Search Results

Most collision reconstructions implicitly assume the same tire/road friction coefficient for all vehicles, despite evidence that friction varies between tires, surfaces, and individual trials. Here we assess the errors introduced by an assumption of a single, universal friction coefficient when reconstructing a collision where vehicles actually had different tire frictions. We used Monte Carlo methods to generate 20,000 synthetic two-vehicle impacts and rest positions using different, randomized friction coefficients for each vehicle and randomized impact speeds. These rest positions were then used to reconstruct both vehicles’ impact speeds assuming a single, common friction coefficient. High and low bounds on the impact speeds were reconstructed using high and low bounds on the common friction. We found that more than 97% of the true impact speeds were in the ranges reconstructed using upper and lower friction bounds.

Numerical parameters describing suspension stiffness and damping are required for 3D simulation of vehicle trajectories, but may not be available. This paper outlines a simple, portable method of measuring these properties with a coefficient of variation of 5% on stiffness. 24 of 26 vehicles tested were significantly stiffer in roll than pitch, complicating analyses with models that don't include anti-roll. Suspension parameters did not correlate with static wheel load distribution, and damping coefficient did not correlate with natural frequency. Computer simulations of the speed required to initiate rollover in an S-curve were highly sensitive to the suspension parameters used. When pre-impact tire marks and rollover distance were considered, the simulations became almost insensitive to suspension parameters.

The critical curve speed formula used for estimating vehicle speed from yaw marks depends on the tire-to-road friction and the mark’s radius of curvature. This paper quantifies uncertainty in the radius when it is determined by fitting a circular arc to three or more points. A Monte Carlo analysis was used to generate points on a circular arc given three parameters: number of points n, arc angle θ, and point measurement error σ. For each iteration, circular fits were performed using three techniques. The results show that uncertainty in radius is reduced for increasing arc length, decreasing point measurement error, and increasing number of points used in the curve fit. Radius uncertainty is linear if the ratio of the standard deviation in point measurement error (σ) to the curve’s middle ordinate (m) is less than 0.1. The ratio σ/m should be less than 0.018 for a radius found using a 3-point circular fit to be within 5% of the actual value 95% of the time.

PC-Crash simulations of staged collisions require dozens of parameters describing vehicle and impact parameters. The Collision Optimizer will vary initial speeds and impact parameters to obtain a best fit to a desired end state, but vehicle parameters are left unchanged. The present paper allows these other parameters to vary in thousands of combinations, re-optimizing the solution in each to find the relationships between the previously fixed parameters and the resulting impact speeds. The results show that tire friction and vehicle inertial properties have the most influence on impact speeds. Other parameters have little influence on the results.