Computerized Crash Reconstruction of Real World Crashes Using Optimization Methodology

Vikas Hasija (GESAC, Inc), Erik G. Takhounts, Stephen A. Ridella (NHTSA)

CHALLENGE - Computerized crash reconstruction must deal with many unknown parameters and can therefore not be solved deterministically with the parametric methodology discussed in the researcher’s earlier publication. In this case study, the researchers modify the parametric methodology to include an optimization scheme to derive an optimal solution for the reconstruction problem in a given range of unknown parameters.








SOLUTION - Real world crashes were selected from the CIREN database and were solved using the proposed methodology. modeFRONTIER software was used for optimization and used the identified unknown parameters as design variables. The objective function and the constraints were defined such that they minimize the differences in injuries and the occupant-vehicle contacts between the real world data and the model prediction.

BENEFITS - Results of the simulations showed that the optimal solution correctly predicted both the occupant-vehicle contacts and the injuries sustained during the crash. modeFRONTIER provided correlation coefficients between all the input variables and the output variables. Using the correlation coefficients, the critical parameters were identified for each of the injuries. These design changes could help mitigate all or some of the injuries sustained by the occupant.