Design Of Wheelset Comprehensive Parameter Detection Device Based On Line Structured Light

With the rapid development of society,rail transit has become an important way for people to travel and freight logistics,and the long-term contact between wheelsets and rails will cause wear and damage to the tread,which affects the stability and safety of vehicle operation.Therefore,it is necessary to regularly check whether the various parameters of the wheelset are qualified to ensure the safety of train operation.At present,the detection of wheelset parameters in my country’s railway sections is still based on manual measurement,which makes operators more labor-intensive and inefficient.After long hours of work,human factors also bring large errors,resulting in poor measurement accuracy and reliability.Based on static measurement,this paper designs an automatic detection system for wheelset comprehensive parameters.It uses line structured light measurement and image processing technology to realize automatic measurement of wheelset size parameters.At the same time,it detects tread defects based on the principle of deep learning target detection to improve the accuracy of wheel set parameter detection.The main contents of this article are as follows:(1)Facing the demand for high-precision wheelset detection,a scheme and overall framework of a wheelset tread profile measurement device based on line structured light is proposed.Aiming at the complex contour of the wheelset tread,the linear structured light measurement combined with the positioning block is used to locate the inner side of the wheel set,which improves the measurement efficiency.The overall mechanical structure and motion control system of the detection device are designed,and the key components are calculated and selected.Before the measurement,the line structured light method and the designed contact measurement mechanism are used to measure the same cross-section of the wheel shape to ensure the reliability of the system measurement.(2)The principle of direct and oblique structured light measurement is explained.By comparing the advantages and disadvantages of the two measurement methods,combined with the actual measurement situation,the direct laser triangulation method is selected as the measurement scheme of the system.According to the conversion relationship between the coordinate systems in the measurement process,a mathematical model of line structured light vision measurement is established,and the internal and external parameters of the camera and the structured light plane are calibrated using Zhang plane method and the principle of constant cross-ratio.(3)For the collected line structured light image,the RGB color channel is separated,and the extracted R color channel grayscale image is preprocessed by median filtering,and the gray center of gravity method is used to achieve precise extraction of the center line of the structured light strip.Aiming at the unevenness and non-smooth in the process of centerline extraction,a polynomial curve fitting method based on automatic segmentation is proposed to smooth the data points.The hypercube random sampling method is used for obtaining the initial segmentation points of the extracted center coordinate points of the light strip,the objective function with the smallest sum of square error of the global error is established,and the optimal segmentation point position is determined through the solution of the L-BFGS optimization algorithm.The segmented intervals are respectively fitted with polynomials.According to the average absolute error of the fitting curve,it is judged whether the second segmented fitting is required.Using this method,the extracted centerline can truly express the contour of the tread,thereby improving the measurement accuracy.(4)A tread defect detection method based on the YOLOv4 deep learning target detection framework is used.Through image translation,rotation,zooming,noise addition and other enhancement techniques,the collected tread defect images were sampled and expanded and a data set was produced,and sent to the built YOLOv4 detection network for training.Finally,and finally the mean average precision(m AP)of defect recognition in the test set up to 98.7%,which realizes the accurate identification of tread defects and avoids the influence of tread defects on the parameter measurement results.(5)Based on the Qt application program development framework,the software part of the detection device is developed in a modular manner,and functions such as system calibration,image processing,wheel set parameter measurement,result display and storage are integrated,and the graphical interactive interface is designed.An experimental platform for parameter measurement based on line structured light was built to measure the wheel model,and the measurement results were compared and analyzed with manual measurement results.The experimental results show that the measurement error of the rim height and rim thickness of the detection method proposed in this paper is less than 0.1mm,and the measurement error of the rim width is less than 0.3mm,which meets the actual measurement requirements and can be used for subsequent wheelset status judgment and repair.Provide data support for size determination and wear prediction.