Taking the chassis shell as an example,the product CAE analysis model was built,and the Moldflow software was used to analysis the product defect,and the optimized factors and indicators were selected. The data samples was obtained from using Taguchi test analysis and CAE simulation,through fuzzy comprehensive quality weighted evaluation analysis it would be useful to effectively resolve that the multi-objective problem was transformed into the single objective optimization problem. The BP neural network prediction model was established,mapping the nonlinear relationship of the process parameters and the quality index. Adopted the genetic algorithm for global optimization,the optimal process parameters within the test scope are as follow:mold temperature is 66.3℃,melt temperature is 227℃,filling time is 4.6 s,holding pressure is 109% of the filling pressure,holding time is 10.2 s,cooling time is 22.7 s. The optimization results were verified by CAE analysis,the results show that the prediction results of neural network are similar to the analysis of CAE software Moldflow,and the multi-objective optimization for the products quality indicators are achieved. The optimization design method can improve the quality of products and shorten the production cycle.