Objective: In the widely used National Automotive Sampling System (NASS)-Crashworthiness Data System (CDS) database, summary metrics that describe crashes are available. Crash angle or principal direction of force (PDOF) is estimated by the crash examiner and velocity changes (ΔV) in the x- and y-directions are calculated by the WinSMASH computer program using PDOF and results from rigid barrier crash testing combined with deformations of the crashed car. In recent years, results from event data recorders (EDRs) have been added to the database. The aim of this study is to compare both PDOF and ΔV between EDR measurements and WinSMASH calculations.
Methods: NASS-CDS inclusion criteria were model-year 2000 through 2010 automobiles, frontal crashes with ΔV higher than 16 km/h, and the pulse entirely recorded in the EDR module. This resulted in 649 cases. The subject vehicles were further examined and characterized with regard to frontal structure engagement (large or small overlap) as well as collision properties of the partner (impact location; front, side, or back) or object. The EDR crash angle was calculated as the angle between the lateral and longitudinal ΔV at the time of peak longitudinal ΔV. This angle was compared to the NASS-CDS investigator's estimated PDOF with regard to structural engagement and the collision partner or object. Multiple linear regression was used to establish adjustment factors on ΔV and crash angle between the results calculated based on EDR recorded data and that estimated in NASS-CDS.
Results: According to this study, simulation in the newest WinSMASH version (2008) underestimates EDR ΔV by 11 percent for large overlap crashes and 17 percent for small overlap impacts. The older WinSMASH version, used prior to 2008, underestimated each one of these two groups by an additional 7 percentage points. Another significant variable to enhance the prediction was whether the crash examiner had reported the WinSMASH estimated ΔV as low or high. In this study, none of the collision partner groups was significantly different compared to front-to-front impacts. However, with a larger data set a couple of configurations may very well be significantly different. In this study, the crash angle denoted by PDOF in the NASS database underestimates the crash angle calculated from recent EDR modules by 35 percent.
Conclusion: On average the ΔV and crash angle are underestimated in NASS-CDS when analyzing the data based on collision partner/object and structural engagement. The largest difference is found in small overlap crashes and the least difference in collision scenarios similar to barrier tests. Supplemental materials are available for this article. Go to the publisher's online edition of Traffic Injury Prevention to view the supplemental file.