Detection And Identification Of Rail Surface Defects

Surface defects of rails can endanger driving safety.In order to ensure the safety of rails in service,this paper studies the non-destructive testing of rail surface defects and the depth recognition of defects.This paper is based on laser ultrasonic technology to detect steel rails.Taking 60 kg steel rail as the research object,through numerical simulation of the propagation process of laser ultrasonic waves encountering defects,observing the effect of defects on ultrasonic waves,and deriving the arrival time of different modes of waves according to the formula,and realizing the detection of defects Positioning.The effects of different receiving points and different depth defects on the surface of the rail are studied.From the time-frequency domain,the relationship between the change of the receiving point and the time and amplitude and the relationship between the change of depth and the reflected wave and transmitted wave are analyzed.And use the laser ultrasonic experiment to verify the simulation results.In order to explore the depth of defects in depth,a frequency domain analysis was performed by combining the Fast Fourier Transform and Hilbert Huang Transform to extract a lot of information about the depth of defects.The nuclear principal component analysis algorithm is used to mine the characteristics of the laser ultrasonic signal in the time domain and the frequency domain,and the number of nuclear principal components is determined according to the cumulative contribution rate to achieve data dimensionality reduction,and multiple principal components that meet the requirements are used as the classifier model.enter.A defect classification prediction system based on least squares support vector machine algorithm is proposed,which improves the speed of system classification.Established a classification model based on KPCA-LSSVM and realized the recognition of the depth of rail surface defects.Through the evaluation index of the classification system,the index values before and after the optimization of the system were compared,and the generalization ability of the system was improved.The accuracy rate of the system has been increased from 86.22% to 94.875%,and the recall rate has been increased from 86% to 95.405%.The false positive rate has been reduced,and the accuracy rate has reached 96.5%.