Due to the bad working environment of a tank planetary gearbox, the gear is prone to crack and other faults. The accurate and effective identification of crack fault has guiding significance for timely maintenance. The entropy value of vibration signal changes after gear failure, so the information entropy feature is widely used in planetary gear fault recognition.
Yang Dawei et al. And Ren Guochun et al. Used VMD sample entropy and EEMD sample entropy to realize planetary gear crack fault identification; Wang Zhenya and Huang Darong used multi-scale sample entropy to realize bearing fault identification; Ding Chuang et al. Used LMD component permutation entropy and improved multi-scale symbolic dynamics information entropy to realize fault identification of sun gear and planetary gear cracks; Wang Zhijian et al. Used permutation entropy to optimize VMD parameters to realize gearbox fault diagnosis; Li Yongjian et al. And Qi Xiaoli et al. Used multi-scale permutation entropy feature to realize fault pattern recognition of bearings and planetary gears; Zheng Jinde et al. Used multivariable and multi-scale fuzzy entropy to realize fault diagnosis of planetary gears. However, the calculation efficiency of sample entropy is low, and the detailed changes of signal amplitude are not considered in permutation entropy. For this reason, rostaghi et al. And Azami et al. Proposed dispersion entropy (DE) and fine composite multi-scale dispersion entropy (rcmde) features to evaluate the complexity of nonlinear signals. The feature algorithm is simple and efficient; Li Congzhi et al. Used rcmde for bearing fault feature extraction; Qiao Xinyong proposed VMD multi-scale dispersion entropy to realize diesel engine fault identification; Fu Wenlong et al. Used VMD component dispersion entropy and improved gray wolf optimization support vector data description to realize bearing fault identification; Zhang Yidong proposed multi-scale fusion of discrete entropy to improve the stability and accuracy of features.
However, rcmde features still have some disadvantages, such as low noise robustness and low scale selection efficiency. In this paper, adaptive fine composite multi-scale dispersion entropy (arcmde) is proposed. Firstly, VMD is used to denoise the signal, and feature coincidence index is proposed to rank the scales to improve the scale selection efficiency, Finally, the proposed features are verified by the experimental data of planetary gearbox.
