Innovations in control of injection molding processes

Abstract

Control of injection molding is currently mostly done by operator intuition. The operator controls the set points of the machine based upon his understanding of the effects of each of the controls on the quality of the parts. This situation leads to significant difficulties and variation in the quality of the parts and reliability of the process. An improvement in the intuition-driven process is automated data-driven control strategies among which are artificial neural networks (ANN) and regression analysis. Both of these methods have been demonstrated to give real-time feedback on part quality. Furthermore, our studies have shown that operator-chosen machine control settings are not as effective in predicting part quality as are sensed parameters measured directly from the process. Perhaps equally important is that these techniques focus on quality of the part directly. On the other hand, SPC, which has been assumed to be the improvement over operator intuition, focuses on machine parameters which are, at best, secondary to part quality. These new techniques have been demonstrated in variety of injection molding situations. We have also used the ANN system to recommend the machine control settings that should be used for a new part that has never been made before.