Fault diagnosis of dual-channel CNN-SVM of RV reducer bearing based on multiwavelet transform

The strong background noise characteristics of the dual channel CNN-SVM bearing vibration signal based on wavelet decomposition leads to poor feature recognition ability of the fault diagnosis model. A bearing fault intelligent diagnosis method based on multiple wavelet packet decomposition reconstruction and dual channel CNN-SVM is proposed to address this issue. Firstly, the signal is frequency band decomposed using wavelet packets with six different wavelet bases. Secondly, the method of Joint Approximative Diagonalization of Eigen-matrices (JADE) is employed to carry out feature integration on each set of frequency bands, mitigating the influence of noise and deriving combined features across the frequency spectrum, from low to high. A weight allocation method is introduced to prioritize sensitive features and obscure non-essential feature information through the application of weight allocation. Finally, to achieve intelligent diagnosis of bearing faults, a dual channel CNN-SVM bearing fault diagnosis framework was constructed. The effectiveness of the proposed method was validated through experiments. The results indicate that the method proposed in this paper has a high ability to identify fault features.