Intelligent dimensional prediction systems with real-time monitoring sensors for injection molding via statistical regression and artificial neural networks

Dimensional defect is one of critical issues in injection molding. The consistency of processing conditions can affect the dimensional stability of molded parts. In this paper, a systematic approach is proposed to achieve two objectives. The main objective is to predict the diameter of injection molded parts according to real-time processing data. The second objective is to provide on-line insight from process monitoring by utilizing a monitoring system with in-mold sensors. To meet these objectives, artificial neural network (ANN) and multiple linear regression (MLR) were used to build the prediction model that was integrated with the monitoring system. Taguchi experiments were presented in the study for choosing the optimal parameter settings to meet the target diameter of 50 mm. Five controllable parameters in the study include shot size, nozzle temperature, barrel temperature, cooling time, and holding time. With the processing data collected from the sensors embedded in the surface of mold cavity, regression analysis was employed to build and test the relationship between the processing data and the diameter. The real-time variables from the sensor-based monitoring system such as mold temperature and flow rate were selected as the inputs of the predictive model. The feedforward ANN and MLR models were established to predict the outcome based on the data extracted from sensors, with the prediction accuracy of 99.79% and 99.78%, respectively. It would provide a fast measure and good control of dimensional issues in injection molding.