This study focuses on the utilization of predictive strategies for controlling the parameters of plastic injection molding and the evaluation of the cooling time for cylindrical plastic components. This focus is as a direct need for a control method that can be applied universally to injection molding machines and for accurate prediction of part cooling time in order to reduce operator setup time.

The injection molding parameters controlled were melt temperature for three zones, the plasticating screw speed, nozzle pressure and plastication back pressure. Dynamic responses were obtained for the molding parameters by conducting a series of experiments involving step inputs to the corresponding actuators. The software algorithms for these tests were formulated using National Instruments Corporation Software LabWindows 2.3a with matching hardware components. The process responses were highly non-linear. Discrete process models were formulated from the experimental data.

A predictive control method (Model Predictive Control (MPC)) was applied to plastic injection molding manufacture for the first time. MPC minimizes an objective function consisting of predicted errors and controls signals at every process sampling instant. The controller output and the process predicted profile are then evaluated. MPC was found to provide consistent setpoint trajectories, good transient characteristics and wide operating bands for the injection molding variables when compared to the proportional, integral and derivative (PID) methods currently used. It is noted that the judicious selection of the process setpoints depends entirely on the machine operator and is not a requirement for this study.

This thesis modifies MPC to introduce Self-generating Predictive Control. This modified MPC was applied to the injection molding parameters to provide model formulation, dynamic matrix evaluation, MPC execution, and self-tuning of the control move weights. The modified MPC is designed to be initiated by the operator thereby obtaining process models for different mold geometries utilizing any plastic material.

A multiple input multiple output (MIMO) model was formulated for MPC of the barrel and nozzle plastic temperatures. Previous controller designs neglected the high degree of interaction between them. More efficient temperature controllers having deadbeat characteristics resulted when compared to single input single output (SISO) MPC.

Utilizing a two-dimensional non-linear conservative model of the heat conduction equation, an algorithm was developed for predicting the plastic part temperature distribution and the cooling duration within the mold cavity. The analysis incorporated cylindrical coordinates and variable thermal properties for amorphous plastics only. This simulation resulted in smaller part cooling durations than those evaluated from cooing time equations currently used. The experimental cooling durations for several part thickness were in good agreement with the theoretical predictions.

In summary, using a conservative model for predicting part cooling time in sequence with this self-generating predictive control strategy provides robust control allowing operator interaction to any injection molding machine.