Compound-foult Diagnosis Of Rolling Bearings In A High-speed Train Based On Improved Fast Kurtogram

The condition monitoring and fault diagnosis of critical components in a high-speed train is one of the most important technical to ensure the safe and stable operation of trains.Wheelset bearing is one of the key components of the bogie,it bears the weight and load from the car body,as well as the complex forces from the track.Therefore,it is of great significance for condition monitoring and fault diagnosis of wheelset bearings.Resonance demodulation technique is one of the most accurate and efficient methods in the fault diagnosis of rolling bearings.The basic principle of the resonance demodulation technique is to find the fault resonance frequency band,and then perform envelope demodulation on the signal of the resonance band,finally,the fault type is judged by comparing with the fault characteristic frequency.The most critical step in the resonance demodulation technique is to find the fault resonant frequency band,and the most widely used of which is the Fast Kurtogram.In this paper,the algorithm and frequency band division structure of Fast Kurtogram are reviewed firstly.To discuss the limitation of Fast Kurtogram,the compound-fault simulation signal of outer loop and rolling element is designed,the result is,only low-density impulsive series can be diagnosed usually.In order to solve the limitation of the Fast Kurtogram,this paper studies five kinds of measurement index.It is found that,kurtosis,spectral kurtosis and sparsity index are sensitive to low-density impulsive series,moreover,envelope spectrum kurtosis and crest factor are sensitive to high-density impulsive series.According to this,the envelope spectrum kurtosis is selected to replace the spectral kurtosis in Fast Kurtogram,the Fast ESKTgram is obtained,and the compound-fault diagnosis method based on Fast Kurtogram and Fast ESKTgram is proposed.Next,the feasibility of this method is verified by the compound-fault simulation signal mentioned above.However,the method of combine Fast kurtogram and Fast ESKTgram can only diagnose two faults at most.In order to diagnose multiple faults of wheelset bearings,this paper proposes correlation kurtosis diagram based on empirical wavelet decomposition,named CKgram.Next,The influence of date length,shift order and shift length on correlation kurtosis are discussed through simulation signal.Then,the compound-fault diagnose method bases on CKgram is proposed,and the feasibility of this method is verified by the simulation signal.Finally,the three methods mention above are verified by bench tests.