Process models used to determine part quality in-process could potentially be used to reject defective parts as they are ejected, and also as a basis for more advanced forms of process control. This paper discusses experiments where process models based on Neural Network and Nearest Neighbor approaches were developed from both machine parameters and high speed pressure traces from the machine nozzle. Both methods proved to be very accurate in predicting defective parts produced in separate experiments, while not rejecting good parts produced. Extension of these techniques to multicavity molds required additional study to determine the optimal location for pressure sensors; this was determined to be in the machine nozzle in this study.
展开▼
