| With the development of high-speed rail technology,it was important for railway safety to develop rail fault diagnosis technology.In this paper,a new method of rail fault diagnosis was studied and a mathematical model for fault diagnosis was proposed.What’s more,the feature vector of rail vibration signals was extracted.According to the inherent characteristics of the signals,the rail faults could be clearly observed by comparing the rail signals in a healthy state.In this sinusoidal model,the rail vibration signal was first to preprocess,low-frequency noise was removed using time-domain synchronous averaging to ensure smooth signal trends.Then the Fourier triangle expansion was used to expand the vibration signal and the signal was optimized according to the developed and actual rail signal characteristics.In the next,the least square strip learning method was used to modify the process of the vibration signal.The expanded sine factor was regularized to ensure the accuracy of the obtained feature vector.Finally,the magnitude and phase feature vectors were obtained by using the fitting error.In addition,the phase and amplitude modulation factors were restored to the amplitude and phase vibration of each order of the vibration signal.At the same time,the pulse vibration signal caused by the fault was also extracted and the degree of the fault was predicted to some certain extent.Through a large number of field tests in different types of rail faults,this sinusoidal model had caused better results and achieved high accuracy,which was of great significance for the safety detection of rails. |
