Optimization of Variable Blank Holder Force Based on a Sharing Niching RBF Neural Network and an Improved NSGA-II Algorithm

摘要

Variable blank holder force (VBHF) is an important process parameter in sheet metal forming. The purpose of this study is to propose a sharing niching radial basis function (SNRBF) neural network for VBHF optimization. Two methods are put forward to improve the prediction accuracy of a RBF neural network. (1) A RBF neural network is trained by a sharing niching technique to achieve global optimal nodes. (2) In Latin hypercube design, Spearman correlation analysis is employed to decrease sample correlation. In addition, in order to improve the performance of the non-dominated sorting genetic algorithm (NSGA-II), excellent individuals in each class of non-dominated individuals are selected by employing immune operators. Based on the Spearman correlation analysis and the Latin hypercube method, training samples are generated and numerical simulations are performed for a double C part. The surrogate models between VBHF and forming quality are constructed by a SNRBF neural network. The Pareto frontier solutions are achieved by employing the improved NSGA-II algorithm. Grey relational analysis is applied to determine the optimal VBHF loading trajectory. The results show decreased wrinkles in the optimized forming part and greater uniformity in the plastic deformation by employing the optimized VBHF, thereby leading to improvement in the forming quality of the double C.