Noise-robust rolling bearing fault diagnosis based on JEPPH-driven HSO-FMD

Reliable rolling bearing fault diagnosis is essential for the operation and maintenance of rotating machinery. However, strong noise, random impulses, and non-fault harmonic components can obscure weak fault signatures. To address this problem, a JEPPH-driven adaptive feature mode decomposition (FMD) method is proposed. The Joint Enhanced Periodic Pulse Harmonic Index (JEPPH) combines harmonic energy concentration, peak amplitude and spectral dispersion, harmonic-amplitude matching, and impulsiveness into a composite criterion for parameter optimization and mode selection. Guided by JEPPH maximization, Holistic Swarm Optimization (HSO) jointly optimizes the filter length, number of filters, and number of retained modes. The mode with the highest JEPPH score is selected for Hilbert envelope analysis and fault interpretation using characteristic frequencies. The method is evaluated using simulated signals, the Paderborn University (PU) bearing dataset, and measured test-bench data. Compared with FDS-MOMEDA, CESM-based IMD, and HSO-VMD, HSO-FMD achieves the highest envelope-spectrum fault-characteristic signal-to-noise ratio in the three analyzed cases: 11.778, 10.227, and 2.011 dB, respectively. The recovered harmonic structures agree with the theoretical characteristic frequencies of the documented faults, supporting fault-feature enhancement under the investigated interference conditions.