Fault diagnosis for multivariable non-linear systems based on non-linear spectrum feature

In this study, a novel fault diagnosis approach based on a non-linear spectrum feature is proposed for a multivariable non-linear system. The non-linear spectrum features are obtained using a non-linear output frequency response function (NOFRF) and kernel principal component analysis (KPCA). In order to improve the real-time performance of obtaining non-linear spectrum features, a frequency domain variable step size normalized least mean square (FVLMS) adaptive algorithm is presented to identify NOFRF. A multi-fault classifier based on the fusion of a support vector machine (SVM) is designed according to different frequency domain scales, and a fusion method by using sub-classifier classification reliability is proposed. A simulation example about a two-input-two-output non-linear system is provided to illustrate the effectiveness and performance of the proposed approach.