Weak signal enhancement for machinery fault diagnosis based on a novel adaptive multi-parameter unsaturated stochastic resonance

The output saturation problem limits the application of classical bistable stochastic resonance (CBSR) in weak fault diagnosis. Furthermore, the adjustment of system parameters is essential to generate stochastic resonance (SR) for achieving the optimal output. In this paper, a novel adaptive multi-parameter unsaturation bistable stochastic resonance (AMUBSR) system based on piecewise linearization of potential function is proposed. The beetle antennae search (BAS) is adopted to optimize the system parameters, and the output signal-to-noise ratio (SNR) is selected as the objective function. The optimization results of the simulation signal prove that the validity and superiority of BAS algorithm in parameter matching compared with particle swarm optimization (PSO) algorithm and ant colony optimization (ACO) algorithm. Finally, the proposed method obtains the further improvement of the output SNR, higher spectrum peak value at characteristic frequency and bigger recognition quantity than the CBSR method from the diagnosis results of bearing inner ring and outer ring fault signals.