Methodologies Of Weak Signal Detection Based On Frequency-exchange Stochastic Resonance

Weak signal detection is a key technology in mechanical fault diagnosis.Its mature application can not only accurately detect the equipment faults as early as possible,but more importantly,it can avoid unnecessary economic losses and casualties.Although nonlinear methods,such as the Stochastic Resonance(SR)theory,Duffing oscillator,have been extensively and deeply studied,their engineering applications are still limited by a certain restrictive condition.Therefore,the condition of the bistable stochastic resonance system is discussed and some solutions are put forward in this dissertation.And finally,the further application of weak signal detection based on SR theory is realized in mechanical fault detection,which is also the main purpose of this research.Although some large-parameter signal detection methods based on SR,such as Parameter-tuning SR and Rescaling Frequency SR(RFSR),can overcome the smallparameter conditions of classical SR,they are also not able to detect the target signals with low sampling ratio.The low sampling ratio of the target signal is defined as the ratio of the sampling frequency to the characteristic frequency.In general,to ensure that the methods based on SR have better detection performance,the sampling ratio of target signal needs to be greater than 50.In addition,in order to make sure that SR system has a maximum output SNR,it has to adjust the system parameters.The adjustment of system parameters will affect the potential well width and barrier height simultaneously,which is not conducive to the intuitive analysis of the influence of the change in potential function on the SR system.In order to solve the above two problems,the following aspects are mainly studied and the corresponding solutions are provided in this work.Firstly,the method of frequency information exchange(i.e.,frequency exchange)is proposed and introduced into the classical SR theory.The mechanism of frequency information exchange(FIE)is demonstrated and analyzed based on the theory of singleside band modulation.And the implementation schemes of the SR method based on FIE in frequency domain and time domain are given in the dissertation.To overcome the sampling frequency ratio constraint of RFSR,a new method called Frequency Exchange and Rescaling SR(FERSR)is proposed based on FIE and RFSR.FERSR can be used to detect single-frequency and multi-frequency target signals.Compared with the traditional FFT and RFSR,FERSR has better detection performance and no limit of sampling frequency ratio.Simulations and engineering applications also verify the effectiveness of the method.Secondly,in order to intuitively understand the influence of potential well width and barrier height on stochastic resonance system,the coefficients of the potential well width and barrier height are introduced into bi-stable potential function.The trends of the escape rate,the transition critical value,and the output response of the bistable system are analyzed in the dissertation.With the increase of the potential well width,the system’s escape rate and transition critical value will decrease monotonically,and the output response amplitude of the system will increase.The increase of the barrier height will lead to the increase of the transition critical value of the system and the decrease of the system output.And the escape rate has a complex trend that increases first and then decreases with the increase of the barrier height.In addition,the introduction of the potential function characteristic parameters(PFCPs)which are barrier height and potential well width,is also investigated.And a new SR method based on PFCPs tuning and frequency exchange(PTFESR)is proposed in this dissertation,which can also detect the target signal with a low sampling ratio.Finally,the two proposed SR methods(FERSR and PTFESR)are used to detect the faults of the CRH’s bogie bearing.To simulate the real failure status of the bogie bearing,the construction of the bogie bearing experimental platform and the bearing fault prefabrication are performed in accordance with the actual operating conditions of the vehicle.The flaking failure detection results of the inner ring,outer ring and rolling elements show that the proposed methods have certain reliability and practicality.In summary,the SR methods based on FIE can overcome the constraints of the sampling frequency ratio and detect the signal with any large-parameter frequency.In addition,the introduction of PFCPs can also realize the adjustment of the bistable stochastic resonance system intuitively.Therefore,the research results of this dissertation provide theoretical and technical foundation for further research and application of the SR methods.