Stochastic Resonance Enhancement Theory And Method For Weak Fault Signal Of Rotating Machinery

Effective extraction of mechanical fault characteristic information through signal processing is crucial for achieving machine condition monitoring and fault diagnosis,ensuring high-quality equipment operation.However,factors such as component coupling between transmission paths,noise in the acquisition system,and environmental noise during equipment operation result in low signal-to-noise ratios of signals acquired by sensors,making it difficult to detect early-stage fault characteristics of mechanical equipment.Unlike feature extraction techniques based on noise suppression or elimination,the stochastic resonance theory can utilize noise to enhance feature signals.In order to improve the accuracy and reliability of early-stage detection of weak faults in rotating machinery,and provide strategies for machine condition monitoring and operation,this paper focuses on rotating machinery and establishes stochastic resonance models suitable for different fault characteristics.The resonant mechanisms of each model are discussed,and corresponding methods for enhancing fault characteristics of rotating machinery based on stochastic resonance theory are developed.A high-order time-delay feedback tri-stable stochastic resonance model is proposed to address the shortcomings of the existing first-order time-delay feedback stochastic resonance model in enhancing fault features.The high-order time-delay feedback is utilized to improve the memory characteristics of the stochastic resonance system,and to enhance the system’s detection and identification of bearing periodic fault feature signals.Simulation and experimental results demonstrate that the high-order time-delay feedback tri-stable stochastic resonance model can achieve higher output signal-to-noise ratio than the firstorder time-delay system,and the system’s fault feature frequency exhibits higher spectral peaks.A method for enhancing fault features of the second-order underdamped tri-stable stochastic resonance model is proposed based on the low-pass filtering effect of the stochastic resonance model on signals.By simultaneously considering the inertial term and the damping term,the stochastic resonance system is endowed with second-order filtering function,which improves the output signal-to-noise ratio of the system.The steady-state solution curve is introduced to investigate the output response of the model,and the model is electrically circuitized to form a second-order underdamped tri-stable stochastic resonance weak signal enhancement circuit.The proposed method is used to enhance the fault features of rolling element faults in bearings and gear wear faults in wind turbine drivetrains.Experimental results show that,under the joint action of system order and potential well,the weak signal feature enhancement capability of the underdamped tri-stable system is superior to that of the common damped stochastic resonance system.A method for enhancing fault features in a bi-stable stochastic pooling network is proposed to address the problem of output saturation in a single stochastic resonance model.The adaptive weight adjustment of network nodes is employed to improve the output stability of the array stochastic resonance system and overcome the inability to balance the output results of each array element.The theoretical output signal-to-noise ratio of the system demonstrates the stochastic resonance effect and weak signal enhancement capability of the bis-stable stochastic pooling network.In the experimental verification part,the proposed method is used to enhance the fault vibration signals of bearings,followed by a bearing fault diagnosis scheme based on current signal feature enhancement.Simulation and experimental results show that the bi-stable stochastic pooling network can achieve bearing fault diagnosis from both vibration signal and current signal perspectives.Taking into account the advantages of periodic potential stochastic resonance for signal feature enhancement,an underdamped modulated periodic potential stochastic resonance model is established.By proposing a modulated periodic potential function,the problem of fixed potential barrier height and inability to change the steady-state point with the position of the potential well in a common periodic potential stochastic resonance model is solved.A network is constructed using the modulated periodic potential stochastic resonance model as nodes.The integral modulation bi-spectrum factor is proposed for detecting signal modulation components,and an adaptive bearing fault diagnosis method based on modulated periodic random pool network is developed.In the experimental verification part,the vibration signals of motor casing and spindle are used to verify the fault detection ability of the modulated periodic stochastic pooling network.The results show that the proposed method can achieve bearing diagnosis under unknown faults,and its ability to enhance weak fault features is significantly better than that of the classical periodic potential stochastic resonance system.