Cross-correlation based fault electromagnetic signature extraction for open-circuit fault diagnosis in NPC inverters

Power MOSFETs are ubiquitously present in the structure of static power converters. The high switching frequency, power MOSFETs are subject to, along with the different inductive and capacitive parasitic elements unavoidably present in the structure of the electric circuits leads to the emission of undesirable Electromagnetic Interferences (EMI). Despite being harmful, the emitted EMIs are rich in information regarding the state of the emitting agent. Recent works have relied on the electromagnetic signature for open-circuit fault detection. However efficient, these methods usually relied on computationally heavy signal processing tools for fault signature extraction. In this work, however, we propose a cross-correlation-based EMI analysis to extract open-circuit switch faults signatures in a three-level NPC inverter. This method is more efficient and it results in a faster diagnosis time. The cross-correlations between the different intervals of the conducted EMI where each MOSFET pair is solicited are modeled in the healthy case. A change in the similarities between intervals, measured by the cross-correlations, is indicative of an unusual situation. The efficiency of the proposed fault diagnosis strategy is validated through various simulation and experimental results.