An adaptive enhanced envelope spectrum technique for bearing fault detection in conditions characterized by strong noise

Rolling bearings are widely used in rotating machinery and have a high failure rate. Regrettably, the task of ensuring dependable bearing fault detection presents a formidable challenge, especially when the bearing fault-related characteristics are non-stationary or even affected by strong noise. In response to this challenge, a novel adaptive enhanced envelope spectrum (AEES) technique is proposed in this study. Firstly, it generates representative intrinsic mode functions (IMFs) using the variational mode decomposition algorithm. Then, based on the analysis of the envelope spectrum normalized mutual information and time-domain fuzzy entropy, a new IMF selection and integration strategy combining time- and frequency-domain metrics is suggested to reconstruct the most informative analytical signal. An adaptive filter is employed to post-process the reconfigured signal to reinforce fault-related impulsive characteristics, the optimal length of which is ascertained through the proposed variable step-size search technique based on unbiased autocorrelation analysis. The efficacy of the AEES technique has been validated through a sequence of experiments conducted under diverse bearing conditions. Its robustness and distinct advantages under strong noise conditions are tested using a publicly available dataset. The validation results show that the AEES technique can effectively identify the health conditions of bearings under high noise conditions (signal-to-noise ratios between 1 dB and 3 dB). Compared with two relevant techniques in the existing literature and a classical method, the proposed AEES technique can achieve signal processing results with fewer interference components and more prominent characteristic frequency information and has a unique ability to identify fault features in some challenging situations.