[SCIE] Design and deployment of a Cloud Large Language Model integrated system for detecting defects in plastic injection manufacturing using lightweight edge devices
- 김두환
- 조회수241
- 2024-03-24
-Title: Design and deployment of a Cloud-Large Language Model integrated system for detecting
defects in plastic injection manufacturing using lightweight edge devices
-Journal/Conference: IEEE Access
-Authors: Doohwan Kim, Yohan Han, Jongpil Jeong
-DOI:
-Journal/Conference Link: https://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=6287639
Abstract: This study presents a novel approach to innovate the plastic injection manufacturing process by designing and implementing a defect detection system that seamlessly integrates cloud computing and Large Language Model using edge devices. We propose a real-time monitoring method linked to a cloud-based learning system, aimed at building and maintaining an efficient learning model and inspection equipment. These insights are then transmitted to the web-based monitoring server, facilitating real-time monitoring and management. This integration significantly enhances the operational efficiency of the factory. For small and medium-sized manufacturing plants burdened by the cost implications of implementing smart factory technologies, we propose a cost-effective defect inspection system that leverages light-edge computing devices. Furthermore, we introduce a defect value algorithm designed to effectively identify product defects, leading to high-quality inspection performance even with lightweight edge devices. By collecting defect detection results through a cloud platform for performing real-time analysis and prediction, as well as monitoring and chatbot systems for inspection equipment, it becomes possible to optimize production, enhance product quality, and reduce operational costs. This holistic approach is expected to accelerate the realization of the 4th industrial revolution through the combination of big data and artificial intelligence, and chart an important direction for the future of plastic injection manufacturing.
-Status: Submitted (2024/03/21)