A novel adaptive stepwise wavelet mode decomposition method for compound fault diagnosis of rolling bearings

When rolling bearings appear compound faults, multiple fault features contained in the vibration signal generated by it are coupled and masked, making it difficult to accurately separate and identify them. This indicates that the compound fault diagnosis of rolling bearings is regarded as a difficult problem in the field of mechanical system state monitoring and diagnosis. Therefore, to accurately separate and extract a single fault component from rolling bearing compound fault signals, this paper proposes a novel adaptive stepwise wavelet mode decomposition method for compound fault diagnosis of rolling bearings. Firstly, the number of decomposition modes is adaptively determined by locating and extracting multiple fault frequency bands of the collected bearing compound fault signals. Secondly, the spectral leakage is reduced by segmenting the fault signal with Hanning windows and adaptively selecting a wavelet basis for each window. Next, the fault-related information is locked while the redundant and mixed modes are removed through the periodicity estimation and iterative updating method. Additionally, the optimal mode is chosen using the impact time-frequency indicator, and key process parameters are optimized with the kangaroo escape optimizer. Finally, based on the wavelet transform and second-order cyclostationarity theory, the compound faults are further extracted and separated. Simulation and experimental data have verified the effectiveness of the proposed method. The results show that the proposed method has the superiority in adaptively separating the compound fault features and reducing the noise interference.