[Purposes] This paper aims to explore the safety performance of a vehicle occupant restraint system. [Methods] An optimization design method of the vehicle occupant restraint system based on the surrogate model update management strategy was proposed. A simulation model of the occupant restraint system was established, and the accuracy and reliability of the simulation model were verified. The sensitivity analysis was performed on the occupant injury response, and the model parameters that had a greater impact on the injury response were screened out. The local evaluation index was used to judge whether to resample the local region of the surrogate model, and the resampling interval was determined. The local sample points were obtained by the optimal Latin hypercube design. Inherited Latin hypercube was employed in sampling, so as to obtain global sample points. The local and global sample points were filtered by the weighted Euclidean distance criterion, and the qualified sample points were added to the initial sample space to update the surrogate model. Finally, the potential optimal solution was searched by the intergeneration projection genetic algorithm. [Findings] The method can reduce the injury of the occupant and ensure the safety of the occupant while ensuring the accuracy of the surrogate model. [Conclusions] The proposed method can efficiently solve the optimal solution of the occupant restraint system.