부산대학교 도서관

Fault diagnosis is becoming increasingly important in improving the reliability of power electronic devices. The research in this paper focuses on the issue of the faulty operation that can occur after partial IGBT open-circuit faults in three-phase PWM rectifier circuits. To promptly and effectively diagnose faults and to determine their locations, a fault diagnosis method based on transient characteristics and random forest classification is proposed. First, the characteristics of single and double IGBT open-circuit faults in three-phase PWM rectifier circuits are analyzed. It is discovered that these faults do not immediately manifest. Instead, they exhibit fault characteristics in the corresponding time sequence. Then the random forest classifier is trained using transient fault samples from the three-phase PWM rectifier circuit. Finally, generalized testing is performed on data that was not involved in the training process, with an accuracy rate of over 98%. The use of frequency distribution graphs for visual analysis of the diagnostic results solves the problem of diagnosing multiple IGBT open-circuit faults.