铁道标准设计2025，Vol.69Issue(6)：46-54,9.DOI:10.13238/j.issn.1004-2954.202311040001
基于贝叶斯正则化的无缝线路轨温荷载预测模型研究
Research on Rail Temperature Prediction Model for Continuous Welded Rails Based on Bayesian Regularization
摘要
Abstract
The substantial temperature stress within the rails seriously threatens the operational safety of continuous welded rails.Therefore,it is necessary to accurately predict rail temperature and grasp its variation patterns in advance.This study focused on the correlation between air temperature and rail temperature.On-site monitoring tests were conducted in Fuyun,Xinjiang,with temperature data for both air and rail collected from January to October within a year to analyze their fluctuation patterns.A Bayesian regularized back propagation(BP)neural network was employed to establish a rail temperature prediction model,which was then used to predict the relationship between air and rail temperature variations at the rail head,rail waist,and other sections.The results showed that both air temperature and rail temperature at different sections of rails exhibited regular periodicity,and the difference in their extreme values did not fully align with traditional patterns.The measured results indicated that the difference between the maximum rail temperature and maximum air temperature ranged from 3 to 15℃,decreasing as the air temperature rose.The difference between the minimum air temperature and minimum rail temperature was approximately 1.08℃,with relatively stable fluctuation.The prediction results of the Bayesian regularized BP neural network model demonstrated that when predicting the minimum rail temperatures at different sections using the daily minimum air temperature,the model achieved the lowest average error of 0.311℃.When predicting the maximum rail temperatures at different sections based on the daily maximum temperature at the rail waist,the average error was the smallest,reaching 0.877℃.In practical engineering applications,priority should be given to predicting the minimum rail temperature using air temperature,and predicting the maximum rail temperature at different sections based on temperature of the rail waist.During the alternating hot and cold periods,timely monitoring of rail temperatures is essential to relieve the temperature stress.关键词
无缝线路/轨温预测模型/轨温监测试验/轨温变化规律/贝叶斯正则化/BP神经网络Key words
continuous welded rail/rail temperature prediction model/rail temperature monitoring test/rail temperature variation patterns/Bayesian regularization/BP neural network分类
交通运输引用本文复制引用
刘兴晨,肖杰灵,庄丽媛,景璞,余思昕..基于贝叶斯正则化的无缝线路轨温荷载预测模型研究[J].铁道标准设计,2025,69(6):46-54,9.基金项目
国家自然科学基金面上项目(52272441) （52272441）
国家科技部重大专项(2022YFB2602905) （2022YFB2602905）
四川省自然科学基金创新研究群体项目(2023NSFSC1975) （2023NSFSC1975）
山东铁投集团科技计划项目(TTKJ2021-04) （TTKJ2021-04）
