PW-FBPNN: A Hybrid Fault Diagnosis Method for Power Circuit Systems Combining Principal Component Analysis, Wavelet Packet Transform, and Fuzzy Neural Networks

Abstract

Due to the complexity of fault states and the non-linear relationship between input and output responses, fault diagnosis in complex power circuit systems faces significant challenges. This study proposes a novel hybrid method, PW-FBPNN, which integrates principal component analysis (PCA), wavelet packet transform (WPT), and fuzzy back propagation neural network (FBPNN) to enhance fault diagnosis. The effectiveness of this method was demonstrated through experiments on the voltage divider basic operational amplifier and the second-order filter circuit of the four operational amplifiers. PW-FBPNN achieved 100% accuracy in diagnosing most types of faults, with a minimum accuracy of 91.67% for challenging faults. This method was significantly superior to existing methods such as FCM-HMM-SVM and KICA-DNN in terms of accuracy and computational efficiency and could complete the diagnosis in just 0.01 seconds. These results indicate that PW-FBPNN has the potential to improve fault diagnosis in power circuit systems, providing a promising solution for enhancing system reliability and maintenance efficiency.