Abstract: By comprehensively utilizing the parameter optimization technique for approximate model and robust optimization method, vehicle occupant restraint system is optimized. The parameters having significant effects on weighted injury criterion (WIC) are selected by global sensitivity analysis. The parameters of support vector regression (SVR) model and kernel function are optimized by using particle swarm optimization (PSO) algorithm, and a high accuracy PSO-SVR approximation model is established. On the basis of deterministic optimization, robust optimization based on Monte Carlo sampling is also carried out. The results show that after optimization the performances of occupant restraint system are apparently enhanced with good robustness

Key words: occupant restraint system, sensitivity analysis, PSO, SVR, robust optimization