Diagnosis and distinguishment of open-switch and current sensor faults in PMSM drives using improved regularized extreme learning machine. (15th May 2022)

• Record Type: Journal Article
• Authors: Xiao, Li; Zhang, Liyi; Yan, Zhi; Li, Yanqin; Su, Xiaoqin; Song, Wenqiang
• Is Part Of: Mechanical systems and signal processing. Volume 171(2022)
• Publication Date: 2022-05-15
• Subjects: Regularized extreme learning machine -- Beetle swarm optimization -- Multiple switch fault -- Current sensor fault -- Three-phase voltage-source inverter; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2022.108866
• Languages: English
• ISSNs: 0888-3270

Abstract

Abstract: Three-phase voltage-source inverter fed permanent magnet synchronous motors drive system is widely applied in high power drive applications. Multiple power switches faults and sensor faults are very common in drive systems. To rapidly and accurately diagnose and distinguish power switches and sensor faults, this article introduces a data-driven diagnosis approach named regularized extreme learning machine (RELM) to distinguish the fault type and identify fault components. In order to shorten the training time of RELM, LU decomposition is used to solve the output weight matrix of the RELM. Meanwhile, the beetle swarm optimization (BSO) algorithm is designed to optimize the weights and thresholds, as well as the Tent mapping reverse learning and Levy flight group learning strategies are also introduced to improve optimization performance. Further, the database required for training is filled with the current fault indicators instead of the universal three-phase currents, which not only good for obtaining a well-learned fault classifier model but also avoid false alarms generated by the inaccurate thresholds in traditional signal-based approach. Experimental tests show that the proposed approach can rapidly and accurately diagnose and distinguish multiple power switches and sensor faults.