Comb-Filtering-Based Signal Reconstruction for Bearing Fault Detection

Faulty bearings generate impulsive vibrations during operation and this provides the basis for bearing fault detection by vibration measurement and analysis. However, in a situation where a mechanical system produces strong impulsive vibrations by its nature even when the system is in good condition, the traditional impulse detection based method for bearing fault diagnosis become ineffective. To solve this problem, this paper proposes a new methodology that first removes the strong impulses generated by normal operation of the system by comb-filtering of the original vibration signal in frequency domain, then reconstructs the time signal by inverse Fourier transform of the filtered spectrum, and finally calculates the crest factor of the reconstructed signal. The proposed method was proved effective in its application to highly impulsive plunger-pump-driving gearboxes.