Modern injection molding machines are much more sophisticated than previous designs and include microprocessors for control of key parameters in order to enhance throughput and quality of molded parts. However, they are susceptible to changes in operating conditions, which can reduce performance, and which are not adequately controlled by current single-loop PID controllers. This research is directed toward development of more advanced multi- variate control methods which can adapt to changes in the machine characteristics and provide more effective control of key parameters, such as melt temperature, than current designs. The Phase I research has demonstrated feasibility by using a combined system identification and model predictive control methodology, using data taken from an operating machine at Cincinnati Milacron, Inc. Phase II will continue development of the identification and control algorithms. Particular emphasis will be placed on adequacy of the identification algorithm for model order selection and handling of dead time in the machine impulse response functions. The control algorithm will be designed for maximum robustness in the face of variations in operating conditions. Algorithms for on-line estimation of unmeasured machine parameters, such as melt temperature, will be developed. A performance monitoring algorithm will be employed to detect the need for reidentification of the system, utilizing statistical tests on the measured machine parameters.
