In order to extract fault, based on nonlinear feature extraction capability of kernel principal component analysis(KPCA) and good classification performance of least squares support vector machine (LSSVM), local mean decomposition (LMD) and KPCA-LSSVM mechanical fault diagnosis algorithm is proposed. When a device failure occurs, a variety of statistical parameters of vibration signals contain a wealth of state characteristics that is a relevant and redundant, which will reduce the generalization ability and the recognition accuracy of the classifier. A series of production function PF component of the signal obtained by LMD, using KPCA to remove redundant features of sample data which include the vibration signal sequence domain characteristic parameters and timing AR parameters, as well as energy entropy PF component. To extract the nonlinear principle component in the input data space, and then fault classification using LSSVM. Experimental results show that, PF energy entropy characteristics are better than temporal characteristics and timing parameters of AR parameters, and KPCA-LSSVM classification model with respect to the direct use of LSSVM has better classification accuracy.

Full text: IJEIR_798_Final.pdf