Parametric Fault Diagnosis Based on Fuzzy Cerebellar Model Neural Networks

This paper aims to propose a more efficient classifier and applies it to parametric fault diagnosis for converter circuit. A novel method for fault diagnosis based on a fuzzy cerebellar model neural network (FCMNN) in a dual-buck bidirectional dc-ac converter circuit is proposed. The proposed method uses fast Fourier transform to analyze the fault signal and effectively extract fault characteristics. The parameter adaptation laws of the classifier are derived to achieve fast and effective training and testing efficiency. After training, the neural network can identify the working state of capacitor and inductor, and achieve parametric fault diagnosis. The training samples and test samples are collected through a dual-buck bidirectional dc-ac converter circuit simulation system and a practical experimental platform. A back-propagation neural network, support vector machine, and the proposed FCMNN are performed for circuit fault diagnosis, and the results of these three methods are compared. Both the simulation results and experimental results show that the proposed FCMNN can effectively diagnose parametric faults, such as the component degradation of the capacitor and inductor, with fast learning and diagnosis speed, higher diagnostic accuracy, and good antinoise ability.