Open AccessOpen-Circuit Fault Diagnosis of Power Rectifier Using Deep Convolutional Neural Network
Author(s) -
Ruoyue Wang
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 - 10.1088/1742-6596/1642/1/012009
Subject(s) - fault (geology) , convolutional neural network , computer science , rectifier (neural networks) , artificial intelligence , feature (linguistics) , deep learning , process (computing) , artificial neural network , power (physics) , bridge (graph theory) , stuck at fault , pattern recognition (psychology) , machine learning , fault detection and isolation , recurrent neural network , actuator , medicine , linguistics , philosophy , physics , stochastic neural network , quantum mechanics , seismology , geology , operating system
Aimed to automatically provide accurate fault diagnosis of data from failed power electronics, various studies have been researched based on different approaches. Recently, data driven methods based on deep learning have required increasing attention because of their automatic feature learning abilities. Nonetheless, one of the challenges using these methods in practice is how to obtain the most representative fault features and ensure the better predication performances at the same time. This paper is in respect of the open-circuit fault diagnosis of the phase-controlled three-phase full-bridge power rectifier using deep convolutional neural network (DCNN) for extracting and further classifying fault features. The process mainly includes four steps. Firstly, a presupposed approach of DCNN which is applied to automatically capture the paramount fault features from the raw data is briefly introduced. Then a structure of DCNN is designed to extract the features from output data. Furthermore, the model and framework of the fault diagnosis system are developed to diagnose the open-circuit fault of the power rectifier. Finally, the effectiveness of the proposed method is validated using simulation results. Experiments illustrate that the DCNN model can achieve high accuracy in different fault cases and present great capability of diagnosing fault types in open-circuit fault of power rectifiers.