The goal of this research was to find vehicle characteristics which may contribute to steering maneuver induced rollover accidents. This work involved studying vehicle handling dynamics using the Vehicle Dynamics Analysis, Non-Linear (VDANL) computer simulation. The simulation was used to predict vehicle responses while performing 28 different steering induced maneuvers for each of 51 vehicles. Various measures of vehicle response (metrics), such as response times, percent overshoots, etc., were computed for each vehicle from simulation predictions.
These vehicle directional response metrics were analyzed in an attempt to identify vehicle characteristics that might be good predictor/explanatory variables for vehicle rollover propensity. The metrics were correlated, using the Statistical Analysis System (SAS) software and logistic regression, with single vehicle accident data from the state of Michigan for the years 1986 through 1988.
The analysis found that the vehicle directional response metrics were not good predictors of the observed rollovers per single vehicle accident rate for each make/model. Stepwise logistic regression was performed including environmental and other, non-directional response, vehicle metrics in addition to the directional response metrics. The most significant directional response metric found was less significant than two environmental and four other vehicle metrics. This metric explained only a small portion of the observed variability in rollover rates.