close


Features Every LS-DYNA Simulation Must Have


LS-DYNA is a general-purpose simulation software. This is a good thing for analysts since it provides tremendous flexibility in simulating a wide range of events. The downside to this is the fact that every simulation requires ‘best practices’ to ensure it meets the quality and accuracy requirements of th event that is being simulated. Additionally, LS-DYNA is super backward-compatible. You can run legacy models that could be a couple of decades old using the latest release with no change to the input file. Backward compatibility however requires that default values that was found to work in the past may no longer be applicable today as better alternatives may be available. This post highlights some important non-default settings that must be present in all simulations.

1. *CONTROL_ACCURACY

  • INN, Invariant node numbering
  • OSU, Objective Stress Update (Complete model or for a subset of parts)

2. *CONTROL_SHELL

  • ESORT, Full Sorting of Triangular Elements
  • THEORY, use BWC as default element formulation
  • BWC, Warping Stiffness
  • PROJ, Full Projection
  • NFAIL1 and NFAIL4, Remove and Proceed when Element Quality Deteriorates

3. *CONTROL_CONTACT

  • SSTHK, Actual Shell Thickness for SINGLE SURFACE contact
  • RWPNAL, Treat rigid bodies impacting RIGIDWALL
  • IGNORE, Ignore initial penetrations
  • SPOTTHIN, Ignore interaction of segments used in spotwelds

4. *CONTROL_BULK_VISCOSITY

  • TYPE = -2 to include shells

5. *CONTROL_SOLID

  • ESORT , Auto sorting of tetrahedrons, wedges to use improved element formulations

6. *CONTROL_OUTPUT

  • NPOPT, Suppress nodal coordinate and element connectivity information
  • NEECHO, Node and element printing

7. Strain-rate Effects
Viscoplasticity is now available to almost all material models that has strain-rate effects. Using VP=1 greatly reduces the noise in the simulations and also provides a more accurate stress as a function of strain-rate. It must be warned that VP=1 requires an accurate input such that the convergence for the shells (plane-stress treatment) is achieved.

8. Airbag Contact
Older versions of LS-DYNA allowed a time-dependent of thinning the airbag thickness to allow interpenetration to exist in folded bags. This is no longer necessary as the segment-to-segment contact (SOFT=2) takes care of this automatically with improved contact treatment.

9. *DATABASE_GLSTAT
This is an important file that must be reviewed for every run.

10. *CONTROL_HOURGLASS
Use type 2 (Flanagan Belytscho Formulation).

11. *CONTROL_ENERGY
HGEN =2 to account for hourglass energy balance.

The above list will be updated so do check back.

  • Heguo OU says:

    Dear Suri, by ur experience & pratices, can u give me some recommened steeings for these keyword?
    I am a auto engineer-CAE, just on the begin stage of studying crash simulation.
    TKS.

  • Kevin says:

    Hi Suri,

    regarding to the hourglass control, you recommend here IHQ=2 (Flanagan Belytscho Formulation). But at your old post in “crash_guidelines for Crash analysis”
    http://blog.d3view.com/2006/11/01/general-guidelines-for-crash-analysis-in-ls-dyna/

    IHQ=4 & QM=0.03 was recommended for metal and plastic parts.

    Are there any different thinking between them?

    – Normally, hourglass energy should be no more than 2~3% of the total energy plotted by glstat (right?). Once, after I was not successful with under-integrated shell elform for shell, I turned to elform=16, then the hourglass energy was gone, but I got unreasonable element failure at some indirect impact area (including triangle element). How could that be?

    – Regarding to VP=1, is that an extrapolation of the present curves/table or a fairly different approach? The LS-DYNA manual at MAT_024 indicate the effect would be dramatic, are the responses still reliable because of that?

    Best regards

    Kevin

  • Dear Suri,

    I am running one of full vehicle impact analysis with rigid wall ( 150 ms run)

    When I see glstat energy balance, there is sudden drop in kinetic energy curve between 47 – 48 ms. But there is no currusponding increase in global internal energy curve. Total energy curve is fairly constant throughout duration of run. also no other misc. energies in glstat shows the energy drop/hike in that time zone. But the analysis however continues and finishes 150 ms of run successfully

    I have also checked IE / KE / HE for all parts in matsum plot. I dont see any abnormal energy curve behavior in that time slot

    Now how to trace this sudden drop in KE energy in our global / partwise energy balance ??

    FYI. I have used element fail option for few plastic parts in the model but there is no element failure taking place after 30 ms run as seen in message file

    I have even extracted B Pillar pulse and I dont see peaks or troughs in pulse also in that ime slot

    I have discussed this problem with ls dyna suppliers here in India but nobody could really help to resolve the issue

    Also for your information I am using same model in front ODB impact there the model finishes entire run without any such problem

    I am writing to you as last resort. Please help me out

    Best Regards

    Ganesh Shanbhag

  • Suri Bala says:

    Ganesh,

    Some things to consider based on information you have provided.

    1. Ensure all energy balance options are turned on in CONTROL_ENERGY
    2. You could use GLSTAT to output model mass as a function of time.
    3. Ensure no nodes exist with initial mass and velocity. When these hit any rigidwalls, the KE could be misleading.
    4. Some basic checks of momentum at time zero.

    Suri

  • Thank you..

    We could identify the source of the problem

Leave a Reply

Your email address will not be published. Required fields are marked *


*