Evaluation of Optimization Algorithms for Crash and NVH problems

The purpose of the studies presented here was to establish a standard optimization strategy for crash and NVH (i.e. noise, vibration, and harshness) problems. Crash simulation is a CPU consuming task, the optimization for this type of problems requires efficient strategies. The approaches should be rather general, which is inevitable for their integration into the standard design process. Monte-Carlo-search strategies, evolutionary and genetic algorithms, kriging, simulated annealing, and some methods based on regression analysis were tested. Mono- and multi-criteria optimization problems were considered. Finally, a standard strategy for optimizing is proposed and tested on a real MDO problem with five crash load cases, statics and dynamics with a finite element model of about 800,000 elements.     «

The purpose of the studies presented here was to establish a standard optimization strategy for crash and NVH (i.e. noise, vibration, and harshness) problems. Crash simulation is a CPU consuming task, the optimization for this type of problems requires efficient strategies. The approaches should be rather general, which is inevitable for their integration into the standard design process. Monte-Carlo-search strategies, evolutionary and genetic algorithms, kriging, simulated annealing, and some...     »