Abstract:
Condition-based maintenance is important for reducing maintenance cost and increasing the life of rotating machines. Conventionally, it relies on fault diagnosis using frequency domain features observed in the envelope spectrum of the vibration signal. The proposed method improves such features by calculating the fractional envelope of the signal. Our study shows that out of available methods of calculating the fractional envelope, the fractional Fourier transform based method is most suitable for bearing fault detection and diagnosis. The results are further improved using maximal overlap wavelet packet transform. The advantage of the proposed method is shown in comparison with a newly introduced autogram method using several cases from the Case Western Reserve University’s bearing dataset.