Fault Diagnosis of Recessive Weakness in Superbuck Converter Based on KPCA-IPNN

This paper mainly analyses the recessive weakness in Superbuck Converter and proposes a diagnosis method through kernel principal component analysis (KPCA) and incremental probabilistic neural network (IPNN). The Superbuck Converter presents excellent performance in many applications and is nowadays faced with demands of higher reliability and steadier operation. Firstly, the character of recessive weakness in both time and frequency domain are demonstrated to clearly reveal the inner actual deterioration of the circuit system. Furthermore, wavelet packet decomposition (WPD) and Parseval's theorem are utilized to extract the feature of recessive weakness. KPCA is also used to the dimensionality reduction of feature vectors. IPNN is applied to the automatic classification of the recessive weakness aroused from different components on the basis of the extracted features. Finally, the classification accuracy of the proposed algorithm is also verified through simulation experiments, the results of which present both satisfying classification accuracy and rapid detection speed.