Rolling Bearing Fault Detection Based On Wavelet Packet Energy Spectrum - Principal Component Analysis

Rolling is one of the most widely used in almost all kinds of mechanical devices, simultaneously, their components are vulnerable to damage. Many failures of mechanical equipment are concerned with the state of the rolling bearing operation, which runs in good condition or not directly affect the function of the entire device. Therefore, it is very significant in rolling bearing fault detection. For the vibration signal data of rolling is analyzed, which is a effective way for monitoring the operating state of the rolling bearing. Therefore, the method of fault identification was proposed in this paper, which is based on wavelet packet energy spectrum. It used wavelet packet analysis method, extracted the energy feature of vibration signal. And as a basis for establishing fault recognition model, it is used for he detection and identification of the vibration signal fault information. In the context of experimental data from the American Case Western Reserve, the proposed method for simulation analysis. The outline of the work is as follows:(1) The method of fault detection is used wavelet packet energy spectrum and principal component analysis for establishing fault detection model. Firstly, the vibration signal was decomposed using wavelet packet, extracted eigenvectors of wavelet packet energy spectrum. And it as the input of principal component analysis, in order to establish fault detection model. It can detect bearing failure by SPE statistics and T2statistics.(2) The method of fault detection are presented, it adopted wavelet packet energy spectrum and parse kernel principal component. Kernel principal component needs to build nuclear matrix, whose dimension is equal to the number of samples, resulting in the curse of dimensionality, computational complexity and other issues. For this reason, a method of sparse kernel principal component was used for fault detection in this paper. In this method, the sample base of energy spectrum was extracted through the method of incremental sample base. A kernel principal component model was built by the sample base for detecting bearing failure. Compared with the standard kernel principal component analysis, sparse kernel principal component method can effectively reduce the dimension of the nuclear matrix. Thereby, the computational complexity of kernel principal component was reduced, and the computational efficiency was improved.(3) It extracted vibration signal of wavelet packet energy spectrum in this paper, and constructed energy feature vector of signal. And using principal component, kernel principal component and sparse kernel principal component for the energy feature vectors to establish fault recognition model. And which is applied to the rolling bearing vibration signal data of American Western Reserve University, it provided to verify the effectiveness of these three methods of analysis.(4) Finally, a bearing fault detection software was designed in this paper, it is a mixed programming based on MALTLAB and C++Builder software. The proposed fault recognition model is the core of the detection software.