Optimization of car body under constraints of noise, vibration, and harshness (NVH), and crash

Abstract

A car body structure was optimized for minimum weight under the constraints of noise, vibration, and harshness (NVH), and a crash event, using up to 254 concurrently operating processors. The crash analysis alone, if executed on a single processor and repeated the number of times this optimization required, would have taken 257 days of elapsed computing time. Parallel processing has compressed the elapsed time to one day demonstrating how a multiprocessor machine may be useful in solving engineering tasks that heretofore were regarded as intractable. The optimization procedure transformed the structure initially infeasible to one having its weight reduced and all the constraints satisfied. The experience gained in the reported application indicated it is important to tailor the solution method to the characteristics of the multiprocessor computer architecture and to understand the data handling options offered by that architecture. Another conclusion drawn from this case is that the coarse-grained parallelism whereby an existing code is being replicated over an array of processors should be regarded as an effective way of utilization of multiprocessor machines, immediately available in the interim before solutions are redeveloped from ground up specifically for that class of machines.