Vehicle change in velocity, often referred to as delta-V, is a widely used measure of crash severity. The National Highway Traffic Safety Administration (NHTSA) uses the WinSMASH computer code to estimate delta-V in several of its in-depth crash databases. This study examined the accuracy of WinSMASH longitudinal delta-V estimates in 140 rear crashes by comparison with direct measurements of delta-V recorded by EDRs. In the entire dataset, WinSMASH longitudinal delta-V was 4.5% lower on average than the delta-V based on direct measurements of acceleration from EDRs.

WinSMASH accuracy varied widely by the degree of overlap. WinSMASH was only 2% lower than EDRs in full engagement rear crashes, but 18-22 % lower in small and moderate overlap rear crashes. WinSMASH accuracy appears to be a function of struck vehicle type. WinSMASH delta-V estimates were only 3-4% lower for rear-struck LTVS, but were 10% lower than EDRs for passenger cars. The lower accuracy of WinSMASH car delta-V estimates did not appear to be the result of LTV-to-car or car-to-LTV structural incompatibilities. The lowest agreement between WinSMASH and EDRs occurred in car-to-car crashes.

This paper is the first of a kind assessment of rear-struck vehicle delta-V reconstruction accuracy when using energy- based methods, e.g. WinSMASH. In our dataset, WinSMASH delta-V estimates were within 5% of EDR recorded direct measurements. However, this level of accuracy was unevenly distributed across both crash overlap and struck vehicle type, and suggests opportunities for further improvements in WinSMASH accuracy.