Fault diagnosis of gearbox using multi-time-delayed feedback stochastic resonance

Abstract

Gearbox vibration signals with defects often show some important characteristic information. However, the feature information is difficult to extract since it is often flooded in the noise. To solve this problem, we present a novel detection approach for weak signals based on the multi-time-delayed feedback stochastic resonance model (MTFSR). Because the shape of the bistable potential well can be changed by the number of time-delayed terms, the MTFSR method can use historical information which is formed by the superposition of multiple time-delayed feedback items to enhance signal periodicity and obtain better output waveform and higher signal-to-noise ratio (SNR). Moreover, the presented method is also insensitive to the noise, and can detect the weak signals with different noise levels. Additionally, by selecting the proper calculation step, the approach has the ability to detect weak signals with different driving frequencies. With these properties, the proposed MTFSR is considered to be very suitable for gearbox fault diagnosis. Both simulation study and practical application confirm that the proposed strategy is feasible and superior in comparison with some traditional methods.