BP Neural Network in Deformation Prediction for Cold Extrusion CAPP System

Abstract

Traditional Group Technology (GT) based CAPP system can provide the cold extrusion process parameters and die designs, but can not predict components’ deformation results. This paper proposed a new way to quickly predict the cold extrusion deformation using back propagation (BP) neural network method. The new way is presented by a kind of step shaft components with forward extrusion as example. First, the main factors of forward extrusion deformation characteristics were summed up for the representation. Then, the predicting model was established by the BP neural network method and trained by the designed orthogonal samples. Finally, the model was used to predict the stress and strain at the component’s biggest deformation zone under the given parameters through the trained model. Compared to DEFORM simulation, the BP neural network prediction uses less of the time and obtains the same precision. This method can be applied to other types of cold extrusion components and supply a function of deformation prediction for the cold extrusion CAPP system. It also improves the efficiency of the cold extrusion process design.