150 ms, which gives a
loading rate of 1333 mm/second. A test simulation with a significantly slower loading rate
(5x) gave essentially the same results. The loading rate over 150 ms, i.e.1333 mm/second,
was therefore considered slow enough for the quasi-static loading. The loading as well as
the fixed conditions were applied by *BOUNDARY_PRESCRIBED_MOTION_RIGID. In
order to get the rotation point moved to the rear edge of the grip, a PART_INERTIA definition
was used.
Output
The main results from the simulation is force vs. displacement of the grip, which is assessed
from combining the displacement of NODOUT with force from BNDOUT.
Simulation check
Mass scaling and kinetic energy is kept at a low level. No remarks in that aspect.
What is interesting is that an initial test simulation with nominal setup, i.e. just merging the
parts together without including HAZ or increased thicknesses at the weld gave a too weak
responds compared to the physical test, see green curve in the figure below. The ultimate
load was 18% too low. By adding the weld modelling as described in this document the
ultimate load was increased to be within 1% of the physical test (blue curve in figure). The
difference in linear stiffness between test and simulation is probably due to that the stiffness
of the test rig/grips is not considered in the simulation. Due to lack of data this was not
further investigated.

Comment: The test curve (red) is an estimated average from 5 tests documented in the
Master Thesis.

2019 © DYNAmore Nordic AB 28 LS-DYNA

Example 8: Rubber seal

Example 8: Rubber seal

This example exemplifies the modelling of a rubber seal. In this case a seal between the
door frame and the side window of a car. The purpose of this 3D quasi-static simulation is
to test the ability of the seal to prevent water and air leakage. The model consists of the
deformable rubber seal, which is made of two different rubber qualities, and the rigid door
frame and window. This example demonstrates, among others, the use of *MAT_77
(*MAT_HYPERELASTIC_RUBBER) in combination with a MORTAR-contact to handle the
complicated contact situation with high pressures that develops as the rubber deforms.

Simulation data
# nodes 30k
# elements 23k
Timestep size (ms) 0.311e-3 (constant)
Termination time (ms) 540.0
Solution time (minutes) 130
Elements and material
The rubber is modelled with solid (mostly) hexahedron ELFORM 1 elements in combination
with hourglass formulation IHQ=6 and QM=1.0. The material model is *MAT_77 with a
poison’s ratio of 0.495. The hyperelastic constants (C10, C01) are set directly. Frequency
independent damping is accounted for by G and SIGF. It was needed to decrease TSSFAC
from default 0.9 to 0.7 to achieve numerical stability of the material. See figures below for
material curves.

2019 © DYNAmore Nordic AB 29 LS-DYNA

Example 8: Rubber seal

Comment: The stress-strain curve was produced through axial tension and compression of
a one-element model.

Comment: The damping curve was produced by simulating the damped oscillations of a
solid element beam.
Contact definitions
The contact is treated by a *CONTACT_AUTOMATIC_SINGLE_SURFACE_MORTAR
contact. Default setting other than some added damping, VDC=40. DEFINE_FRICTION is
referenced by setting FS=-2.

2019 © DYNAmore Nordic AB 30 LS-DYNA

Example 8: Rubber seal

Control cards
The control cards follow the setup as described in chapter “Control cards” other than that
ISYM on *CONTROL_CONTACT references to a node set on the symmetry plane. The
MORTAR-contact picks this up and uses it the contact algorithm.
Loads and constraints
The movements of the rigid parts, i.e. smooth velocity curves, are controlled by
*BOUNDARY_PRESCRIBED_MOTION_RIGID. The symmetry constraints (in z-direction)
is defined by *BOUNDARY_SPC_SET.
Output
No special output for this this case. Animations that show the behaviour of the rubber are
produced through D3PLOTs.
Simulation check
The mass-scaling (conventional) is kept at a reasonable low level, i.e. about 50% at the end
of the simulation. This is acceptable considering the quasi-static nature of the problem.
Energy curves look good in general. The contact performs very well in this simulation, which
is crucial. It needs to be accurate.
For demonstrative purposes this simulation is performed in 3D. However, this example can
with advantage be run in 2D. A test simulation in 2D (not documented in this document)
was 10 times faster than this 3D simulation.

2019 © DYNAmore Nordic AB 31 LS-DYNA

Record of revisions

Record of revisions
Rev. no Release Author Comment
date
1.0 2019-03-21 Klas Engstrand Release of document.
Anders Bernhardsson

Copyright and Trademark Notice

All brands, trademarks and images mentioned are property of their respective registered
owners: LS-DYNA and LS-PrePost of Livermore Software Technology Corporation
(www.lstc.com) and ANSA and META of BETA CAE Systems (www.beta-cae.com).

Disclaimer
By using this Technical Guide, you hereby consent to this disclaimer and agree to its terms.
All the information in this Technical Guide, comprised of this document and the
accompanying simulation models, is published in good faith and for general information
purpose only. Neither DYNAmore Nordic AB nor the authors make any warranties about
the completeness, reliability and accuracy of the information in this Technical Guide. Any
action you take upon the information you find in this Technical Guide is strictly at your own
risk. Neither DYNAmore Nordic AB nor the authors will be liable for any losses and/or
damages in connection with the use of the Technical Guide.

2019 © DYNAmore Nordic AB 32 LS-DYNA