Survey on Generative and Discriminative Fault Detection Approaches with Focus on SiC Components

Abstract:
As the power electronics applications are developing, the power electronic components and systems are gaining a more critical role, hence the assessment of their condition, and subsequently, the reliability of components becomes more significant. Unarguably the most crucial step in their fault detection would be selecting a proper approach that fits better to the system characteristics. In this study, generative and discriminative approaches are explained for fault detection. Then, statistical estimation methods are classified into these classes. finally, it is discussed how these methods have the edge over each other, specifically considering Silicon Carbide (SiC) devices' characteristics, and this enables researchers to systematically find a few of methods to implement, according to intrinsic characteristics of systems and constraints.