A support vector machine framework for fault detection in molecular pump. Issue 1 (2nd January 2023)

• Record Type: Journal Article
• Authors: Yuan, X.L.; Kai, J.; Chen, Y.; Zuo, G.Z.; Zhuang, H.D.; Li, J.H.; Hu, J.S.
• Is Part Of: Journal of nuclear science and technology. Volume 60:Issue 1(2023)
• Page Start: 72
• Page End: 82
• Publication Date: 2023-01-02
• Subjects: Fusion device -- fault detection -- molecular pump -- support vector machine; Nuclear engineering -- Periodicals; Nuclear physics -- Periodicals; Nuclear energy -- Periodicals; 621.4805
• DOI: 10.1080/00223131.2022.2080124
• Languages: English
• ISSNs: 0022-3131

Abstract

ABSTRACT: The molecular pump fault results in a vacuum leak that can lead to accidents in a fusion device. The fault detection for the molecular pump can provide an early warning to avoid a greater loss in a fusion device. This paper proposes a Support Vector Machine (SVM) framework for fault detection in a molecular pump. This study investigates the key technologies, framework, and system development methods of the data-driven fault detection system for a molecular pump. This study proposes a molecular pump test system to simulate the fault of a molecular pump to solve the highly unbalanced data problem and verify the framework availability. The packaged development framework, implemented in the test system, realizes the fault detection function. The developed system offers better practicability and can effectively detect perform faults in a molecular pump. The results obtained would support new ideas and methods for early warning in molecular pumps and pave the way for stable and reliable vacuum pump systems in a fusion device. Graphical Abstract: uf0001