Abstract: The prediction accuracy and applicability of the radial basis function (RBF) and the response surface method (RSM) are studied, and an optimization idea is put forward to solve the problem of selecting an applicable shaping parameter of the RBF model when the surrogate model technology is employed in structure crashworthiness optimization. The genetic algorithm (GA) is used to optimize the crashing energy absorption of a thin-walled structure and find the global optimum which can be viewed as the target of the surrogate model. The necessity of testing surrogate model’s prediction accuracy via numerical examples is demonstrated and the applicability of several methods of design of experiments is tested. Thaping parameter of RBF model is treated as a variable, and a few random sample points are used to update radial basis function model suitable. The results show that the updated radial basis function model will be more suitable for structure crashworthiness problem.

Key words: crashworthiness;surrogate model;radial basis function (RBF);shaping parameter