Study On The Weak Feature Extraction And Fault Diagnosis Of Rolling Bearings Under Strong Noise Background

As one of the most widely used key components in rotating machinery,rolling bearings are highly liable to malfunction due to harsh working environment.Thus,the normal operation of the entire equipment will be affected,which may lead to not only the shutdown of the production chain,resulting in some economic losses,but also catastrophic casualties and serious social hazards.Therefore,to carry out rolling bearing fault diagnosis technology research,especially for early fault diagnosis,has important scientific significance and engineering value.However,in practical engineering,the early fault features of rolling bearings are so weak that the obtained bearing vibration signals are contaminated by strong noise and have very low signal-to-noise ratio,which brings some difficulties to the fault diagnosis of rolling bearings.Therefore,this paper is based on stochastic resonance theory and proposes two weak feature extraction methods under strong noise background to provide some references for the early fault diagnosis of rolling bearings.Details are as follows:Firstly,based on the common failure mode of rolling bearings,the bearing vibration mechanism is summarized.Meanwhile,the rolling bearing fault simulation test rig is set up to analyze the characteristics of the vibration signal when different components are faulty.Secondly,the classical bistable stochastic resonance theory and its weak feature detection mechanism are introduced.For the output saturation phenomenon of the classical bistable model,a piecewise-linear model of unsaturated stochastic resonance is established.At the same time,an adaptive piecewise-linear stochastic resonance method is proposed to extract the weak features.Here,the general scale transformation principle is used to preprocess the large parameter signal,which makes it meet the requirements of stochastic resonance.In addition,to obtain the best stochastic resonance output,the quantum particle swarm optimization is used to search the optimal parameters of the piecewise-linear system.The validity of this method is verified by numerical simulation.Then,on the basis of the single stochastic resonance,a cascade adaptive piecewise-linear stochastic resonance method is proposed.In the proposed method,a series of single adaptive piecewise-linear systems are connected so that the high-frequency energy is continually transferred to the low-frequency region.Consequently,the low-frequency weak feature energy is gradually enhanced.Through numerical simulation and experimental signal analysis,it is verified that this method can further improve the output signal-to-noise ratio and realize the weak fault feature extraction of rolling bearings under strong noise background.Finally,aiming at the poor quality of empirical mode decomposition of the bearing fault signal under strong noisy background,an improved empirical mode decomposition method based on the cascade piecewise-linear stochastic resonance denoising is proposed.In the proposed method,the cascade adaptive piecewise-linear stochastic resonance method is used to preprocess the strong noisy signal.Afterwards,the denoising signal is decomposed and the features are extracted.Numerical simulation and experimental signal analysis show that this proposed method can improve the quality of empirical mode decomposition and realize the accurate extraction of the weak fault features of rolling bearings under strong noise background.