Multivariate process analysis for the prediction of injection molded part quality

The injection molding process is a highly complex system in which multiple variables influence the final molded part quality. The influence and interactions of all these variables requires the use of a system that analyzes the process in a multivariate manner. This thesis investigates a multivariate approach to on-line quality monitoring and quality predictions that can be fully implemented in a fairly short time period. This multivariate approach consists of determining and monitoring the most important process data features that are generally available and correlating them to critical part dimensions. A multivariate model is then created from these correlations; parts are then compared to this model and accepted or rejected based on multivariate statistics.; The implemented research investigates the use of different process sensors and different signal data features with different molds and materials. The results determined that this approach can be implemented and used successfully. Not only can it be used successfully, but it can be utilized on a molding process while monitoring only the sensors that are typically included on most machines. Multivariate models were created based on acceptable molded product. Validation trials were then run in which unacceptable parts were intentionally created. The unacceptable parts were rejected while acceptable parts were accepted.