A Numerical and Experimental Investigation of Neural Network-Based Intelligent Control of Molding Processes

Abstract

The current investigation focused on the development of intelligent injection molding processes by utilizing a neural network based control unit. In this study, the emphasis was on the control of flow front progression during injection molding processes. The progression of a flow front into a mold, cavity is crucial since it dictates the locations of possible air voids and weld lines. It is desired that the flow front progresses towards the vent locations and that weld lines coincide with locations where their quality decreasing influence has a minimum impact on the overall part performance. The intelligent control scheme developed is based on a neural network that was trained with data obtained from a first-principles based process model rather than actual molding experimentation. The control strategy was developed such that one can specify a desired flow progression scheme and the controller will take corrective actions during the molding process to realize this scheme. This is done by controlling the inlet flow rate at various inlet gate locations. Experiments were conducted with a 2-D, complex shaped, mold cavity to test the performance of the control unit during actual injection molding processes. The mold had two inlet gates and three different desired flow progression schemes were considered. In all cases, the first principles model/neural network based control unit was able to steer the flow front along the corresponding desired flow progression path.