1) A large-scale stochastic automotive crash simulation was performed using 128 parallel simulations on a Cray T3E supercomputer. This allowed analysis of the statistical effects of uncertainties in vehicle properties and crash conditions.
2) Results showed the deterministic single-point analyses produced conservative designs and did not capture the most likely responses. Intrusion values from stochastic analysis had higher means and different most probable values than the deterministic analyses.
3) The impact angle had a large influence on responses like intrusion based on scatter plots, showing a chaotic relationship and inability to control intrusions through angle variation. The stochastic analysis provided more insight than deterministic analysis alone.