Open-circuit fault diagnosis of NPC inverter IGBT based on independent component analysis and neural network

Abstract

Power switching devices are the core component of inverter, the fault diagnosis of power switching devices has very important significance for the reliability of inverter. The IGBT is usually used as power devices in the neutral-point-clamped (NPC) inverter, and it has 12 IGBTs totally. NPC inverter is the typical application scenario of the IGBT fault diagnosis. The premise of fault diagnosis method based on signal processing is fault feature extraction. A novel fault feature extraction method is proposed in this paper, which is based on the joint approximative diagonalization of eigenmatrix and independent component analysis (JADE–ICA). A neural network (NN) is used as the fault classification method. Through the JADE–ICA algorithm, the source signal and the separated signal can be effectively one-to-one correspondence, and the effects of nonlinearity and time difference can be overcome. The input of NN is reduce through the JADE–ICA algorithm effectively, which can reduce the time necessary to train an NN, and improve the classification accuracy. The proposed method is verified in the simulink simulation environment, and the fault diagnosis is more than 95.1%.

Keywords

• NPC inverter
• Independent component analysis
• Feature extraction
• Fault diagnosis
• Neural network