HOWTO Calculate CRASH3 A & B Coefficients, AKV (for SMAC)

[MSI]

This is a very preliminary writeup

• NOTE: If you don't have our msmac3D software, Please also see the post below which includes links to directly get the NHTSA crash test data and HOWTO calculate the A&B Coefficients
  

For 3D Graphics interface and/or Medit3D Tool menu:
We have a link in the 3D Graphics HELP Button to direct link to the NHTSA CASH Test database

lsqfit and crash test databse on Help menu.jpg (19.38 KiB) Viewed 1670 times

• If from Medit3D:
  • Tools menu of msmac3D we have the Lsqfit - A&B Coefficient Calculator (same program as discussed below)
  • Menu item Databases and select NHTSA Crash Tests which brings up the test selection menu (on Left see below) which allows you to select Make, then Model, Then Year
  and all Tests are then listed

Select a Test from the list of tests for the make/mode/yr/etc and then:

• From the datasheet take the average crush, test weight,
  From the data on the sheet Record the following (numbers listed in red):
  1. the average crush
     • C1,C2,C3,C4,C5,C6
       Add these 6 crush measurements and divide by 6
  2. the Veh Speed Test Impact Speed
  3. The Weight aka Test Weight
  4. The LENCNT aka Length of Contact
     • for some reason they don't present in inches in () like other nums, so simply divide by 25.4
  5. The Test# aka Test number
  6. Note that if photos, videos, reports available these buttons will activate you can review them/collect them/etc

  

  NHTSA Crash Test Database Dialog v3.3.jpg (254.44 KiB) Viewed 1671 times

  We also recommend you check the Sisters and Clones Database from Greg Anderson
  • We have an older version of that Sisters Clones Database available if you pop out to Medit3D Grid: Database Menu->

    

    Medit3D Grid Database Menu.jpg (28.45 KiB) Viewed 1670 times

    It is probably 10 or so years old so for vehicles made in past decade and for refinements of the database please consider a subscription to the Sisters and Clones database

Then select the LSQFIT Program which brings up a file selection menu (to load prior tests)

• An instructional video on the LSQFIT will be posted and linked here soon

The file menu will have an array of files (note some of these will be in the ver 3.31 update to be updated shortly

lsqfit file selection menu.jpg (73.54 KiB) Viewed 1668 times

We selected the 1985 Ford Escort

• From the Prasad SAE 900412 Report A Study Using the Repeated Test Technique
• one of the datapoints was bad (50 mph/50 inches) or simply skewered the results (see below for details)
• We added 3 more tests from 81-84 which the sisters & Clones Database said are the same body

Here are the inputs we put into the program:

output of pressing enter.jpg (112.43 KiB) Viewed 1669 times

And the chart created and A,B CRASH3 Coefficients and also AKV Stiffness were as follows:

prelim display of Ford Escort tests with 3 plus minus outlier.jpg (101.81 KiB) Viewed 1669 times

CRASH was developed as a preprocessor for the SMAC simulation program
To Help NHTSA with their statistical NASS study get starting speeds for SMAC
NHTSA Later adopted CRASH since 'the errors will wash out in the large sample size' and that they were 'looking for trends'
However now CRASH and damage analysis have been used in individual crash reconstrctions
Reasons to be cautious when using CRASH or other damage analyses:

• Most CRASH or damage analysis techniques are based on a single data point and assumption of 'no-damage' intercept
• Be sure to check the source of the data for any crush coefficients you use
• below we present the data for a 1985 Ford Escort
  • Some of the data was first presented by Prasad SAE 900412 Report

and now we show in more practical format:
First a summary of what would be the Crash B coefficient (and corresponding AKV for SMAC) with the data as presented and then dissected:

• Note that these charts and comparisons demonstrate how depending on a single crash test and assumption of 'no damage' intercept may produce very different results for a vehicle's CRASH3 A&B Coefficients virtual crush stiffness .

1985 Ford Escort Summary.jpg (94.38 KiB) Viewed 1667 times

Now here are the six items presented in graphical form so you can see that depending on the amount of crush, the results can have major differences.
And realize that most Crush Coefficients sold are based on a single Crash Test at approx 35 MPH and then an assumed 'no damage' intercept.
More write up on this will follow in the next week or so
this first collection from the Prasad SAE 900412 Report has an outlier (the green dot on chart behind data) which skewers the fit away from the other points:

1. Prasad 6 Ford Escort Tests.jpg (52.14 KiB) Viewed 1655 times

Here is what the fit looks like if you remove that outlier:

2. Ford Escort Prasad less outlier.jpg (51.59 KiB) Viewed 1655 times

Next we added the extra crash tests (sisters/clones of the test vehicle) but again with the outlier (green dot on chart behind data)
Note once again the fit is skewered

3. Ford escord Prasad plus 3.jpg (57.32 KiB) Viewed 1655 times

And now with that outlier removed, note how much better the least square fit matches the data

4. Ford Escort Prasad plus 3 less outlier.jpg (55.1 KiB) Viewed 1655 times

Now we simply go with 'what if only single tests available (we found 3):

5. Ford Escort 3 tests.jpg (49.3 KiB) Viewed 1655 times

We do it again with a low speed test from that Prasad paper:

6. Ford Escort 3 tests plus low speed test.jpg (52.55 KiB) Viewed 1655 times

These next 3 are what if only 1 test available for the Escort?

• And which test?
• and how much do the A/B fits change?

here are all 3 side by side and note the change in the A & B Values, depending on which single test the fit is made from?

• This demonstrates the sensitivity and possible changes in the A & B CRASH3 Coefficients depending on test results
• and why it is important to use MORE than a single test and an assumption for the 'no damage' intercept!
  Below you see depending on which individual test is chosen (or if only one ran, which one?)
  the CRASH3 A & B Coefficients would be:
  If based on all 3 tests:
  • NHTSA 3 Test: A = 213 lb/in^2 B = 58 Lb/in
  If Based on all 3 tests with a low speed test:
  • NHTSA 3 Test and Low speed Test: A = 207 lb/in^2 B = 64 Lb/in
  If based on eachg individual test:
  • NHTSA Test #407: A = 188 lb/in^2 B = 55 Lb/in
  • NHTSA Test #452: A = 221 lb/in^2 B = 63 Lb/in
  • NHTSA Test #509: A = 190 lb/in^2 B = 70 Lb/in

test 6 7 8 single point fits.jpg (49 KiB) Viewed 1622 times

Below are the individual fits:

7. Ford Escort test num 407.jpg (44.86 KiB) Viewed 1655 times

8. Ford escort Test num 452.jpg (46.85 KiB) Viewed 1655 times

9. Ford Escort test num 509.jpg (44.84 KiB) Viewed 1655 times

[MSI]

A recent posting on another forum included a vendor who sells CRASH3 A & B coefficients

• I have removed any vehicle specific or vendor references to make this more generalized information:

The following values for the CRASH A&B crush coefficients were posted up by a vendor in response to a request for information on the A & B Coefficients for a specific vehicle:

• They wrote...
  • "Depending on how you calculate the A-B values:
    • A= 441 B= 143
    • or
    • A=689 B=348"

Our response to the posting:
Preliminary thought:These wildly different values either demonstrate a flaw in the calculation procedure or something else wrong with things...
This obviously indicates a wild variability in at A & B Crash3 Crush Coefficients so one wonders HOW? WHY? WHAT data?

• They said "Depending on how you calculate the A-B values"
  • Are there different ways to calculate A & B which can skewer the results?
  It appears basically a 2.43 to 1 difference for the Crash3 "virtual" Linear Force-Deflection Rate?
  • For B the Crash3 "virtual" Linear Force-Deflection Rate: 143 lb/in vs 348 lb/in??!!
  • For A the CRASH3 Unit width Force Level for ‘no residual damage’: 441 lb/in^2 vs 689 lb/in^2??!!
  • And so obviously any analysis will be skewered by which one is chosen and used? Or a wide range of results.
  I looked at the test mentioned, The test had an average crush of approx 10" at 35 MPH!
  and if A&B values calculated it indicates a quite high value for B.
  But i also looked at the videos for the test!
  • The 35" front overhang is crushed to wheels so more like 20" crush??
  • is that how they got such wildly different results?
  And also see elastic response of tires into walls which was originally posted to the NAPARS facebook page and i mentioned in my followup post:
  • On Restitution indicated in crash tests
  I also checked the sisters/clones and they seem to indicate that 2009-2017 share the same body (quick look, obviously double check)
  
  These tests may all be a product of the 'SUV tire/wall interaction' elastic rebound which makes the frontal structure seem extremely stiff (once the tires encountered and if they don't deflate you get a large amount of elastic rebound and reduction in residual crush.
  
  Only reason i post this up is that folks who use damage analysis need a explanations for wild variations in crush stiffnesses.
  
  The main points:
  • Before accepting ANY damage analysis crush values for A&B be sure to look at the tests.
    • NOTE Feb 2022 updated link:NHTSA has a great crash tests resource online to be able to look at the reports, pictures, videos
  • Below is our calculations of the A & B for the test vehicle, with adjustments of the average crush of 0,1,2,4,6,8,9,and 10 inches.
    Best match of the two Vendor posted values is:
    • use of 5 MPH assumption for 'no damage intercept'
    • For the Higher stiffness values (A=689/B=348) the Vendor adjusted the measured average crush by 2.5 inches
    • For the Lower Stiffness values (A=441/B=143) the Vendor adjusted the measured average crush by 9 inches

    

    compare adjustments of test results with vendor A and B.png (86.61 KiB) Viewed 1611 times

  Here's the sanitized test results (VIN & Veh identification removed to keep this topic more generalized):

  

  sanitized veh data.png (151.34 KiB) Viewed 1557 times

[MSI]

In response to this thread got the following question which i answer in a separate thread:

• Can you use Maximum Crush for the creation of CRASH Crush Coefficients?

[MSI]

A recent post on the crash forum asked:

• Q: I am a newly certified reconstructionist. My training was conducted by a local police department with most of the topics taught by IPTM. One concept we did not dive into very much was calculating stiffness coefficients for crush profiles. A and B values are the ones I am struggling with. I can get G values based on A and B values just fine. Can anyone point me in the right direction of being able to do this?

A: In response we posted a link to this topic and we also realized we need to add links to getting the raw NHTSA data

• • From our forum Specs, Accel, Braking, Crash Tests. Motorcycles
    • Here's a link to that:
    • Crash Test Databases: NHTSA and Transport Canada
      Enter NCAP and Make/Model & Years

      

      NCAP.png (83.2 KiB) Viewed 353 times

also need

• Least Square fitting
  • Many online, Here's one:
    • Math is Fun Least Squares Calculator
    • And a step by step too Least Squares Regression
    Be sure to include the 'no damage intercept as a point (either 4 to 5 mph and 0 crush unless you have low speed data for the vehicle)
    see below a duplication of a test from above etc

    

    LSQ.png (188.18 KiB) Viewed 356 times

Note to go from the fitted values to A& B
use the conversion factor as listed here:

• Crush Coefficients Conversions: CRASH (McHenry), Campbell, WinSMASH (Prasad)

on the other forum they reminded me i didn't include that and the mentioned the IPTM recommended formulas which are the same:
DeltaV= change in velocity
b0= zero crush speed (in/sec)
L= overall length of crush (in)
W= weight (lbs)
G= gravity (in/secsquared)

b1= (DeltaV - b0)/ L = (1/sec)
A= (W x b0 x b1)/ (G x L) = (lb/in)
B= (W x b1squared) / (G x L) = (lb/ insquared)

to check your work, work through the examples in this first part of this thread to check your work since they include all the information you need to go from Crash test to A & B.
Thanks to the posted on the other forum for reminding me i left out the last step.

Please let us know if any questions.