Two knowledge-based approaches to fault detection and isolation on a PWM inverter

This paper presents an approach based on knowledge models to detect and isolate faults in a PWM inverter supplying a synchronous machine. These faults don't affect the system protections. A diagnostic system which uses the input variables of the drive, the current measurements, is presented. It is based on the analysis of the current vector trajectory and of the inverter input current in faulty mode. These two methods have been successfully applied to an experimental system.