Rotor vibration fault diagnosis method based on WCFSE-FSVM

To improve the diagnostic precision of rotor vibration fault samples with the noise and outliers, the fault diagnosis method based on the wavelet correlation feature scale entropy (WCFSE) and fuzzy support vector machine (FSVM) (WCFSE-FSVM) was proposed. The fault diagnosis model of the WCFSE-FSVM was established by fully fusing the strength of the WCFSE feature extraction method and the FSVM fault diagnosis method. The original fault data were gained through simulating four typical faults on rotor test-bed. The WCFSE values of these data were extracted by the WCFSE method, and W₁ and W₂, which are the WCFSE values on the scales 1 and 2, respectively, in the high band of fault signals, were selected to construct the fault vectors of vibration signals as the fault samples for establishing the FSVM diagnosis model. As shown by instance analysis, the WCFSE-FSVM in four methods possesses the highest diagnosis precision that the fault type and severity diagnosis precisions of rotor vibration are 94.49% and 95.58%, respectively. This paper demonstrates the validity and feasibility of proposed WCFSE-FSVM and provides an effective method for rotor vibration fault diagnosis.