An Enhanced Hybrid SVD-DBO-VMD Framework for Rolling Bearing Fault Feature Extraction Using Vibration Data.

To address the challenging problem of extracting fault information from rolling bearings, this study proposes a novel hybrid method that integrates singular value decomposition (SVD), Dung Beetle Optimization (DBO), and Variational Mode Decomposition (VMD) for enhanced fault feature extraction. The method consists of three key steps: (1) adaptive SVD denoising via singular-value difference spectrum, where the collected signal is first reconstructed into a phase-space Hankel matrix, and the maximum extremum point of its singular value difference spectrum is identified as the optimal rank order for SVD denoising; (2) DBO of VMD parameters using minimum envelope entropy-to prevent over-decomposition, where DBO is employed to automatically determine the optimal number of modes K by minimizing the envelope entropy, and the signal is decomposed into a set of Intrinsic Mode Functions (IMFs); (3) kurtosis-peak-to-peak joint screening of sensitive IMFs, in which a joint criterion based on kurtosis and peak-to-peak values is applied to select the IMF that contains the richest fault information. This sensitive IMF is then subjected to Hilbert envelope spectrum analysis to accurately extract the fault characteristic frequency of rolling bearings. Experimental results from two different test rigs demonstrate that the proposed method can more effectively highlight periodic fault impulses and identify fault types, offering a reliable approach for rolling bearing fault feature extraction.