Artificial neural network-based online defect detection system with in-mold temperature and pressure sensors for high precision injection molding

Abstract

Injection molding is widely used for mass production of thermoplastic parts with complex geometry and tight dimensional tolerance. However, due to the unavoidable shrinkage and uncontrollable process condition variations, defective parts may occur. Thus, dimensional control and online defect detection are extremely important for quality control, particularly for high precision injection molding. The conventional monitoring and control are based on machine setting parameters, but it may not capture the molding condition variations under the unchanged machine settings. This paper develops an artificial neural network (ANN)-based online defect detection system with the real-time data extracted from in-mold temperature and pressure sensors. Both multilinear linear regression (MLR) and ANN models were developed based on the real-time data, but the ANN model is much better than the MLR model. The ANN model has a high prediction accuracy of 98.34% with the coefficient of determination R2 of 91.37%. When applied to defect detection, the ANN model has a defect detection accuracy of 94.4% in consideration of type I and type II errors. This research demonstrates the feasibility of integrating such an ANN-based expert system to injection molding process, to improve online dimensional monitoring. The ANN model also can be easily adapted for detecting other quality characteristics of injection moldings, which would be helpful for the advances in intelligent injection molding.