Development of e-Dart-based artificial neural network for multiple quality characteristic online defect detection system in injection molding processes

Most of the thermoplastic parts with complex geometry and narrow acceptance tolerances are produced by injection molding processes. However, the remains of unavoidable shrinkage and random variations during cooling and processing can result in the production of defective parts. Therefore, to achieve better production reliability and capacity, an online defect detection system is needed, particularly for high-precision and fast-paced injection molding. The conventional way of monitoring and modeling deals with only a single quality aspect at a time, but many products nowadays require more than one quality characteristic inspection before delivery to customers. To fill the gap between single and multiple quality aspect monitoring, this paper presents the development of an integrated e-Dart-based artificial neural network (e-ANN MQOD) system for sensing multiple quality characteristics with real-time process information extracted from in-mold sensors. The ANN prediction model of weight shows its defect detectability by sorting out 99.95% of the defects with its mean squared error (MSE) of \(5.33\times {10}^{-5}\) $5.33\times {10}^{-5}$ , while the ANN prediction model of dimension shows its precision of 99.97% with its MSE of \(3.00 \times {10}^{-5}\) $3.00\times {10}^{-5}$ , based on the real-time information. This research successfully develops a multiple-objective online defect detection system that reveals the potential feasibility of integrating different predictive models regarding the priority of quality characteristics, contributing to the development of intelligent injection molding processes in the era of Industry 4.0.