Scale-Demodulation Adaptive Chirp Mode Decomposition With Application to Rolling Bearing Fault Diagnosis

Rolling bearing fault signals under variable-speed conditions commonly represent nonstationary and nonlinear characteristics. It is a challenging problem to implement the bearing fault diagnosis under variable-speed conditions. In recent years, adaptive chirp mode decomposition (ACMD) has been proposed as a nonstationary signal decomposition technique with strong adaptability. However, ACMD exhibits unsatisfactory performance in analyzing signal components with mode-close or instantaneous frequency (IF) intersected. Hence, this paper presents a novel signal mode decomposition method called scale-demodulation ACMD (SDACMD). First, utilizing the Hilbert transform enhances the concentration of fault features and then initializes the IF of maximum frequency modes using the scale demodulation algorithm. Next, based on the orthogonalization criteria and the original ACMD algorithm, a bandwidth-adaptive criterion is derived. This criterion effectively alleviates the interference of noise by automatically updating the bandwidth. In addition, an optimal parameter selection algorithm based on mutual information (MI) is introduced to further enhance the precision and adaptability of SDACMD. Finally, simulation and experimental results demonstrate the effectiveness of SDACMD in handling simulated nonstationary signals and applying bearing fault diagnosis under variable-speed conditions.