Dual genetic neural network method for robust optimization of square box forming

To solve the problems of high computational cost and difficulty in obtaining satisfactory coordinated solution in sheet metal deep drawing robust design, a dual genetic algorithm back propagation (Dual-GABP) neural network for multi-objective robust design is proposed by combining Kreisselmeier-Steinhauser (KS) function and GABP neural network. Firstly, the first GABP response surface model of sheet metal deep drawing is established by using experimental design samples and finite element simulation. Secondly, the second GABP model of the mean and standard deviation of sheet metal forming quality is constructed using the first GABP model and Monte Carlo simulation, and the computational cost can be reduced by replacing the finite element model with the first GABP model. Finally, the multi-objective optimization problem is transformed into a single-objective problem using the KS function, and the satisfactory coordinated solution is obtained by optimization. The multi-objective robust optimization of NUMISHEET’93 square box drawing is carried out by using the proposed process. Compared with the deterministic optimization results, the standard deviation of the thinning and the wrinkling index of the robust design are reduced by 41.5% and 18.1% respectively, and the robustness of forming quality is enhanced significantly.