Abstract: Based on the surrogate model technology, this paper proposes an optimization design method on the complex structure, and the method is applied to the optimization design on the ventilated disc brake. The proposed method integrates the CAE analysis, experimental design, construction of the surrogate model and the nonlinear optimization. Expermental design adopts the Latin hypercube sampling stragety, the surrogate model adopts the improved response surface methodology and the optimization algorithm applies the sequential quadratic programming algorithm. However, there is a problem that the differences between the predicted values and experimental values for some sampled points are relative large in the traditional response surface methodology; in the improved method, it is assumed that the model is feasible only if the relative error between the predicted value and the experimental value is within a margin (?) at each sampled point. The optimal design of the ventilated disc brake is obtained, regarding the maximization of the fatigue life as the objective. The fatigue life is predicted by Coffin-Manson equation, and the maximum stress and temperature on the disc are obtained through the thermal-mechanical coupled finite element analysis on the emergent brake. The result shows that the proposed method is an effective optimization design method on the complex structure.