Abstract: Aiming at the problem that the nonlinearity of gearbox vibration signal made it difficult to effectively extract the damage feature, a fusion damage identification method based on dual-frequency refined composite multi-scale permutation entropy (DFRCMPE) and whale algorithm optimized support vector machine (WOA-SVM) was developed. First, the wavelet packet decomposition (WPD) was used to decompose the gearbox damage vibration signal in two layers to obtain the low-frequency and high-frequency components reflecting the gearbox damage characteristics. Then, the refined composite multi-scale permutation entropy (RCMPE) was used to analyze the two groups of frequency band components to fully extract the damage information embedded in the vibration signal, and constructed the damage characteristics. Finally, the damage features were input into the whale algorithm optimization support vector machine classification model to achieve intelligent damage identification; and the effectiveness of the fusion damage identification method based on DFRCMPE and WOA-SVM was discussed by taking the gearbox vibration signal collected in the experiment as the research object. The research results show that comparing with the feature extraction methods based on refined composite multi-scale entropy (RCMSE), refined composite multi-scale fuzzy entropy(RCMFE), RCMPE and refined composite multi-scale dispersion entropy(RCMDE), the fusion damage identification method based on DFRCMPE and WOA-SVM has higher accuracy and stability, and the average recognition accuracy reaches 100%; this method can provide a feasible way to solve the problem of gearbox fault identification in practical applications.
Key words:  gear transmission; damage feature extraction; gearbox vibration signal; dual-frequency refined composite multiscale permutation entropy (DFRCMPE); whale algorithm optimized support vector machine (WOASVM); wavelet packet decomposition (WPD)