Novel Bearing Fault Diagnosis Algorithm Based on the Method of Moments for Stochastic Resonant Systems

The principle of stochastic resonance (SR) in the noisy Duffing oscillator model has shown benefit for designing novel mechanical fault diagnosis algorithms, where noise is utilized rather than being eliminated. However, there is a clear gap between the model progress and the experimental application. In this article, effort is made trying to narrow the gap by applying the method of moments to obtain the spectral amplification factor within the linear response range to improve the algorithm design, which avoids the conventional time-consuming direct simulations. A strategy for estimating noise, which is critical for programming the algorithm, is proposed and evaluated. Through simulation and experimental data analysis, it is confirmed that the new algorithm has advantages over the overdamped system-based methods as it does not depend on the signal preprocessing techniques such as envelope extraction and high-pass filter. Also, the new method has advantages over the existing underdamped system-based methods as it can decrease the computational time for seeking the optimal parameter by at least one order of magnitude.