Abstract: The energy absorbing box needs multi-pass stretch forming due to its large ratio of height to diameter. The stretching coefficient has great influence on the forming quality of the energy absorbing box. In order to determine the stretching coefficient of each pass objectively and reasonably, a optimization method of multi-pass stretch coefficient of the energy absorbing box based on BSO-BP neural network was proposed. Taking a certain type of car energy absorbing box as the research object, the maximum thinning rate and FLD forming field safe ratio of the box were used as the forming quality evaluation criteria. Firstly, with the tensile coefficient of each pass as the experimental factor, the Latin hypercube experimental design method combined with the finite element analysis technique was used to establish experimental samples. Secondly, the weight and threshold of BP neural network were optimized based on the brain-storming algorithm (BSO), and the BSO-BP neural network model of multi-pass stretch coefficient was established. Thirdly, multi-objective particle swarm optimization algorithm (MOPSO) was used in the established BSO-BP neural network model to obtain the optimal multi-pass tensile coefficient that met the forming quality evaluation criteria. Finally, the effectiveness of the proposed method is verified by actual production, which provides a new method for objective and reasonable determination of tensile coefficient of deep drawing products in engineering.