Simulation Analysis: Effects of Input Variations

General questions on the SMAC, msmac3D and other Collision Simulation programs
Post Reply
MSI
Site Admin
Posts: 1719
Joined: Thu Jun 18, 2009 12:37 pm

Simulation Analysis: Effects of Input Variations

Post by MSI »

New: Use this link to go to the latest post on this thread
This is response to the question posted on INCR about Momentum which morphed into a question on simulation:
  • "I wonder how many of the simulation program users have played with some of the unmeasurable factors that go into the analysis, such as tire cornering stiffness, suspension parameters, inter-vehicle friction, etc. to see what the actual effects are? There is a tendency to blindly use default values without ever really knowing what the effects of these parameters are"
OUR RESPONSE:
  • Main points:
    • What was mentioned as 'unmeasureable" are measureable
      And then to point out the need to differentiate between simulation programs:

SMAC Type Simulation programs
  • First i'd mention that what you mention as 'unmeasureable" are measureable.
    • Cornering stiffness and suspension parameters are by default approximated however they can easily be changed via inputs based on measured values or to test sensitivities. In general typical approximated values for these variables correlate well in validation tests.
    • The inter-vehicle friction coefficient is a variable to represent and reflect how "sticky" the vehicle interactions are (how sticky is their interaction?) and the default approximation of 0.55 SHOULD be tested for sensitivities if an impact configuration and interaction seems to indicate more or less than 'typical' type interaction
      • msmac3D is also the only version of SMAC which includes the ability to model a SNAG option which can be used to model tensile forces in the event the collision interaction goes beyond simple collision forces and coulomb friction (say for example if a component like a tire gets ripped out during the collision interaction?
    .
Instantaneous Exchange Momentum Simulation Programs
  • When you get into Momentum simulation programs like PC-CRASH, Virtual Crash, Planar Impact Models, and others the inputs for suspension and tire properties are measurable, however...
    • Some inputs are NOT measurable, are subjective and arbitrary with no objective input instructions:
      For example, HOW to set the LOCATION, the ANGLE and the LOCATION for the INSTANT of the momentum to be "instantaneously exchanged"
      How do you best approximate the 50 to 150 or more instants during a collision interaction as a SINGLE INSTANT?
      • We paid to watch 401 Staged Collisions: Validation 401 a pc-crash instruction video. We were hoping to find objective guidance on the inputs for the momentum exchange approximation. There was nothing about how to objectively set the instantaneous exchange. They merely adjusted it to produce their desired/known result in the video.
        • Simplified momentum equations and programs require the user to pick a point and angle/direction to instantaneously exchange the momentum of a vehicular crash.
          Any vehicular crash takes 50-150 milliseconds (or more) for the collision interaction which these programs/equations consider a SINGLE 'instant'.
          When applying a momentum solution it is subjective and random HOW a user sets the LOCATION and the ANGLE/DIRECTION for the INSTANT of the momentum to be "instantaneously exchanged"

          Please see
        Until then BE SURE to test moving and changing your assumed point and angle or however characterized for the INSTANTANEOUS MOMENTUM EXCHANGE to see how much it changes your results. Sometimes minor changes can make dramatic differences.
        • In our experience in evaluating applications of momentum based programs which formed a basis for "expert" opinions we have been generally able to make minor changes/adjustments in the subjective inputs and "prove" an alternative opinion.
Main point is when talking about vehicle crash simulation programs be sure to differentiate between There IS a MAJOR difference!

When Raymond R McHenry invented SMAC (and the other programs he invented) he brilliantly found the right combination of scientific sophistication and modeling simplicity for the time and budget constraints. And the continued prevelance of so many of his ideas and programs is a great demonstration of his brilliance.His analytical modeling and inputs were and are based on his extensive experience in vehicle suspension design.

Please give our mSMAC3D program a try, 2020 year end specials in effect.

NEW!! LEASE pricing of msmac3D Software!!
FOR MORE TOPICS see:Forum Index & McHenrySoftware.com
Question? Comment? Please email us (all communications considered confidential)
(c)McHenry Software, Inc ALL Rights Reserved.
MSI
Site Admin
Posts: 1719
Joined: Thu Jun 18, 2009 12:37 pm

Re: Simulation Analysis: Effects of Input Variations

Post by MSI »

Another response on the INCR thread included the following statement/question on Simulation:
  • Q: For those who are doing the simulations, correct me if I am incorrect on this, but if you do a simulation, and it overlays the physical evidence, vehicles hit in the correct locations, stop in the correct locations, etc ... havent you, in effect, "checked" the simulation? isnt part of running the simulation multiple time4s with multiple variables is to get that "overlay"?
RESPONSE

The first response is what we ALWAYS tell folks:
Use simulation to test and refine your momentum and damage analysis estimates.
  • Why? Because you want to form a general opinion and then use simulation to test and refine speeds/results.
    Also so you make sure you're not blindly applying simulation.
    And note this is exactly what we did a while back**
The response to your question is Yes, and No.
First, for ANY simulation program:
  • Response is YES as the main challenge is input parameters:
    • Do you have the weights, moments of Inertia, dimensions correct?
    • What did you use for ground friction coefficient and on what basis?
    • What did you use for inter-vehicle friction coefficient:
      • if lower or higher than typical, why?
    • As far as specific input variables like suspension stiffness, bump stop locations, etc
      • if using the defaults, OK since those generally are used in the validations. If using values different or if you have measured quantities, have you input those specific measured values and re-iterated to get a match and how does is compare.
        • I say re-iterated because like the butterfly in the amazon ANY changes you make to the inputs will probably make some change to the specific simulation match with evidence HOWEVER minor tweaks of impact positions, speeds etc normally brings the results once again into agreement with the evidence and then you can compare the results.
          • In published tests on SMAC (will add a citation) where researchers varied things like the moment of inertia and other measurable inputs and compared to default inputs the final results, once iterated to a "best match" of the evidence, both trajectory and damage, the results are essentially the same.
  • Now onto the WHY sometimes a YES and sometimes a NO answer:
    • The answer is YES for SMAC type simulation programs.
      • These programs include the simulation of forces during the entire collision interaction, every millisecond for 50 to 150 or more milliseconds, and include the forces and moments and therefore the areas and extents of damage as the vehicles crush and interact: you check how you compare to the measured damage and trajectories and try to match as best as you can the vehicle travel and areas and extents of crush (particularly when a sideslap impact)!
        • The actual vehicle stiffness values and inter-vehicle friction coefficients are of somewhat secondary effects
          • as long as within +/-25% or so of value, obviously using crazy numbers WAY out of range will make significant differences
          Variation of these inputs will also affect the responses of the vehicles. So if you match the trajectory, the damage areas and extents you have a good result you will have a good test/refinement of your momentum solution.
    • The answer is NO for Instantaneous Exchange Momentum Simulation Programs like PC-CRASH, Virtual Crash, Planar Impact Models.
      • There is NO FEEDBACK on areas and extents of damage.
        Since they include 'instantaneous momentum exchange' the user has to subjectively and somewhat randomly pick a "point of momentum exchange"
        THERE IS NO FEEDBACK ON WHETHER THAT POINT WILL PRODUCE THE PROPER AREAS AND EXTENTS OF DAMAGE
        Meaning whatever you pick to approximate the 50 to 150 millisecond or more collision iteration as A SINGLE INSTANT you have no idea whether it will produce damage areas and extents that match the particular collision you are reconstructing.
        THAT is an issue with those types of programs:
      In our experience in evaluating applications of momentum based programs which formed a basis for "expert" opinions we have been generally able to make minor changes/adjustments in the subjective inputs for the point and angle of "instantaneous momentum exchange" and "prove" an alternative opinion.



**Footnote: A While Back: As part of our Automatic Iteration of SMAC research: we automatically used CRASH (both momentum and damage analysis) to get preliminary speed results and then used those speeds to start the SMAC Iteration/optimization process. All hands off!
  • This was a demonstration that CRASH, with momentum and damage analysis solutions, and which was created for initial approximations of SMAC, can the be used to start a SMAC analysis and if you (or an auomated process) iterates to a good match of the evidence (both trajectory AND damage) your results will be within +/1 10 of the truth! This is what we published in our 2003 SAE paper.
NEW!! LEASE pricing of msmac3D Software!!
FOR MORE TOPICS see:Forum Index & McHenrySoftware.com
Question? Comment? Please email us (all communications considered confidential)
(c)McHenry Software, Inc ALL Rights Reserved.
Post Reply