A Fault Feature Extraction Method Based on Second-Order Coupled Step-Varying Stochastic Resonance for Rolling Bearings

In mechanical equipment, rolling bearings analyze and monitor their fault based on their vibration signals. Vibration signals obtained are usually weak because the machine works in a noisy background that makes it very difficult to extract its feature. To address this problem, a second-order coupled step-varying stochastic resonance (SCSSR) system is proposed. The system couples two second-order stochastic resonance (SR) systems into a multistable system, one of which is a controlled system and the other of which is a controlling system that uses the output of one system to adjust the output of the other system to enhance the weak signal. In this method, we apply the seeker optimization algorithm (SOA), which uses the output signal-to-noise ratio (SNR) as the estimating function and combines the twice-sampling technology to adaptively select the parameters of the coupled SR system to achieve feature enhancement and collection of the weak periodic signal. The simulation and real fault data of a bearing prove that this method has better results in detecting weak signals, and the system output SNR is higher than the traditional SR method.