The adaptive bearing fault diagnosis based on generalized stochastic resonance in a scale-transformed fractional oscillator driven by unilateral attenuated impulse signal

Abstract

Bearing fault diagnosis is vital to guarantee the safe operation of rotating machines. Due to the enhancement principle of energy conversion from noise to weak signal, noise-assisted stochastic resonance (SR) methods have been widely applied. In this paper, to utilize the memory-dependent property of the mechanical degradation process, we develop a scale-transformed fractional oscillator (SFO) driven by a unilateral attenuated impulse signal, and reveal the active effect of generalized SR (GSR) on the energy conversion from internal multiplicative noise to signal. By applying the quantum particle swarm optimization algorithm in the multi-parameter regulation, we propose the adaptive GSR-SFO diagnosis method to realize the enhancement of weak fault characteristics. The experimental results demonstrate that the proposed method is valid and exhibits superiority in diagnosis performance, especially in several typical difficult cases, such as smeared bearing fault caused by mechanical looseness, smeared bearing fault disturbed by strong random pulses, and corrupted bearing fault disturbed by patches of electrical noise.