Sensitivity Analysis of Various Vehicle Dynamic Simulation Software Packages Using Design of Experiments (DOE) 2020-01-0639

A previous paper on this topic presented the use of design of experiments (DOE) to evaluate the sensitivity of vehicle dynamics simulation of the postimpact motion of a vehicle that included high initial rotational rates. That investigation involved only one software package and thus was confined to one simulation model for the purposes of developing and refining the analysis method rather than including a variety of simulation models for broader application. This paper expands the application of the method to investigate the comparative behavior and sensitivity of several other vehicle dynamic simulation models commonly used in the field of crash reconstruction. The software packages included in the studies presented in this paper are HVE (SIMON and EDSMAC4), PC-Crash and VCRware. This paper will present the results of the study, conducted using DOE, involving these models. The eleven factors selected for the study presented here were chosen based largely on the results of the prior study. The experimental design was expanded from 16 trials to 32 trials to provide additional insight into the interactions between the factors. Three response variables were used: the x-coordinate, y-coordinate and total rotational displacement of the vehicle at rest.

The results of the analysis show the sensitivity of all four programs to the eleven factors as well as the interactions between the factors. The results show that, consistent with prior results, and the expectation of the authors, the factor that all four programs showed as the most sensitive is the tire-roadway drag. Additionally, consistent with a reconstructionist’s intuition, is that the programs are sensitive to the initial (postimpact) velocities. However, the level of the sensitivities of the four programs to a given change in the same factor are different in many cases.

As compelling as the results are that show the largest sensitivities, the results show factors that some, or all, of the programs show little sensitivity. For example, for response using the y-coordinate and rotational displacement of the vehicle at rest, all four of the programs were relatively insensitive to changes in the lateral and longitudinal stiffnesses of the tires over the range selected.