Sensor/Model Fusion for Improved Process Understanding and Control in Injection Molding

Abstract

Many types of sensors have been investigated to monitor the process conditions in an injection mold during the molding process. Sensors such as thermocouples, pressure sensors, optical sensors, and ultrasonic sensors have been used to monitor the material, mold, and machine status during molding. Users have always found disadvantages or constrains in application for each type of sensor. Certain sensors can only be applied below a certain temperature. They may be hard to install at a critical location, or have difficulty in making an on-line measurement. A model of the process can predict molding conditions and polymer behavior at any location in the process, but the result is not on-line and the accuracy may be unacceptable. In this work, the signals from a cavity pressure sensor and an ultrasonic sensor are used in conjunction with a finite difference model to predict conditions in an injection mold during molding. The combination improves the model predictions and allows monitoring of variables that are not easily measured. Using this system one sensor is used to provide feed back to improve the model accuracy, while the model acts as a “virtual sensor” predicting the output of a variable that is not as easily measured.