A novel maximum impulse amplitude deconvolution for rolling bearing weak fault identification enhanced by spectrum information guided VMD

The analysis of vibration signals constitutes a widely employed diagnostic methodology for detecting bearing faults. In real harsh operational environments, periodic fault-induced impulses, which represent a characteristic manifestation in vibration signal, pose significant challenges in feature extraction under the strong noise contamination, particularly for weak faults. Deconvolution methods are commonly used for weak feature extraction, but the existing classical methods still facing the problem of insufficient feature extraction capability under complex strong noise interference. Therefore, this paper proposes a new periodic impulse amplitude ratio (PIAR) index based deconvolution algorithm for improving this issue. To further enhance the weak feature and suppress noise, an improved VMD method based on spectral information (SIVMD) is further proposed to obtain the frequency band that contains the most fault information. Based on the full combination of SIVMD and MIAD, weak fault information can be effectively enhanced. To be specific, firstly, the penalty factors, decomposition numbers and initial center frequencies of VMD are adaptively optimized by the spectral trend and the nonlinear function, instead of using the traditional intelligent optimization method, to obtain the optimal frequency band. Secondly, the proposed MIAD is used to filter the noisy signal. Finally, the enhanced extraction of fault characteristics can be implemented through envelope analysis. The simulated and experimental signal analysis are utilized to verify the promising performance of proposed SIVMD-MIAD method compared with existing widely used methods.