An effective identification method of gear fault

In view of problem that feature information of gear fault often submerged by noise in signal, an identification method of gear fault based on harmonic wavelet package, sample entropy and grey correlation degree was proposed. Firstly, the method uses rank-order morphological filter and selects the simplest line structure element combining reality to conduct denoising pretreatment with rank-order morphological filtering for actual measured gear vibration signals. Then, it uses harmonic wavelet package to decompose the gear vibration signals of different faults into eight frequency bands in three levels, and calculates sample entropy of each band. Finally, it takes the sample entropies as elements to construct feature vectors, and calculates grey correlation degree of feature vector of standard fault pattern and test sample to identify working state and fault type. The test result shows that the method can be applied to fault diagnosis of gear system effectively.