Coherence assessment of accident database kinematic data

The analysis and research of accidents aimed at improving the safety of vehicles and infrastructures are typically based on the retrospective investigation of data that are collected in in-depth accident databases. In particular, kinematic data related to accidents (impact velocity, velocity change of the vehicles, etc.) make possible the identification of correlations between impact severity and injury risk (IR), as well as assessing the effectiveness of vehicle protection systems. The necessary condition to conduct suitable and significant analyses is to utilise data which are correct and representative of national statistics, i.e., congruent with physical laws governing the accident phenomena. Whereas representativeness can generally be retrospectively verified, the checks on kinematic data coherence during codification are rarely performed. The present work describes a procedure to verify the internal coherence of kinematic data collected in indepth accident databases. The introduced checks allow the identification of parameters, which are not internally coherent because the accident reconstruction model employed is inappropriate or improperly used. These checks pertain to physical laws on which road accident reconstruction is based, i.e., momentum conservation, compatibility of velocity triangles, and energy conservation. Moreover, they can be modified and expanded to consider other parameters, making the methodology virtually applicable to any database. In the case of vehicle-to-vehicle collisions, the application of the procedure to detect incongruent data inside two real databases demonstrates how their number is often not negligible. Furthermore, consequences can be substantial for both direct and secondary analysis, i.e., determining IR curves (for example, logistic regression on input data) and identifying IR associated to an accident. Accordingly, the application of checks is particularly recommended during both analysis and collection phases to confirm the congruence of collected data; consequently, the quality of investigation is enhanced.