Injection molding, an important polymer processing technique, is a complex cyclic process during which material properties, machine variables and process variables interact with each other in determining the final product quality. This project developed an overall control system for the process with a three-layer structure: a process parameter control layer, a process variable setting layer and a quality control layer....[ Read more ]

Injection molding, an important polymer processing technique, is a complex cyclic process during which material properties, machine variables and process variables interact with each other in determining the final product quality. This project developed an overall control system for the process with a three-layer structure: a process parameter control layer, a process variable setting layer and a quality control layer.

Injection velocity, a key variable during injection, has been selected to demonstrate the process parameter control level design. A generalized predictive controller has been implemented to the velocity control. The controller has inherently fast set-point tracking and superior robustness against the model mismatch than the previously used pole-placement design. Based on the GPC design, a fuzzy multi-model has been proposed for the molding process variables, to provide accurate prediction over a wide range of operating conditions. The experimental tests have shown the performance and robustness of the controllers.

An optimal iterative learning controller (ILC) is adopted to exploit the cycle-to-cycle nature of the molding process, and to deal with the slow response of the actuators. Through robustness and convergence analysis, it is found that a proper setting of the weighting matrices is critical to the success of the control. A scheme has been proposed to reduce the weighting of feedforward action from cycle to cycle resulted in a successful implementation of the optimal learning control with good experimental results.

The injection velocity profiling is proposed to achieve a constant melt-front-velocity to minimize the non-uniformity within the molded product. The profiling problem is transformed into a cascade control problem that is solved through a novel iterative learning approach. The automatic profiling system has been experimentally tested with different mold geometry with good results.

In the quality level design, the current work mainly focuses on the prediction and control of the product weight. An on-line weight prediction model has been developed with the process variable trajectories as the inputs, using principal component regression (PCR). A novel nonlinear enhancement has been made to improve the prediction accuracy of the PCR weight model. Based on the inferential on-line prediction, a closed-loop weight control system has been developed and tested experimentally. The good experimental results have illustrated the effectiveness of the developed system.