Health State Estimation and Remaining Useful Life Prediction of Power Devices Subject to Noisy and Aperiodic Condition Monitoring

• Published in: IEEE transactions on instrumentation and measurement Vol. 70; pp. 1 - 16
• Main Authors: Zhao, Shuai, Peng, Yingzhou, Yang, Fei, Ugur, Enes, Akin, Bilal, Wang, Huai
• Format: Journal Article
• Language: English
• Published: New York IEEE 2021; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accelerated aging tests; Aging (metallurgy); Algorithms; Condition monitoring; Degradation; Degradation modeling; Diagnostic systems; Electronic devices; Field effect transistors; gamma process; Heterogeneity; Life prediction; Measurement uncertainty; Model accuracy; MOSFETs; Noise levels; Noise measurement; noisy and aperiodic measurements; Numerical analysis; Parameter estimation; particle filter (PF); Pollution measurement; Predictive models; remaining useful life (RUL) prediction; Robustness (mathematics); Semiconductor devices; SiC metal–oxide-semiconductor field-effect transistors (MOSFETs); State estimation; State space models; Uncertainty; Useful life
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2021.3054429

Condition monitoring of power devices is highly critical for safety and mission-critical power electronics systems. Typically, these systems are subjected to noise in harsh operational environment contaminating the degradation measurements. In dynamic applications, the system duty cycle may not be periodic and results in aperiodic degradation measurements. Both these factors negatively affect the health assessment performance. In order to address these challenges, this article proposes a health state estimation and remaining useful life prediction method for power devices in the presence of noisy and aperiodic degradation measurements. For this purpose, three-source uncertainties in the degradation modeling, including the temporal uncertainty, measurement uncertainty, and device-to-device heterogeneity, are formulated in a Gamma state-space model to ensure health assessment accuracy. In order to learn the device degradation behavior, a model parameter estimation method is developed based on a stochastic expectation-maximization algorithm. The accuracy and robustness of the proposed method are verified by numerical analysis under various noise levels. Finally, the findings are justified using SiC metal-oxide-semiconductor field-effect transistors (MOSFETs) accelerated aging test data.