The Technical Studies Of Rail Profile Classification And Rail Wear Detection

With the development of China's rail traffic,transportation missions are getting heavier,while the train running density and speed of the train are constantly improving.In this case,the wear of the rails becomes increasingly serious,which is of great necessity to make an accurate and efficient detection of the rail wear in time.The traditional method by contact detection requires a lot of manual operation,existing problems of low detection efficiency and accuracy etc.However,the rail profile detection system based on CCD and structured light adopts optical non-contact measurement method,which has high detection accuracy.By combining with the railway inspection train,the detection efficiency can be greatly improved,and the comprehensive management of detection data can be realized as well,which is the development trend of rail wear detection.Based on the rail profile detection system,two major problems are studied: the rail profile classification and the rail wear detection.The significance of the rail profile classification mainly lies in the identification of line features,so as to provide the positioning information for the rail inspection data;while the rail wear detection is to realize the automatic calculation of the rail vertical and side wear.In the research of rail profile classification,the characteristics of each rail profile in the actual line data has been analyzed,with the proposal of the feature extraction algorithm of rail profile.Based on the profile data of rail,the pattern recognition method by Support Vector Domain Description(SVDD)algorithm has been adopted after that.Finally,through the experimental verification of the data,the characteristics of the algorithm and the influence of each parameter have been analyzed,with the performance of the classifier model improved by optimizing the algorithm model.In the study of the rail wear detection algorithm,the structure of the rail and the definition of wear has been first introduced,followed by the study of nonlinear least squares contour matching method based on the geometric characteristics of the rail waist.Due to the anomalous data of the measured rail profile,the method using object function by the sum of the distance instead of least squares has been proposed,which solves the optimal parameters through the Nelder-Mead simplex optimization algorithm and completes the matching of the measured contour with the standard contour.Simulation results suggest that this method has better matching accuracy than the least square method.At the end of this paper,some measured data are selected to test above methods,and the validity of the proposed methods is verified as well.