Research On Intelligent Recognition Technology Of Heavy-Haul Train Couplers Fracture And Application

Heavy-haul train couplers are prone to unpredictable failure fractures under long-term high-load operating conditions,potentially threatening railway transport safety.Coupler fractures are important evidence for analyzing coupler fracture mechanisms and summarizing the causes of failure,which is of great significance for improving the coupler production process.The current manual inspection-based coupler fracture failure analysis method has many drawbacks such as poor timeliness and subjective factors,which is not conducive to the rapid and quantitative analysis of coupler fracture.Therefore,this thesis proposed an automatic and intelligent coupler fracture recognition method to achieve fast and accurate identification of different failure patterns of coupler fractures,which provides the basis and prerequisite for subsequent coupler failure analysis and mechanism deduction.To address the problems of subjective factors and low recognition efficiency in the current manual inspection method,a lightweight feature-reused network is proposed to achieve fast segmentation and recognition of brittle fracture regions on the coupler fracture surface.The experimental results show that the proposed method has achieved a mean intersection over union of 88.64% and a segmentation speed of 58 FPS,which has a strong real-time performance.To improve the global information capture capability of the convolutional neural network,a hybrid convolution and transformer network is proposed,which is aimed at the problem of low accuracy of fracture recognition due to the weak ability to capture long-local information in convolutional networks.Experiments show that the proposed method has achieved a mean intersection over union of 91.70% and a speed of 51 FPS.It effectively improves the recognition accuracy of brittle fracture regions of coupler fracture.To address the problem of low accuracy of convolutional and Transformer networks in composite failure form recognition,a fracture failure pattern segmentation network incorporating convolutional and local perceptron is proposed.Experiments show that the proposed method achieves a mean intersection over union of 79.10% for the composite failure form of the coupler,and effectively improves the accuracy of the composite fracture details.To address the shortcomings of low automation and complex operation process in the field of coupler fracture analysis,we develop a client-side fracture recognition platform and web-site platform in offline mode.In addition,the proposed methods are deployed offline to realize the automatic and intelligent recognition and analysis function of one-click detection.