A Multi-Stage Hybrid Open-Circuit Fault Diagnosis Approach for Three-Phase VSI-Fed PMSM Drive Systems

The performance of Hybrid Electric Vehicles (HEVs), especially in series architecture, is highly dependent on the reliability of electric drive-motor systems. Any failure in power semiconductor devices, such as Insulated-Gate Bipolar Transistors (IGBTs), used in three-phase Voltage Source Inverters (VSIs) for Permanent Magnet Synchronous Motor (PMSM) drive systems, causes a reduction in the reliability and leads to unscheduled maintenance of HEVs. This paper aims to present a three-stage combined model-based and data-driven fault diagnosis approach, the so-called hybrid fault diagnosis approach, to detect, locate, and clear open-circuit faults in IGBTs used in VSI-fed PMSM drive systems in HEVs. Field-Oriented Control (FOC), which is a model-based technique, is used to control the electric drive-motor system. The proposed method, which is based on phase voltage analysis, estimates the current in each phase of VSI using the normal operating conditions dataset to detect open-circuit faults in IGBTs. Once a fault is detected, it is located using the faulty conditions dataset and an online data-driven approach, called the Modified Multi-Class Support Vector Machine (MMC-SVM) algorithm. Thereafter, the faulty IGBT is bypassed by closing the corresponding backup switch, ensuring the continuous operation of the electric drive-motor system. The proposed method can accurately and quickly detect, locate, and clear open-circuit faults in IGBTs without the need for additional sensors. Additionally, it demonstrates robustness against back-to-back and simultaneous faults in IGBTs used in VSI-fed PMSM drive systems in HEVs.

Automated summary

This article proposes a hybrid fault diagnosis approach to detect and address open-circuit faults in IGBTs used in electric drive-motor systems of hybrid electric vehicles. The method combines model-based and data-driven techniques, utilizing phase voltage analysis and the modified multi-class support vector machine algorithm for accurate and quick fault diagnosis.