Research On Key Techniques For Detection Of Micro Scratches On High Speed Rail Axle

Wheelsets are an important part of high-speed rail EMUs.In high-speed EMUs,in order to ensure the reliability of operation,the wheelsets need to be returned to the factory for inspection every 1.2 million kilometers.However,during the repair and disassembly process,they are subject to non-compliance.The uneven force caused by the action of a single axial force will leave tiny scratches on the metal surface of the axle at the connection with the wheelset.Such tiny scratches will cause oil leakage at the wheel and axle meshing position when oil is injected,which will cause damage.The pressing of wheel sets and high-speed trains in high-speed operation cause safety hazards.If the depth of the scratch on the metal surface exceeds 0.5mm,the axle of the moving wheel pair should be ground before pressing to reduce the degree of damage on the metal surface.At present,the detection of such small scratches by enterprises mostly relies on manual visual inspection or empirical judgment,and it is impossible to accurately and effectively obtain the scratch data detection report.As a result,the safety parameters of the originally qualified axles after the overhaul are exceeded,and they are directly scrapped.Serious economic loss;at the same time,it greatly affects its maintenance efficiency,greatly reduces the quality of wheel repair,and in severe cases,it also affects the dynamic balance of the train during high-speed operation and the safety performance of the train.Therefore,this article focuses on the detection of small scratches on the surface of high-speed rail axles as follows:(1)Aiming at the scratches on the surface of the axle during the disassembly process of the high-speed rail wheel set for maintenance,a small scratch detection scheme on the surface of the high-speed rail axle based on the line structured light active vision is proposed,and the corresponding mechanism design and design of the detection system are carried out according to the scheme.Electrical design,and finally complete the selection of various hardware parts in the line structured light three-dimensional measurement system according to the expected technical indicators.(2)Research the key technologies of the line structured light three-dimensional measurement system.The improved BAS algorithm is used to optimize the initial calibration parameters of the camera until the calibration accuracy requirements are met.In order to improve the calibration accuracy of the line laser,a direct calibration method based on a special-shaped calibration block is proposed,which establishes the mapping relationship between the image pixel coordinate system of the feature calibration point on the calibration block and the world coordinate system,and then any point on the laser line The space world coordinate system of can be obtained by interpolation based on this mapping relationship.In order to accurately extract the characteristic calibration points on the center line of the laser line,an extraction method based on the combination of the DBSCAN clustering optimization algorithm and the improved PSO shortest path optimization algorithm is proposed,and the results are compared with the results obtained by the gray barycentric method.Analysis has achieved a good extraction effect.The speed calibration method is used to complete the speed calibration of the mobile detection mechanism,and the detection and repair process of the metal reflective area on the surface of the high-speed rail axle is realized through the adaptive multi-level threshold and the progressive dynamic template.(3)Research on scratch measurement algorithm for processing point cloud data.Use the RANSAC algorithm to segment the point cloud data obtained from the scan,and use the method of combining straight-pass filtering and statistical filtering to filter the point cloud data,which reduces the amount of calculation of the later scratch detection algorithm,and uses the Delaunay quantization processing algorithm to make the point cloud The data is more in line with the actual scratch morphology.Finally,through the depth map mapping of the point cloud data,the edge extraction of the scratches from the disparity map realizes the length and width measurement of the scratches,and the measurement of the length and width of the scratches is achieved by performing a cubic B-spline curve fitting on the single frame point cloud data obtained by scanning.Measurement of the depth of the scratches.(4)Experimental platform construction,software system design,experimental result analysis and error compensation model establishment.The software system design is divided into three modules: calibration module,data acquisition module,and scratch detection module.The calibration module gives the calibration results of each part of the camera,the line laser and the movement detection mechanism,and carries out the accuracy verification analysis.The data acquisition module and the scratch detection module respectively carry out the corresponding software interface design.Then use the cylindrical axis for experimental measurement,compare the measurement results with the three-coordinate measurement results,and meet the measurement accuracy requirements of0.05 mm in the expected technical indicators,and use the improved GA-BP algorithm to establish an error compensation model to make the measurement results more accurate.Close to the standard value.Finally,the built-up line structured light three-dimensional measurement system was used to realize the detection of small scratches on the surface of the high-speed rail axle.