Research On Wheel Pair Dimension Detection System Based On Machine Vision

In recent years,with the rapid development of social economy,the passenger flow of highspeed rail and urban rail transit has risen sharply.Any slight fault may pose a serious threat to the safety of trains.The wheelset is an important part of the train.It directly contacts the track,bogie and other units,carries the weight of the car body,and is very easy to wear during operation.The good state of the wheelset is an important condition to ensure the safe and stable operation of the train.However,due to the unique shape of the wheel flange,the detection,repair and maintenance work is complicated and time-consuming.Therefore,studying the dynamic online detection of wheelset size and predicting the loss of dimension parameters can help relevant departments to formulate efficient maintenance strategies,ensure the safe use of wheelsets,and optimize the safety,comfort and economic performance of trains during operation.Based on machine vision,this study researches the dynamic detection system of train wheelset size.Firstly,the basic knowledge of wheelset is expounded,and the structure of the system is introduced from two aspects of software and hardware according to the specific functional requirements.Secondly,the basic principle of binocular structured light measurement and the calibration method theory of structured light camera are explained,which lays a foundation for wheelset size detection.Then,the contour fusion algorithm is studied,the Canny algorithm is improved,the image is preprocessed,and the denoising and edge detection effects are optimized.In the frequency domain,the point cloud is registered to obtain the rotation parameters,and the least square method based on the geometric constraint condition is used to fit the arc part at the top of the rim to obtain the translation parameters.Finally,the matching of the profile is completed and merged into a complete wheelset curve.By calculating the coordinates of the feature points,the size detection is realized.In addition,on the basis of wheelset size detection,this study proposes a wheelset size prediction model based on NPSOLMBP algorithm,which combines niche idea with particle swarm optimization algorithm to optimize the initial weights and thresholds of traditional BP neural network,and further optimizes the network with LM algorithm to improve the convergence speed.Taking the wheel diameter and flange height value as the prediction object,the prediction performance is tested by the measured data of the wheel size of the urban rail train.At the same time,compared with the prediction results of BP neural network and LMBP neural network,the superiority and rationality of the algorithm are verified.The algorithm is applied to the dynamic detection of subway wheelsets,and the prediction is completed based on historical data to verify the feasibility and accuracy of the algorithm.Experiments show that the detection system based on this algorithm can meet the requirements and standards of the maintenance department,which is conducive to the maintenance and repair of train wheelsets.