Unknown bearing fault diagnosis under time-varying speed conditions and strong noise background

The bearing vibration signal shows strong non-stationary property under time-varying speed conditions. In addition, the weak bearing fault characteristic is often submerged in strong background noise. How to accurately extract the unknown fault characteristic from the non-stationary vibration signal is the primary problem of bearing fault diagnosis. Stochastic resonance has been proved to be an effective weak signal enhancement method. Therefore, an unknown bearing fault detection technology of speed variation is proposed, which breaks through the periodicity limitation of the classical stochastic resonance on the input signal. It enables stochastic resonance suitable for the enhancement of non-stationary fault signal. Firstly, the non-stationary vibration signal is processed by the computed order tracking to obtain the stationary signal in angular domain. To extract the potential feature information, the bearing imaginary fault order index is constructed from the angular domain order spectrum. Then, the resonance response at the imaginary fault order is obtained. Finally, the coherence resonance theory is introduced to judge the bearing fault pattern through the resonance factor index of response order spectrum. The proposed method overcomes the fuzzy mapping relationship between the signal symptom and the bearing fault caused by speed variation. The experimental data analysis results provide effective support for the proposed method.

Automated summary

The bearing fault diagnosis is an important problem due to the non-stationary nature and background noise interference of bearing vibration signals. This study proposes a method for extracting unknown fault characteristics from non-stationary vibration signals using stochastic resonance technology. The method successfully determines the bearing fault pattern through order tracking and coherence resonance theory.