Fault Diagnosis for Closed Loop Nonlinear System Using Generalized Frequency Response Functions and Least Square Support Vector Machine

Abstract

In this paper, a new fault diagnosis method for closed loop nonlinear system is proposed based on generalized frequency response functions(GFRFs) and least square support vector machine(LSSVM). A closed loop estimation method is used to identify GFRFs of controlled plant under closed loop condition. The kernel principal component analysis(KPCA) method with mixed kernel function is used to compress extract nonlinear spectrum feature. After obtained nonlinear spectrum feature, the LSSVM classifier is constructed for fault recognition. A simulation example about fault diagnosis of a nonlinear closed loop system is provided to illustrate the effectiveness of the proposed method. The results indicate that the proposed method has high accuracy, which can meet the requirements of fault diagnosis.