Abstract:  Aiming at the problem of fault feature extraction and pattern recognition of planetary gearbox, a planetary gearbox fault diagnosis method combined honey badger algorithm (HBA) optimal improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), hierarchical weighted permutation entropy (HWPE) and grey wolf algorithm (GWO) optimal support vector machine (SVM) was proposed. Firstly, HBA was used to optimize the white noise amplitude weight and noise adding times of ICEEMDAN, and HBA-ICEEMDAN decomposition was performed on the vibration signal of planetary gearbox to obtain several mode functions, and the components with larger correlation coefficients were selected for reconstruction. Then, sensitive feature values of reconstructed low noise signals were extracted by HWPE method to obtain fault feature vectors. Finally, GWO was used to optimize the penalty coefficient and kernel coefficient of SVM, and GWO-SVM multi fault classifier was trained to realize damage identification of planetary gearbox, and the experiments were carried out with the vibration data of planetary gearbox, and the effectiveness of the algorithm was verified. The research results show that the fault diagnosis method of planetary gearbox combining HBA-ICEEMDAN, HWPE and GWO-SVM can accurately identify typical single point faults and composite faults of planetary gearbox, with an identification accuracy rate of 98.15%. Comparing with other combination methods, the method has more advantages and effectiveness in fault diagnosis of planetary gearbox.
Key words:  gear transmission; honey badger algorithm (HBA); improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN); hierarchical weighted permutation entropy (HWPE); grey wolf algorithm-optimal support vector machine (GWO-SVM); planetary gearbox; fault diagnosis