Research On The Identification Of Acoustic Source Motion Parameters Based On Time-Frequency Analysis Of Wayside Acoustic Signal

Train wheel bearing is the main component of train running gear,and its health is directly related to whether the train can run safely and stably.Therefore,it's necessary to monitor the running state of the train wheelset bearing.Due to the high speed of the train,the wayside acoustic signal often have serious Doppler distortion,multiband aliasing and strong background noise.Because of the influence of Doppler distortion,the instantaneous frequency of signal will change with time.For this kind of nonlinear time-varying signal,the time-frequency analysis method usually has a good effect.In this thesis,the artificial fault of the wheelset bearing is identified by the wayside acoustic signal of the train.Taking the sound source as the research object,the characteristics of the time frequency distortion of the wayside acoustic signal are introduced.Based on time frequency analysis,this paper combines image processing,traversal optimization and other signal processing methods to identify the motion parameters of wayside acoustic signal.The distortion signal is corrected by the identified motion parameters,and finally the fault information is extracted from the acoustic signals.Three kinds of algorithms for identifying the motion parameters of the wayside acoustic signals is the main work of this thesis.The first is the motion parameter identification algorithm based on the time frequency ridge extraction of the signal.This method use mean filtering to the time-frequency matrix that obtained by Short Time Fourier Transform(STFT),then,the threshold processing is used of the regions where some signal energy is gathered after filtering.Finally,the ergodic peak search algorithm is used to extract the time frequency ridges of the target ridge area.The motion parameters of the sound source can be fitting according to the motion model of the sound source.The second algorithm is motion parameter identification based on variable scale iterative fitting.The algorithm is designed to reduce the dependence of the algorithm on human intervention.The frequency distorted signal can be filtered by variable frequency band filter based on Doppler window structure.Through the sound source motion parameter of each recognition to construct new Doppler window for signal processing,time-frequency ridge extraction and parameter fitting iteration.Finally,get the optimal parameters of the sound source movement.The third algorithm is adaptive acoustic source motion parameter identification.The main innovation of this algorithm is use Doppler window in time-frequency matrix to select local signal traversal.The source movement parameters used to construct the Doppler window is closer to the true value,the local signal energy of the Doppler window value is bigger.According to this relationship,we can preset the motion parameter set of the sound source to traverse the whole time and frequency domain with the idea of matching pursuit,then,the optimal motion parameters of the sound source can be matched.The research of experimental signals shown that the three methods proposed in this thesis can accurately identify the motion parameters of sound sources,According to the research results,it is proved that time-frequency processing is an effective way to analyze nonlinear time-varying signals.This work is a beneficial explore to the condition monitoring and fault diagnosis of wayside acoustic signals.