Prediction of CADI Chemical Composition and Heat Treatment Parameters using a BPNN Optimized with the Genetic Algorithm

Abstract

Due to the increasing application of the Carbidic Austempered Ductile Iron (CADI) with carbides, it is of great significance to predict the CADI chemical composition and heat treatment parameters to meet the requirements of process prediction in the complete design process of CADI parts. This study combines a backpropagation neural network (BPNN) and the genetic algorithm (GA). Based on the domestic production data, six key influencing parameters are selected to establish the BPNN prediction model. The prediction results of the non-optimized BPNN and the BPNN optimized using the genetic algorithm (GA-BP) are compared with the real industrial data. The results show that the optimized prediction model can meet the design requirements for the accuracy and stability.