Open-Circuit Fault Diagnosis of an Inverter Based on Bayesian Network

摘要

The paper establishes an open-circuit fault diagnosis Bayesian network (OFDBN) to diagnose the faults of inverter's IGBTs. The topological structure of the OFDBN includes two layers, an inverter faults layer and a fault symptoms layer. The faults layer includes various IGBT open-circuit faults, and the symptoms layer includes various fault features extracted from the output line voltage of the inverter. A number of fault training data and test data are obtained by the method of changing carrier frequency and modulation ratio. The study uses the maximum likelihood algorithm to train the OFDBN parameters by the training data simulated in Matlab. The trained OFDBN is used to diagnose IGBT open-circuit faults of inverters. The results show that the proposed OFDBN has a very high accuracy for both single IGBT fault and double IGBT fault, and it can accurately locate the open-circuit fault location.