Open AccessOpen-circuit fault diagnosis of traction inverter based on improved convolutional neural network
Author(s) -
Jun Liu,
Junnian Wang,
Wenxin Yu,
Zhenheng Wang,
Guang’an Zhong
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1633/1/012099
Subject(s) - insulated gate bipolar transistor , convolutional neural network , fault (geology) , inverter , traction (geology) , computer science , artificial neural network , electronic engineering , engineering , artificial intelligence , electrical engineering , voltage , mechanical engineering , seismology , geology
In the traction system of high-speed EMUs, the inverter’s Insulated Gate Bipolar Transistor (IGBT) often occurs open-circuit (OC) faults. However, traditional fault diagnosis relies mainly on signal processing to extract fault features, which is susceptible to environmental interference, resulting in poor generalization ability of the model. Aiming at this problem, a fault diagnosis method based on an improved convolutional neural network is proposed. Firstly, the three-phase stator current signal is preprocessed by wavelet domain denoising. Secondly, fault features are independently learned through a convolutional network. Finally, a fully-connected layer is used for fault diagnosis. The experimental results show that this method could resolve the OC fault diagnosis problem of the inverter’s IGBT effectively, and also achieve higher accuracy under the interference of noise, and the diagnosis can be made at 0.01s after the fault occurs.