Adaptive Detection Method Of Resonance Frequency Band And Its Application In High-speed Train Axle Box Bearing Fault Diagnosis

As one of the key components of the running part of the high-speed train,the bearing of the axle box has been in service in a relatively harsh environment for a long time,which makes it a component that is prone to failure.Therefore,the condition monitoring and fault diagnosis of bearings are very important to ensure the safe operation of trains.Vibration response is highly sensitive to bearing fault characteristics,and the technology of vibration data information acquisition is becoming more and more perfect.This makes the technical means of bearing fault monitoring through vibration information widely used.However,how to extract bearing fault features from complex vibration information is still a very difficult task.Fast kurtogram is the most commonly used signal processing method for extracting bearing fault characteristics using spectral kurtosis.The fault type is determined by finding the optimal fault resonance frequency band and filtering the signal and performing envelope demodulation.However,in practice,the fast kurtogram still faces two main problems: one is that the fixed allocation rule of the signal frequency band is easy to make the diagnosed signal frequency band unable to cover the complete fault information,or to extract the failed components;another is that the kurtosis index is very sensitive to the shock component in the signal,which seriously affects the effective localization of the real resonance frequency band,so it is not suitable for the diagnosis of bearing composite faults.Aiming at the above problems,taking the axle box bearing as the research object,this thesis improves the defects existing in the fast kurtogram fault diagnosis technology.A method is formed with the use of spectrum trend to detect resonance frequency bands as the core technical framework,and a diagnosis method for compound faults is proposed.The thesis mainly includes the following research contents:(1)Aiming at the defects caused by the fixed division of fast kurtogram,a bearing fault diagnosis method based on adaptive kurtogram is proposed.Firstly,an adaptive spectrum trend division rule is formulated,and the kurtosis value of each frequency band is calculated.Then,the optimal center frequency and bandwidth for fault diagnosis are obtained by extracting the frequency band where the maximum kurtosis value is located.The effectiveness of the method is verified by simulation signal and measured data.(2)The correlation kurtosis index superior to the kurtosis index is used as the criterion for constructing the kurtogram.A new adaptive correlation kurtogram method is proposed by combining the adaptive frequency band division method with the robustness index.The method not only enhances its anti-interference ability,but also can complete the diagnosis of bearing compound faults.(3)Aiming at the problem that the correlation kurtosis index needs to be calculated by artificially given a priori period,an adaptive windowed correlation kurtogram method is proposed.The method utilizes the autocorrelation function for fully adaptive detection of a priori period.And the windowed correlation kurtosis is introduced for post-processing,which realizes the function of continuously isolating each failure mode and determining the number of failures.Finally,the superiority of the method is verified by the simulated bearing fault signal and the measured bearing fault data respectively.