Research on multi-fault identification method of axial piston pump based on correlation dimension

Aiming at the problem of multi-fault identification method of axial piston pump, a new method combining correlation dimension and support vector machine was proposed. Firstly, the envelope demodulation was performed to several typical fault vibration signals of pump by using wavelet packet decomposition and Hilbert transform method. Then, the useful signals corresponding to feature band were obtained. Secondly, using the embedded dimension as characteristic attribute, the correlation dimension feature information of the useful signals was extracted by correlation dimension analysis method. At the same time, the feature sample set consists of the correlation dimension feature information and the energy feature information was built. Finally, the feature sample set was utilized as input, and the multi-fault modes of the axial piston pump were identified by support vector machine. Moreover, the experiments were respectively performed to the pump including slipper wear, loose slipper failure and central spring failure. The results confirmed that the fault recognition accuracy of SVM can be improved, and the faults can be effectively identified.