Data driven injection molding process monitoring using sparse auto encoder technique

Injection molding process monitoring is quite essential for the stabilization of product quality. One of the most important things is to identify the character of injection batch process. In this study, sparse auto encoder technique is applied to extract features from the raw trajectories of system pressure and screw position. Subsequently, the process condition is identified by performing a classification on the features, in comparison with the raw trajectories data, and the principal components. The mean reconstruction error and the classification accuracy are selected to evaluate the representation capability of the extracted features. The experimental results show that the sparse auto encoder is an effective method of extracting features from the injection processing batch data, indicating that it is useful in injection molding process monitoring.