More Discussion on issues with PC-CRASH type simplified momentum solutions

Topics Related to Analysis of Motor Vehicle Collisions
Post Reply
MSI
Site Admin
Posts: 1484
Joined: Thu Jun 18, 2009 12:37 pm

More Discussion on issues with PC-CRASH type simplified momentum solutions

Post by MSI »

  • NOTE: This is a somewhat follow-up information on pc-crash. We recently posted which was posted in response to that question being posted on another forum (see the link above for details)
    Recently a colleague paid for a pc-crash training video and saw my postings to VIMEO and let me know he agreed with me....
    read below for the details.
Back on December 2, 2019 i paid $100 for and reviewed a PC-CRASH demonstration video on Vimeno called:
vimeo pc-crash validation.png
vimeo pc-crash validation.png (101.87 KiB) Viewed 56 times
  • I expected it to provide the HOW TO STEPS for determination of the required inputs for a validation. Of particular interest was HOWTO determine objectively the 'point and angle of instantaneous momentum exchange' through a scientific procedure.
  • A HOW TO should talk about HOW TO initially place the 'point and angle of instantaneous momentum exchange' and then HOW TO objectively and scientifically find the optimal location and angle, NOT simply by making arbitrary changes KNOWING the desired answer wanted or needed, but from objective scientific iterative steps to achieve a match of the evidence.
    • NOTE: the PC-Crash program and other momentum simulation programs DO NOT include damage modeling or predictions so NO feedback is included in the iterative process as to whether the impact configuration, speeds, and "point and angle of the instantaneous momentum exchange' will produce damage that matches the actual evidence.
BACKGROUND: Programs like PC-CRASH, Virtual Crash and other Momentum based solution procedures REQUIRE the user determine and specify the location and other factors associated with the assumption of instantaneous momentum exchange which is a simplifying assumption.
  • None of these vendors include guidelines for HOW TO determine the 'point and angle of momentum exchange' by an objective scientific procedure.
    OBVIOUSLY when you know the answer you want and need, it is easy to arbitrarily move and adjust the point and angle until a match of results is achieved..
    However what if you don't know the answer?
    • I expected the video to include some form of HOW TO determine the 'point and angle of instantaneous exchange'
      It did not.
    In early August 2020 a colleague who had also paid for the video and saw my comments emailed me. He agreed with my comments and so I decided to post up my FULL comments on the video which applies to ANY AND ALL 'instantaneous momentum exchange programs'.
COMMENTS made on VIMEO to PC-CRASH
  • The first of the following comments, and responses, were posted on the VIMEO website. I repeat here for convenience.
    I also emailed some additional comments which i include below.
My Initial Comments Dec 2, 2019 (sent after viewing the video):
  • Couple of comments:
    1. bottom of video obscured by the [url=VIMEO player band, need to consider that as some items lost
    2. damage: why don't you display damage measured from the full scale tests to aid in setting the 'point of momentum exchange'??
      • You mention in one test that damage needed, you have a CRASH3 damage option, why not put in the measured damage (in many full scale test results) and use the display to match vehicles for 'point of maximum engagement' and also compare CRASH damage energy to momentum energy...some form of approximate guide to scientifically adjust things
      • The Damage areas could be put together like a puzzle and then guidelines provided on how to determine where along the interface betwen the two vehicles damage lines to put the 'point of momentum exchange'
      • NOTE: From Momentum Misconceptions there is the assumption of the impulse of momentum for the CRASH simplified damage analysis as acting though the 'centroid of the damage area'. This was an assumption created for mathematical convenience as part of the development of the CRASH damage analysis portion. The centroid of damage for the two vehicles are NOT at positionally coincident and most likely will NOT work as the assumed `point of instantaneous exchange of momentum'. One needs to test this and other possible theories to come up with an objective scientific basis for determining the 'point of momentum exchange'.
    3. You rely too much on secondary impact and 0.6 restitution for those??
      • we have found a major error in your secondary impact logic (it has no clue where prior collision interactions were?) and you still haven't fixed
    4. when you KNOW the answer it's easy to pick/tweak a point/angle/friction of instantaneous exchange, but what if you don't?
      • What is a scientific basis for picking the point aside from 'it looks good'??
    5. The lengths of these videos could be reduced SIGNIFICANTLY if you left out a lot of the fumbling around and simply demonstrated how you scientifically came up with the basis for a reconstruction.
    6. what is good enough for optimization?
      • you changed the rules adding arbitrary weights and turning on/off options, but when is good enough good enough? More guidance on that would help your users.
    In most cases where i've encountered a pc-crash user i can make a single simple change in the point or angle or friction of instantaneous exchange and get the 'opposite opinion'. That is not good.
    • I hoped and expected these demonstrations would include what guidance you provide to help your users properly and scientifically come to a conclusion on a given case (FIRST point i make and you should make is look at the evidence, look at the damage location and extents and use those as a guide to the simulation)
    sorry to be a bit critical but seems a waste of my dime and my time to watch this,
    ONLY 5 min of useful information in your 90 minutes of video
PC-Crash Response Dec 3rd:
  • Sorry that you did not find the video useful. I will look into the VIMEO player band issue - thanks for bringing it up.
    Regarding the fumbling: it is a basic characteristic of using a detailed time-forward simulation to find an initial state that results in a simulation matching the given constraints. The real world does not have a closed-form solution, so you can't arrive at the initial state all in one go. If I just gave the final solution, the viewer would not learn the process of arriving at that solution.
    I would really like to discuss your other perceived problems. You should contact me directly - it is difficult to correct a "major error" if you never raise it with me, other than in a public comment section. We also hold regular live workshops - it would be good to see you at one, so we could sit down and work through your questions.
    Regarding your perceived errors and problems, I suggest that you consult the following papers that investigate the accuracy and sensitivity of impact speeds found using PC-Crash: We recently also published a study on the sensitivity of the solution to errors in the point of impact placement and vehicle engagement. It apparently isn't online yet, but here's the citation:
    • Heinrichs B, Hunter R, Moser A (2019). The effect of using residual versus dynamic crush on the accuracy of impact speed simulations. EVU 28th Annual Congress, Barcelona, Spain.
    Hope to speak with you soon,
My Comment Dec 3, 2019 on VIMEO:
  • my response got too wordy so i'll send via message on LinkedIn...
    i am very familiar with the pc-crash program (and all simulation programs) and so if any others on VIMEO want further information on this please contact me with your email address.
My Comment Dec 3, 2019 via Email:
  • I saw a post on the validation videos posted here by Dave on Linkedin and so was very interested in them.
    Of most importance was what scientific basis is used to determine any pc-crash results are good (obviously knowing the answer helps but what if a user doesn't know the answer?)
    i was hoping for more insight into what objective scientific protocol is used to determine the location, angle/friction value for the 'instantaneous exchange of momentum'.
    Your validation papers and software do not include the inputs for your validation simulations.
    If would be good if for your users you come up with some objective scientific guidelines and protocol for determination of the 'point and angle of momentum exchange' and publish/make a videos of the process/procedure.
    I am very familiar with all the papers you cited.
    I don’t have the 2019 paper The effect of using residual versus dynamic crush on the accuracy of impact speed simulations
    • If you could send it along I’d appreciate it.
    You have a very popular software product but I see and hear of too many people using and abusing the ease of the program to 'get the result they want/need' instead of objective scientific results.
    i am very familiar with simulation programs and the iterative process.
    I would recommend for future videos, particularly about validations, you include a little fumbling around but get to the point more quickly and not require revisiting due to fumbling. (think one of them you came back to 3 times!!)
    There are many options with all simulation programs and most important thing is to demonstrate how you 'blindly' got to the solution, why you made any/all adjustments, and what affect the adjustments caused, etc.
    (and would be nice if VIMEO rolled to the next in a series instead of requiring one to exit and browse to next selection)
    You also left out a very important point:
    • Before turning on ANY simulation program that requires a starting speed like pc-crash, ALL users should
      1. calc separation speeds based on distance and direction sep to rest (see ourCRASH-97 paper for ideas on that) and then
      2. do a simple 1 step momentum calculation based on the separation speeds to get a 'starting speed'.
      3. then use that speed in the simulation and tweak to test and refine the initial approximation

        For starting speeds we use the CRASH program since that is what it was invented for: as a starting point for use with the SMAC simulation program. It uses damage analysis and a simple momentum calc (it has an automatic iteration and that is what we covered in that 1997 paper but in 2003 we were able to do that with SMAC so no point anymore (the exponential increase in computer power and capabilities is simply amazing!)
        Main point is it is important that folks evaluate the evidence FIRST so they aren't simply trying to produce results to match preconceived opinions:
        They need consideration of damage location and extents and damage effects on steering and braking, and other factors.
        And roadway evidence and all that.
    Too often experts are ‘you pay, we say’ and so START with a conclusion and then try to use simulation to PROVE it!

    As far as the ‘major pc-crash error’
    • 10 years back? I was hired to evaluate one of your power users of pc-crash (who’ll you’ll know if I dig up the case file)
      They used scanned vehicles and video and all that to bury the other side in bullshit (sorry had to say that 😊)
      Something was odd about it (mighty purdy video, but something off!)
      When I reran the simulation with SMAC for same speeds etc something was very different and so something was very wrong with the pc-crash results.
      A flaw in your secondary contact allowed the vehicle to pass thru the actual structure damage area of the vehicle (it didn’t know of the initial crush from the first impact and so got a much improved moment arm for secondary contacts!)
      I documented and presented it in deposition and the case settled.
      The lawyer who hired me asked that I not publish on it any time soon (since he bought me pc-crash for that case)
      So I didn’t and haven’t and don’t plan to but realized I never let you know.
      With each pc-crash update I reload the PRO file from that case and BAM still has the same major issue!
      Sorry to blast it on VIMEO and Linkedin but far enough down most folks don’t read so probably few have seen it.
      And I’ve been meaning to bring it up with some of the pc-crash folks so did that to remind me/you I got something you need to look at.
      If anyone contacts me about that comment I will ask they contact you/pc-crash since I expect it will be a simple logic fix.
    And probably another reason I reacted was that one thing that rattled me was that you use ‘secondary contact’ for these cases with coefficient of restitution of 0.60!
    • Now on some of them there may be a secondary contact but not all and so I wonder what is happening such that your modeling technique needs that?
      Why doesn’t the initial contact handle the collision ‘good enough’ such that the vehicles don’t recontact (and again, depends on the collision)
      A good point would be to show the video of the crash test where there is a secondary contact for illustration.
      I’ll have to look through the cases you used and see if any do (I recall the RICSAC might have)
    And I don’t have ‘perceived errors and problems’…they are actual issues the software has and needs to address: the inability of it to discern in many cases which result is correct
    • I’ve been against pc-crash folks in cases like ‘wrong side of road’, that secondary contact case, and many others where a simple change to the ‘point or angle or friction’ of the secondary contact produces a significantly different opinion in the case.
      So how to tell which is correct?
    For those cases I sometimes use something like SMAC (on many I can prove their errors simply using their pc-crash program inputs with a few variations so no point in bringing in another program and all that entails),
    However in some cases using SMAC which includes collision modeling of the 150-200 ms crash forces essentially makes it a mathematical full scale test, it can help discern the issues.
    Again that is why I had hoped your validation video would include the logic and protocol for determining how to be sure the result of the program is correct and not an anomaly of the simulation.

    But it is disturbing that in a lot of cases I’ve had people on the other side using pc-crash they have misused it to get ‘the results they want/need’ instead of a scientific result.
    This is not meant to ‘trash’ the program just as a heads up that the pc-crash folks need to reel in flagrant irresponsible users by hopefully establishing a protocol to insure they are getting the most accurate results possible.
    Guess that’s all for now.
    I don’t mean to insult the program. All programs have room for improvement.
PC-Crash Response on Dec 10, 2019:
  • Thanks for getting in touch. I’m not quite clear on your specific concerns.
My Response Dec 10, 2019:
  • Not sure how you missed my specific concerns.
    FYI main points which i realized from watching your videos:
    1. Pc-Crash (and other 'momentum instantaneous exchange type programs') have NO clear scientific protocol for determining the location of the ‘point/angle/friction of momentum exchange’? (unless you know the answer you want or need) and so the program is easily abused by folks to ‘get whatever answers they want’ instead of a result with a clear scientific basis and validity.
    2. Secondly the major flaw I mentioned in the secondary contact has not been addressed or corrected. I checked and that case was 13 years ago!
    3. Third you have an issue with the primary collision since why does the secondary contact (with its flaws) come into play in your ‘validation’ runs? Were there secondary contacts in those crash tests (easy to check) or is it due to questionable choice of point/angle/friction such that the primary impact is incomplete so it requires secondary impacts?
    I was merely trying to suggest to you, and the developers, that you need to investigate and address these issues since they represent a shortcoming in the software.
PC-Crash Response Dec 10, 2019:
  • Can you be more specific about your concerns with secondary contacts? I’m not clear on the problem. The only secondary impacts in the online Validation examples are essentially sustained contacts, so the results aren’t sensitive to the specific parameters for the secondary contacts. You can try this, or I can go over it with you if you like.
    Regarding the protocol for POI and properties determination, my 20+ years of PC-Crash experience is that one cannot get whatever answer one wants. We’ve demonstrated (SAE 2012-01-0604) that impact speeds are mostly sensitive to tire friction and inertial properties, just like any momentum-based method.
    Are you asking for a closed-form solution to determine POI and impact parameters? There isn’t one, but lots of science and engineering involves searching for solutions that aren’t algorithmic. There are physical principles which are used to determine the starting point for some fine-tuning, as we work through in the videos.
    I suggest that you take some of the solutions from the course, then try find alternate POI and impact parameters that also satisfy the constraints but give you “the answer you want”. Send me the project files showing your results and let’s discuss.
My Summary: (I did not respond to their last email)
I believe my comments covered the points and areas of concern.
I will not waste more of MY time to "take some of the solutions from the course, then try find alternate POI and impact parameters that also satisfy the constraints but give you “the answer you want. Send me the project files showing your results and let’s discuss".
I mentioned several instances of serious issues of which I found that they could have asked for the PRO files to discuss.
We covered one of them in this related topic recently: And they also mention "I’ve demonstrated (SAE 2012-01-0604) that impact speeds are mostly sensitive to tire friction and inertial properties, just like any momentum-based method"
That paper should have also included a test for sensitivity to the arbitrary and subjective placement of the "point and angle of instantaneous momentum exchange"?? See the figure below for additional comments.
Then the paper could have included guidelines for placement of the "point and angle for instantaneous momentum exchange" for the validations. They did not address that issue and they did not list for each validation of full scale tests the inputs or at least the 'point and angle of the instantaneous momentum exchange'

A HOW TO should talk about HOW TO initially place the 'point and angle of instantaneous momentum exchange' and then HOW TO objectively and scientifically find the optimal location and angle, NOT from KNOWING the answer wanted or needed, but from iterative steps to achieve a match of the evidence
  • NOTE: the PC-Crash program does NOT include damage area and extent modeling or predictions so NO feedback is included in the iterative process as to whether the impact configuration, speeds, and "point and angle of the instantaneous momentum exchange' will produce damage that matches the actual evidence.

With the 2019 holidays and then the COVID issue (which continues) this get lost this thread with Pc-crash got lost under the pile of things to do.

Aug 2020: A colleague who had also paid for the VIMEO video and saw my comments emailed me to say he supported my comments. Below is a figure which demonstrates the arbitrary "Impact Point", Coefficient of Restitution" and "Contact Point Angle and MU"
  • First for the Validation Optimizations Paper 2001-01-0507 the input are listed and were set to ACHIEVE CORRELATION WITH KNOWN TEST RESULTS and NOT Optimized or changed to determine sensitivity as part of the Optimization "Validation". What if you DO NOT KNOW the results you want/need how to you objectively set these input variables to approximate the 50-150 millisecond collision by an instantaneous point and angle?
  • Second lists to right are from paper 2012-01-0604 Table 1 and the Table 2 list of variables which show they DO NOT include any test or variation on the sensitivity of the arbitrary and subjective "Impact Point", Coefficient of Restitution"(it is listed but minor/no effect? (see figure below)) and "Contact Point Angle and MU (MU listed but minor/no effect? see figure below))".
    a protocol needs to be established to objectively and scientifically determine these inputs.
Comparison of variables tested Ver2.jpg
Comparison of variables tested Ver2.jpg (436.42 KiB) Viewed 78 times
The following is a comparson of the Global Variables tested in 2002-01-0604 vs those used in 2001-01-0507:
Comparison Global variables tested.jpg
Comparison Global variables tested.jpg (389.8 KiB) Viewed 78 times
The following is a schematic of the pc-crash COLLISION MODEL with Friction Cone, Contact Plane & Crash Impact Vector displayed
PC-CRASH COLLISION MODEL.jpg
PC-CRASH COLLISION MODEL.jpg (94.29 KiB) Viewed 81 times
The following is a comparison of Impact Friction and Restitution for Various Pc-Crash References
And note for 2012 reference little or no variation indicated
Comparison IMpact friction and EPS from various references for Validation tests.jpg
Comparison IMpact friction and EPS from various references for Validation tests.jpg (299.58 KiB) Viewed 70 times
The following is from the 2012 refrence
restitution and impact friction variations.jpg
restitution and impact friction variations.jpg (64.93 KiB) Viewed 70 times
Question? Comment? Please email us
MORE TOPICS see:Forum Index & McHenrySoftware.com
(c)McHenry Software, Inc ALL Rights Reserved.
Post Reply