Unanticipated background noises often obscure fault information within gearbox vibration responses. The Blind Deconvolution strategy has been extensively employed to enhance fault impulses for aiding in gear fault detection. Existing deconvolution approaches predominantly involve designing optimal filters applied in the time domain. Gear tooth wear induces excitation of Gear Mesh Frequency harmonics, prompting the use of spectral analysis in gearbox fault detection. Thus, feature enhancement in the order domain proves more practical than current blind deconvolution approaches. This study introduces a Spectral Kurtosis-based blind deconvolution strategy, applying filtering in the order domain to facilitate gear fault detection. Experimental analyses demonstrate enhanced performance for constant and real-world speed operations, respectively, supporting the proposed method's utility in ai ding spectral analysis-based fault detection.