Research On The Online Detection System Of Rail Surface Defect Before Welding

As an important part of railway infrastructure,the quality of rails directly determines the safety of railway transportation.As the current level of manufacturing technology is getting higher and higher,the number of safety accidents caused by rail internal defects is getting fewer and fewer,and the rail fracture probability caused by rail surface defects is gradually increasing.Therefore,rail surface defect detection is an important part of rail detection.This paper aims at the problems of low detection efficiency and detection accuracy of manual visual inspection methods,as well as the two-dimensional defect detection based on machine vision strict with the detection environment,and the inability to obtain depth information of defects.So machine vision-based three-dimensional defect detection methods were using for rail surface defects.Tests were conducted to develop an on-line inspection system for rail surface defects with high detection accuracy.This paper mainly studies the design and implementation of rail surface defect detection system,surface data preprocessing and defect detection algorithm,as follows:(1)The current research on the detection technology of rail surface defects based on machine vision has problems of low accuracy and incomplete defect information.A detection system of rail surface defects based on linear structured light sensors was developed.The sensor installation method and foundation,rail support guide mechanism were designed to achieve a real-time online,unmanned,higher detection efficiency and accuracy of on-line detection system.(2)Analyze the external environment and the influence of the vibration during the rail movement on the detection results.The paper proposed the method of the speed limit filter to reduce the noise,analyzed the composition features of the rail profile curve,studied the feature points and proposed the alignment method of the rail profile data based on the feature points.A method based on statistical window-based moving indirect adjustment method was proposed to search the tangent point of the tangent arc with large difference between the radii.The feature points were searched and the datas of each frame were aligned.And the accuracy of the alignment algorithm was verified to ensure that the initial value that met the accuracy requirements could be provided for subsequent defect detection.(3)According to the characteristics of rail surface defects,a method of detecting single-profile data and combining multiple contours for defect identification was proposed.According to the composition characteristics of the rail profile curve,after detailed analysis and multiple tests,the dynamic interval fitting curve detection algorithm based on the quadratic curve fitting model was proposed to detect the rail surface defects.The detection results of each frame of data were compare,then using multiple contour combinations to identify defects.The effect of rail waist characters and rail surface oxide scale on the detection results of the algorithm was examined.The results show that the algorithm is simple,efficient,and stable.The detection system was developed under the Visual Studio 2015 platform.The Geomagic Studio software was used to simulate the dent defects on the rail surface and the accuracy of the system was tested.Experiments show that the detection method has high accuracy.At present,the system has been applied to the welding base of Wuhu north and has achieved good results.