Crashworthiness Optimization Using Response Surface Methodology Nielen Stander Livermore Software Technology Corporation Summary
Response Surface Methodology has been applied to develop trade-off curves of crash induced deformation vs. mass. The response calculations involve explicit dynamic analyses of a full vehicle executed using LS-DYNA. A crash simulation model which was constructed using 100,000 elements, is used to demonstrate the methodology. The model represents a fully detailed vehicle and invokes surface contact features necessary for crash simulation. The runs are scheduled in parallel on a parallel computer featuring 48 processors. Eight main member gauges are chosen as design variables.
All the design functions relate to the deformation of the vehicle or its component mass. The four deformation criteria are combined into a multi-objective formulation. In anticipation of a conflict with the component mass, the deformation objective is traded off against the mass.
The construction of the design surfaces requires at least 45 points. Because of the number of processors available (30) it was decided to conduct 60 runs from D-optimal designs. Based on five checkpoints, root-mean-squared prediction errors of between 2.4 and 11.5% were obtained.