Intelligent Modeling of Injection Molding Process Based on PLS-NN

Abstract

Considering the problems of small sample, non-linear, multi-variable coupling and dynamic of plastic injection molding process, an intelligent modeling methodology based on partial least-squares (PLS) regression and neural networks (NN) was presented to predict the quality index of injection-molded parts in this study. This modeling approach extracted relatively few latent variables as inputs of intelligent model by using correlation analysis and principle components extraction. Then a constructive approach and program was used to optimize training process and improve the precision of modeling. Furthermore, the performance of quality prediction model was evaluated and tested by its application to verification tests. Compared with normal neural networks, PLS-NN model has better prediction performance and generalization ability, the simulation result proved that this approach was effective.