Application of Radial Basis Function Networks Combined with Genetic Algorithm in Predicting the Geometric Parameters of Drawbead

Abstract

The drawbead plays a very important role in automobile covering part forming processes. Traditional drawbead design mainly depends on designers’ experience. In order to obtain proper restraining force during die try-out, it is often necessary to adjust the drawbead through a very complicated procedure. It is thus meaningful to study the relationship between the parameters used to reflect metal forming effects and the geometric parameters of drawbead and then create a prediction model for them. This paper employs the radial basis function neural network technology to predict the geometric parameters of drawbead used in forming processes, where the genetic algorithm is used to optimize the neural network structure. Simulation results show that the proposed approach outperforms the curve fitting method.