Rotor fault diagnosis based on wavelet packet energy spectrum and adaptive fuzzy weighted support vector machine

In this study, a novel application of a wavelet packet energy-weighted support vector machine (WPE-WSVM) is proposed to perform fault classification of helicopter rotor. Because the helicopter rotor fault signal is weak, it is difficult to extract fault feature. The wavelet package is adopted to decompose the vibration signals on the fuselage into different frequency bands, and to eliminate the noise. And then single signal was reconstructed to extract the energy in each frequency band serving as fault feature vectors. And support vector machine was applied for classifying the failure mode of the helicopter rotor. For classification task support vector machine is used due to its good robustness and generalization performances. But the classification accuracy of standard support vector machine is relative slow when the number of samples of different classes is dramatically different. So a fuzzy weighted support vector machine was proposed, which added weight coefficient to samples of different classes. A comparative analysis of standard support vector machine and proposed fuzzy weighted support vector machine is done. The proposed fuzzy weighted support vector machine improved the classification accuracy of class with fewer samples. The proposed method is sufficiently accurate, fast, and robust, which makes it suitable for use in helicopter rotor fault diagnosis applications.