Fractional Stochastic Resonance for Weak Fault Detection of Machinery

Conventional integer order stochastic resonance (SR) based fault diagnosis methods overlook the natural memory and hereditariness of fractional order calculus. These properties can more accurately model the dynamic behaviour of real world systems, and their absence limits the diagnostic performance. A novel fault diagnosis approach grounded in fractional order stochastic resonance (SR) is presented, aimed at boosting the extraction of faint fault signatures in settings with low signal to noise ratios (SNR). Tests on early stage faulty bearings with ball spalling show its effectiveness in extracting early stage fault features. By comparing it with the conventional fast spectral kurtosis based denoising method, the superiority of the proposed method is further highlighted.