Optimum Design of Cold Extrusion Combined Die Based on NN and GA

摘要

In the design of structural size for cold extrusion combined die, the diameter of each layer and shrink range between fitting surfaces must be optimized under the condition of working in order to make full use of the potential of die material and ensure the best performance of die. Finite-element method, neural network and genetic algorithm are combined together to optimize the size of combined die. By taking three-layer combined die as an example, a parameterized model for FEM analysis is built. Some groups of structural sizes are obtained by orthogonal experiment, and simulations under each group are performed to get corresponding equivalent stress distribution in the die. The results are used to train BP neural network, so that the nonlinear mapping relation between structural sizes of combined die and equivalent stress values is obtained. According to uniform strength design philosophy, genetic algorithm is applied to optimize combined die sizes. Comparing the optimal result with the theoretical one, a good agreement is found. It shows that the intelligent design method is feasible, which provides a foundation for optimum design of complicated non-linear problems.