Wikipedia:Reference desk/Archives/Science/2018 December 19

= December 19 =

How many floating point operations and TB of storage would be needed for a good car crash test?
Is it a lot? Then a container ship crashing into another or something complex like that would be computationally expensive as hell. Or are auto regulators just crashing them in real life out of caution? Sagittarian Milky Way (talk) 00:09, 19 December 2018 (UTC)


 * You are talking about simulating car crashes? Why would anyone do such a thing? The expense would be far higher than crashing a car would be, even if computing resources were free. Just entering in the data for the shape and thickness of every part would be far more than a car. Then there is the liability cost; even if the simulation was perfect, would a jury believe that with the lawyers on the other side trying to confuse them? Would every jury believe it? --Guy Macon (talk) 01:37, 19 December 2018 (UTC)


 * In fact, IIHS wrote this year-2000 op-ed on the topic of the NHTSA Driver Simulator: Advanced driving simulator is costly, value is questionable.
 * Of course, computer modeling is important, but actual crash tests are also important.
 * Nimur (talk) 01:44, 19 December 2018 (UTC)
 * Those links seem to be about simulating the driving experience, not the crash. --Trovatore (talk) 20:28, 20 December 2018 (UTC)
 * For a few years I was employed under the same manager as was responsible for computer analysis of car crashes. The results are good enough these days that the physical tests are done to verify the CAE runs. The calibration for the airbags etc is done from the sensor signatures generated by the CAE test. Yes it is lots of data. Yes the computer tests take a while to run on supercomputers. The fidelity of the model used is a mesh size of about 1-3mm. The main programs used are LS Dyna and Nastran. The physical tests that are done by NHTSA and IIHS and NCAP and so on are to verify that the auto manufacturers are telling the truth and to provide an absolute number. The error is rarely more than 0.5 stars. These are crashes into concrete blocks, car vs car is rarely modelled. A test has to be reliable, repeatable, and, representative. Car vs car scores 1/3, car vs block scores 2/3. In practice car vs block seems to be pushing the casualty rate the right way. Greglocock (talk) 04:43, 19 December 2018 (UTC)
 * How expensive are individual runs? Could you test those two, car vs. standard guardrail, car versus highway divider, car versus moose, car versus jogger etc. and start getting composite pictures of overall safety?  (I realize that economically the goal would be to get the best score on the test and never mind the real world, but I mean, potentially...) Wnt (talk) 15:28, 19 December 2018 (UTC)
 * Not amazingly expensive once you have the model. I don't know if they use the same models when they develop pedestrian safety, I'd have thought so. I'm not sure how to implement a composite picture other than the current approach which gives a single figure score made up of several categories. Greglocock (talk) 21:31, 19 December 2018 (UTC)

Why do we shake when we're angry?
Many people shake when they're angry. I looked for this question on Google and I saw many answers that said that it's because the activation of the sympathetic system (by the amygdala) which cause to flow of adrenaline. But I really didn't find the relation between the adrenaline and the shaking of the body muscles. So assuming that the shaking is from the adrenaline, I'm asking what's the relation between them? --ThePupil (talk) 00:12, 19 December 2018 (UTC)
 * Increased blood pressure might also figure into it. ←Baseball Bugs What's up, Doc? carrots→ 00:43, 19 December 2018 (UTC)


 * See The effect of adrenaline on the contraction of human muscle -- one mechanism whereby adrenaline increases the amplitude of physiological tremor. --Guy Macon (talk) 01:56, 19 December 2018 (UTC)