abstract
- Each power electronic converter contains several semiconductor switches which are subject to faults. In this paper, an online model-based diagnosis algorithm is applied to a dc-ac half-bridge (HB) inverter to isolate switch faults. The circuit is modelled with hybrid bond graph (HBG) and the residuals are generated. The fault signature matrix (FSM) is used to determine the minimum number of detectors for isolable faults. A mode detection algorithm is proposed to estimate switch states. Short circuit switch faults are detected and isolated with the proposed mode detection algorithm. Least square method is adopted to identify the component faults. Incipient and abrupt faults in the power electronic inverter in switches, detectors, or other components are isolable using this model based approach.