Feature extraction of rolling bearings in cow dung briquetting machine based on CEEMDAN-WTD-DBO

To address the difficulty in identifying weak early fault impacts and fault characteristic frequencies of rolling bearings in vertical cattle-manure briquetting machines under strong background noise caused by low-speed, heavy-load operation and material compaction, this study proposes a denoising and fault feature extraction method based on the adaptive fusion of CEEMDAN, improved wavelet threshold denoising, and DBO. By constructing a kurtosis–permutation entropy composite evaluation index to select sensitive IMF components and using the DBO algorithm to adaptively optimize the adjustment factors in the improved threshold function, the proposed method effectively enhances bearing fault impulses under strong-noise conditions. Simulation results showed that, at an input signal-to-noise ratio of approximately − 15 dB, the proposed method increased the signal-to-noise ratio to − 1.400 dB and reduced the mean square error to 0.002152, outperforming single CEEMDAN and conventional wavelet threshold denoising. In measured bearing signals from a vertical cattle-manure briquetting machine with rolling-element spalling, the 8.5 Hz fault characteristic frequency and its harmonics were clearly extracted. The method provides an effective signal-processing approach for bearing condition monitoring and early fault diagnosis in agricultural and animal husbandry equipment.