Multivariable dynamic system fault diagnosis using nonlinear spectrum and SVM fusion

The fault diagnosis of multivariable dynamic system is studied by combining nonlinear spectrum and support vector machine fusion method. In order to overcome the calculated amount expansion problem of generalized frequency response function (GFRF), the one-dimensional nonlinear output frequency response function (NOFRF) is used to obtained nonlinear spectrum data. After obtaining nonlinear frequency spectrum data, the spectrum features are extracted by kernel principal component analysis(KPCA) method. To fully consider global characteristics and local characteristics of spectrum data, a kind of mixed function is used as kernel function of KPCA model. According to different frequency domain scale characteristics, a multi-fault classifier based on support vector machine (SVM) fusion is constructed, and the fusion method based on sub-classifier classification reliability is proposed. Experiment results indicate that the proposed fault diagnosis method has higher recognition rate so that it has important practical value.