Proceedings of the
9th International Conference of Asian Society for Precision Engineering and Nanotechnology (ASPEN2022)
15 – 18 November 2022, Singapore
doi:10.3850/978-981-18-6021-8_OR-08-0317

A Deep Learning Based Quality Monitoring System for Injection Moulding Process

Doan Ngoc Chi Nam1,a, Chan Wing Fook1 and Michelle Nguyen Thanh Truc1

1Manufacturing Execution and Control Group, Singapore Institute of Manufacturing Technology, 2 Fusionopolis Way #08-04, Innovis 138634, Singapore

ABSTRACT

Online quality monitoring and defect detection is critical for realising high quality manufacturing. This paper presents a quality monitoring system via deep learning based for early detection of defects at an injection moulding process. Main contribution of this paper is a light CNN (Convolutional Neural Network) -LSTM (Long Short-Term Memory) model implemented on an edge-device to detect possible defects in real-time. The proposed model was designed to use features from both manufacturing parameters and products images; and was trained with a real manufacturing dataset. Its performance was also benchmarked with other candidates to evaluate the effectiveness in prediction accuracy and computational costs. Thanks to its advanced performance, the proposed model has been implemented at Model Factory @ SIMTech for real-time quality monitoring of an injection moulding machine.

Keywords: Injection moulding process, quality monitoring, defect detection, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM)



PDF Download