Finite element (FE) models are commonly used for automotive crashworthiness design. However, even with increasing speed of computers, the FE-based simulation are still too time-consuming when simulate the complex dynamic process such as vehicle crashworthiness. To improve the computational efficiency, the response surface model has been widely used as the surrogate of FE model for crashworthiness optimization design. Before introducing the response surface model in the design optimization, the RSM should satisfy the requirement of the accuracy. However, it’s very difficult to make an once end for all decision on how many samples are needed when building the surrogate model for the output variable with strong nonlinearity. In order to solve the aforementioned problems, a surrogate optimization method called Efficient Global Optimization (EGO) is proposed to conduct the crashworthiness design optimization. An initial DOE samples are generated to build the Kriging models for the responses. The new sample is found by maximizing the expect improvement criterion and then utilized to the update the Kriging models continuously. Since the expect improvement criterion can balance the global search and local search, a relative global optimal design is found after several iterations. Thus, the proposed method will reduce the number of high cost of evaluations while achieve the desired optimal design. The proposed method is demonstrated through a vehicle weight reduction problem while satisfying the safety performance requirements. A comparison study between the proposed method and traditional surrogate model based crashworthiness design optimization method indicates the EGO method is more effective in crashworthiness design.