Abstract: In view of that traditional methods can not efficiently achieve multiobjective optimization, finite element method is combined with surrogate model technology, with vehicle mass, the peak acceleration of B pillar and the intrusion of firewall as the objectives of optimization and their surrogate models are constructed to study the influence of sample size on model accuracy, and an optimization on the thicknesses of panels is conducted by using NSGAII algorithm. The results show that increasing sample size can not necessarily enhance the accuracy of surrogate models effectively. The accuracy of test point evaluation method to a large extent depends on the number and location of test points, so it can not accurately evaluate the accuracy of surrogate models, while the adoption of cross verification method get a better results. After optimization based on surrogate models, significant improvements are achieved: vehicle mass decreases by 41kg, the peak acceleration of B pillar lowers by 844% and the intrusion of firewall reduces by 603%.

Key words: frontal crash, surrogate model, multiobjective optimization