Fault Detection and Diagnosis Technique for a SRM Drive Based on a Multilevel Converter Using a Machine Learning Approach

One of today's well-accepted solutions for the SRM drives is based on multilevel converters. In fact, they present interesting features like an extended voltage range and the capability of fault tolerance. The guarantee of fault tolerance is fundamental in the context of preventive maintenance. However, regarding the power electronic converter, this requires a fault detection and diagnosis algorithm for failures in power semiconductors. Thus, this paper proposes a novel detection and diagnostic approach for the failure of those semiconductors. In this case, it will focus on one of the most commonly used topologies, namely the asymmetric neutral point clamped converter. This approach was developed with the purpose to develop specific patterns that are associated with each semiconductor and fault type. In this way, through the image identification of the multilevel converter current patterns, it will be possible to identify a distinct semiconductor and fault type. Several tests obtained from a simulation tool allowed to show the capability of the proposed approach.