Parameters sensitivity analysis of silicon carbide buck converters to extract features for condition monitoring. (November 2020)

• Record Type: Journal Article
• Authors: Loghmani Moghaddam Toussi, A.; Bahman, A.S.; Iannuzzo, F.; Blaabjerg, F.
• Is Part Of: Microelectronics and reliability. Volume 114(2020)
• Publication Date: 2020-11
• Subjects: Electronic apparatus and appliances -- Reliability -- Periodicals; Miniature electronic equipment -- Periodicals; Appareils électroniques -- Fiabilité -- Périodiques; Équipement électronique miniaturisé -- Périodiques; Electronic apparatus and appliances -- Reliability; Miniature electronic equipment; Periodicals; 621.3815
• DOI: 10.1016/j.microrel.2020.113910
• Languages: English
• ISSNs: 0026-2714

Abstract

Abstract: This paper proposes a study on the possibility of using measurable electrical quantities in a DC/DC converter to infer the state of health of active and passive components. We worked out the dependence of several features of the output voltage waveform on the parametric drift of the main switch, the diode, the tank inductor, and the output capacitor. The goal is to use these findings for the implementation of machine-learning algorithms for indirect condition monitoring, i.e., not relying on the direct measurement of the critical parameters. The case study is a buck converter based on silicon-carbide MOSFETs. Simulation results show the sensitivity of various output voltage signal features to these parameters and also their correlations, and as a result, the most appropriate features for the condition monitoring purpose. The same approach can be implemented for other converters. Highlights: We proposed a simulation-based feature engineering of terminal signals for the condition monitoring of SiC converters. We identified and selected time-domain feature sets for machine learning approaches to detect or predict faults and failures. The time-domain features of the terminal values are investigated, including both the transient and steady-state ones. We analyzed the features using correlation and sensitivity, and also introduced a modified sensitivity criterion. We chose feature sets to monitor parametric drifts of components and checked their compliance with theories and experiments.