The accuracy of the CRASHEX version of CRASH, which uses a Newtonian rather than a Cartesian treatment of impact, is compared with that of a substantially “original” version of CRASH, both evaluated with respect to EDSMAC computer simulations of the RICSAC series of staged, partially braked collisions; and with similar data for EDCRASH, as published but recomputed.
Special problems inherent in treatment of impact with a stopped vehicle, quasi-head-on impact, failure to separate after initial impact, and spin truncated by side slap are discussed.
Appropriate use of the statistical confidence interval in single-event reconstruction is also discussed, as are anomalies in previously published standard deviations for EDCRASH applied to the RICSAC data.
The study shows that the Cartesian simplification incurs dominantly large errors, a finding not generally recognized. For complete and accurate reconstruction, Newtonian treatment, as now provided in CRASHEX, is requisite.