Maintaining tight control of temperature in injection molding machines is of great importance due to the fact that polymer properties are temperature sensitive. Hence, variation in polymer temperature during processing can cause variations in product properties and, in extreme cases, can cause polymer degradation. This temperature control problem is also of broader interest because of the similarity of the nonlinear behavior noted here to other thermal control problems. The barrel temperature close to the nozzle defines a baseline for the melt temperature, variations of the barrel temperature in the middle section affect the viscosity of the melt temperature which directly influence shear heat generation, and the barrel temperature near the hopper define the start of melting which affect the total melting length. This work focuses on the improvement of barrel temperature control loops.;The amount of time and effort spent on manual controller tuning of PID controllers for barrel temperature control is great, because of the combination of slow response characteristics. Auto-tuning of the PID controllers provides more consistency in controlled system response and a more efficient way to determine controller gains. The relay feedback tuning method is simple to implement and does not require knowledge of system model. When this method is used to tune the controller parameters for the temperature loops in an injection molding machine, the lack of symmetry in the heating and cooling rates causes significant difficulties in interpretation of system behavior and selection of controller parameters. A way to overcome this limitation is proposed here, along with extension of this method to multi-loop auto-tuning.;A more advanced control strategy, with incorporation of other nonlinear features, is needed to overcome the limitation of PID controller. The learning ability and nonlinear structure of neural networks make them suitable for controlling nonlinear systems. In this work, a neural network controller, which does not require online training and which has a PID structure, is proposed. The performance of the controller is evaluated using an experimentally validated, non-linear model of the barrel temperature control loops in an injection molding machine. Experimental implementation of the controller is performed on a thermal system. The use of the proposed controller enables us to obtain good performance in controlling temperature in both systems.
展开▼