An improved bearing fault diagnosis method based on variational mode decomposition and adaptive iterative filtering (VMD-AIF)

Rolling bearings are widely applied in rotary machines. Bearing failures can lead to long machine downtime and costly maintenance. To reduce the maintenance time and cost, this article proposes a new denoising approach to remove heavy noise and extract weak bearing fault vibration features. The first step proposes a global variational mode decomposition (VMD) optimization algorithm that adaptively matches the optimal decomposition parameters based on kurtosis. In the second step, an adaptive selection criterion for IMF is developed by kurtosis theory. Meanwhile, an adaptive termination criterion is established based on sample entropy (SampEn), and the fault features are extracted with a Butterworth band-pass filter (BBF). Finally, actual fault-bearing experiments are performed using the proposed denoising approach. The comparison with respect to EEMD and WPD is illustrated in detail. The experimental result shows that the proposed approach is a validated tool for the diagnosis of bearings.