Aiming at the problems that traditional BP neural network surrogate model had deficiency of fitting accuracy and computational efficiency, the Mind Evolutionary Algorithm was used to optimize BP neural network and an improved BP neural network surrogate model reliability calculation method was proposed. Firstly, the Mind Evolutionary Algorithm was used to optimize the weights and thresholds of BP neural network to obtain the optimal initial value. Secondly, the Bayesian Regularization algorithm was used to train the optimized neural network to establish MEA-BR-BP neural network surrogate model and verify the effectiveness of the improved surrogate model used test function. Finally, the reliability calculation results were calculated with the Monte Carlo method. The results show that the proposed method improves the fitting accuracy and gives consideration to the calculation efficiency, which verifies the superiority and feasibility of the proposed method.