Virtual Crash Accident Reconstruction Software | Hacker News Reader

Virtual Crash Accident Reconstruction Software

vcrashusa.com

52 points1970-01-0117 comments

I want to point out, as someone who spent several years as a senior CAE specialist at an automotive OEM doing finite element analysis of automotive body and chassis structures- the type of mostly-rigid-body physics on display here is not the kind of simulation at all that is done at the vehicle design level for say, NCAP testing, where the actual structures and explicitly modeled and the structural dynamics are computed using an FE code such as LS-DYNA. The 'soft-body physics' shown in the above linked kind of reconstructions aren't relying on incredibly realistic computation based on the vehicle crumple zones, because this kind of tool simply would not have access to the level of model, material, and other detail used in an actual car company; These things are several orders of magnitude more complex. So for instance, the 'crumpling' and denting of the fenders and bumpers shown in these renders isn't, per se, very real, comparatively.

I'm not saying there are not uses for this tool- accident reconstruction often can be very 'spherical cow'. But I just want to point out this type of engine is much closer to GTA-5 than it is to, say, NASTRAN.

This software is obviously intended to demonstrate or investigate the vehicles' locations and velocities leading up to a collision, not determine how damage occurs once the collision is in progress.

Yes, I just wanted to point that out. But in addition, in certain impacts, the large deformations and energy absorption differences in modern vehicles would result in significant differences in the crash kinematics compared to these types of simulations. For instance, small-offset impacts are handled very differently between different crash systems even on relatively similar looking vehicles. Some vehicles approach small-offset via substantial A-post reinforcements and taking the tire impact very linearly to absorb energy, while others now use a more kinematic approach of deflecting the tire outwards and pushing off a more angled side sill and this results in a very different movement of while just after impact. In crash analysis involving multiple vehicles or impacts where there is a secondary collision (which is shown in some of the demos there), the relative energy absorption, tire parameters, knowledge of bumper beam engagement, etc completely can change the results. I question if the people using these tools in all situations always understand the full physics.

Looks impressive, but how does this compare to BeamNG?

https://www.beamng.com/

I don't know either, but they have a demo reel at https://www.vcrashusa.com/gallery/ which was interesting to watch.

Exactly my thoughts.. it looks like the physics you find in most half decent games.

When I was involved in a motorcycle accident (oncoming car made a left turn which resulted in me flying over the bonnet of the car, luckily at ~50km/h and fully equipped), I did try to come up with additional proof that I was not speeding, based on basic physics equations (the classic exercise of "where does an X kg projectile being thrown at an angle of Y degrees and initial velocity of Z m/s land").

By varying estimates of input data (mainly, "takeoff" angle) I could "prove" whatever was needed. An insurance expert also told me that usually the inputs to such equations are so fuzzy that another expert can easily refute the conclusion.

Wish I'd had this a year and a half ago when I was trying to explain to my insurance company and the local police supervisor that just because they don't know how to calculate a maximum speed from the speed at impact, the maximum distance of travel, and the ABS deceleration rate for a specific vehicle, it doesn't mean nobody can!

They probably would have mocked you and said that you are using a "video game" to illustrate the accident.

Attorneys both public and private pay through the nose for software like this to make their case to judges and juries. I don't think you'd face ridicule.

If the other driver hadn't had the same insurance company, I'm sure they would have put forth even a basic level of effort to have their own reconstruction experts look at it.

My state's insurance commissioner had to get involved because of how they were treating me. It got ugly.

Whatever tech they've got has to be better than the incorrect conclusions the local sheriff made on my brother's accident. They way over estimated his speed by making an easy mistake. They, of course didn't want to hear about that.

In this case I'm not sure software would help, because bad measurement in likely mean bad data out.

It was good to see firsthand that people who you think might be knowledgeable on a topic are sometimes nearly clueless. I don't thing software can make this area worse.

This looks interesting...

Does anybody know if there's any open-source software in this field? Or any attempts to develop a community or similar software there?

Wow there is no annual fee. The website is filled with duplicated keywords and other annoying seo tricks. I see no videos, but a huge amount of claims about their extreme precise pjysics and stunning graphics. Sorry I don't buy it .

This is what NTSB and other safety boards all over the world need to put humans & ADAS systems against, to see how each would behave. Like a leaderboard for safety instead of long held beliefs.

It's a shame they use 'accident' for collisions. The motor industry back in the 1920s pushed hard for RTCs to be referred to as 'accidents' and to move blame away from the drivers responsible. It's even less appropriate for simulations where the collisions are explicitly designed.

See https://www.vox.com/2015/1/15/7551873/jaywalking-history

Reply on news.ycombinator.com