Abstract: The application of neural network and genetic algorithm to inverse solution of parameters of drawbead used in drawing procedure of corner backstop of front floor of a truck is discussed. A neural network model, the structure of which is optimized in genetic algorithm, is designated to simulate the nonlinear mapping relation between parameters used to reflect metal forming effect and geometric parameters of semi-circular drawbead. A sample design method with gradual local densifing samples is developed. The method is helpful to shorten design period of neural network and to improve simulating degree of neural network. It is especially suitable to the case of sample datum being gotten by finite analysis automatically.

Key words: Drawbead, Genetic algorithm, Inverse problem of parameters, Neural network, Optimization of neural network structure