Research On Rail Crack Identification And Quantitative Estimation Methods

Steel rails,as one of the ferromagnetic materials,are the basis of railway transportation.Therefore,it is of vital importance for ensuring the normal railway operation to non-destructively test and evaluate steel rails,timely and effectively.Based on the actual demand in engineering,the author studies the key technologies in applying Non-Destructive Testing(NDT)to rail cracks.The main points are as follows:For part one,the author first analyses the Metal Magnetic Memory(MMM)method,according to which the author puts forward the characteristics of the MMM signals of the rail cracks.Second,In view of the noise and trend terms of the measured signal,the author combines the Least Mean Square(LMS)algorithm with the least square method to filter MMM signals.Moreover,the improved STA /LTA automatic detection algorithm is used to detect and intercept the rail crack signals.At last,the rail crack database is constructed.For part two,aiming at the problems in MMM about multi-directional rail cracks,such as information redundancy,the author proposed Relief F-GRA feature selection algorithm.It integrates the relationship between features and crack parameters with that among the features themselves.It also optimizes the characteristics of the MMM signals,and thereby improves the accuracy and speed in rail cracks identification and parameter estimation.Aiming at the problem of identification of rail cracks,the author use the Support Vector Machine(SVM)method to identify the types of rail cracks.Experimental results show that the rail crack qualitative identification model,based on the Relief FGRA feature selection algorithm,has higher accuracy and faster recognition speed.For part three,aiming at the problem of parameter estimation of rail cracks,the author designs two parameter estimation algorithms,which are based on BP neural network and ALS algorithm respectively.Through these algorithms,effective estimation of rail crack parameters can be realized.The experimental results show that the author’s research has high reference value for rail crack identification and its parameter estimation.