Damage detection of flexible thin-walled elliptical bearings using optimized sparse maximum harmonic-to-noise ratio deconvolution and multipoint optimal minimum entropy deconvolution adjusted

Flexible thin-walled elliptical bearings (FTEBs), essential to harmonic reducers, generate vibration signals dominated by strong background elliptical impacts and random noise, which obscure weak fault features. To overcome this challenge, we propose a novel hybrid diagnostic framework that integrates optimized sparse maximum harmonic-to-noise ratio deconvolution (OSMHD) with multipoint optimal minimum entropy deconvolution adjusted (MOMEDA). The framework introduces two innovations: (1) a frequency-domain energy characteristic ratio (FECR) combined with the Superb Fairywren Optimization Algorithm (SFOA) to adaptively determine filter parameters, and (2) a fault frequency energy ratio (FFER) index to quantitatively assess feature extraction. By sequentially applying OSMHD and MOMEDA, residual impacts and noise are effectively suppressed, enabling clear recovery of fault characteristic frequencies and their harmonics. Experimental verification on outer- and inner-ring faults demonstrates that the proposed approach consistently outperforms envelope spectrum analysis, OSMHD, MOMEDA, and OVMD–MCKD, while robustness tests under strong noise confirm its superior anti-interference capability.