Abstract: Aiming at the problem of profile machining accuracy out of tolerance (error is up to ±0.343 mm) in the semi finish milling process of aero engine blade, the numerical model of blade milling is established based on Deform and ABAQUS, the contact point of deformed cutter in thermal and mechanical field of blade is analyzed, the data of elastic deformation of blade is collected, and the orthogonal experiment is designed. BP neural network is used to train and predict the data set, and genetic algorithm (GA) is used to solve the optimal value of the BP neural network and verify the feasibility. The results show that the optimal solution of blade milling elastic deformation obtained by BP-GA algorithm meets the profile tolerance of blade profile ±0.05 mm.