A machine learning approach to quality monitoring of injection molding process using regression models

• Published in: International journal of computer integrated manufacturing Vol. 34; no. 11; pp. 1223 - 1236
• Main Authors: Farahani, Saeed, Xu, Bin, Filipi, Zoran, Pilla, Srikanth
• Format: Journal Article
• Language: English
• Published: Taylor & Francis 02.11.2021
• Subjects: In-mold sensors; Injection molding; Machine data; Machine learning; Quality monitoring; Regression models
• ISSN: 0951-192X; 1362-3052
• DOI: 10.1080/0951192X.2021.1963485

Recent advancements in instrumentation and the emergence of industrial IoT solutions have created new possibilities to collect data in injection molding processes. Although many approaches have been introduced to develop quality monitoring systems based on this wealth of data, most of them require human expertise for data evaluation or feature extraction, thus their performance is highly dependent on the quality of the incorporated features. In this work, a data-driven approach to the development of a quality monitoring system is proposed using a wide range of process data obtained from a variety of in-mold sensors and machine data sources. Initially, ten different machine learning algorithms are explored. Among them, the multiple linear regression models show the best performance due to their low training error. Several training strategies are implemented based on allocating different subsets of the available data sources. A set of experiments perturbated by three common process disturbances is used to validate the accuracy of the trained models in predicting the quality indices. The results demonstrate that the average error in predicting the weight, diameter, and thickness of the injected parts does not exceed 0.5%, 0.1%, and 0.4% respectively, which imply the effectiveness of the proposed data processing method.