铁道科学与工程学报2024，Vol.21Issue(2)：837-850,14.DOI:10.19713/j.cnki.43-1423/u.T20230484
基于BP神经网络的混凝土箱梁最大温度梯度预测
Prediction of concrete box-girder maximum temperature gradient based on BP neural network
摘要
Abstract
The concrete box-girder is subjected to the combined effects of various time-varying climatic parameters,such as solar radiation and atmospheric temperature fluctuation,which may lead to significant non-uniform temperature field in the bridge structure.Temperature gradients along the cross-section may lead to excessive temperature stresses and temperature deformations,thereby deteriorating the safety and durability of bridge structure.The purpose of this paper was to explore the influence mechanism of climatic parameters on the temperature field of concrete box-girder and propose a novel method for predicting the extreme temperature gradient in concrete box-girder.First,a numerical model was established to simulate the temperature field in concrete box-girder under insolation conditions.The long-term variation of concrete box girder temperature field in respect of different regions in China was simulated using meteorological data for more than 2 years as input conditions.The long-term variation trend of the cross-sectional maximum temperature gradient of concrete box girders was also analyzed.Then,the principal component analysis(PCA)was conducted to determine the input parameters for the prediction of maximum temperature gradient of concrete box girders.Finally,the BP neural network optimized by genetic algorithm was established to predict the vertical and transverse temperature gradients of concrete box girders.The results were compared with the maximum temperature gradients of concrete box girders in respect of different regions in China.The analytical results indicate that the BP neural network models have high accuracy for predicting the maximum temperature gradient in concrete box-girder,with the absolute average errors(AAE)less than 0.8℃,the root-mean-square errors(RMSE)less than 1.2℃,and the coefficient of determination(R2)greater than 0.9.Based on local meteorological conditions,BP neural network prediction model can be used to determine the maximum temperature gradient of concrete box girders in different regions of China.The developed prediction method in this paper can provide an efficient tool for calculating the maximum temperature gradient for the design and construction phases of the concrete box-girder.关键词
桥梁工程/日照温度作用/BP神经网络/混凝土箱梁/温度梯度Key words
bridge engineering/solar thermal action/BP neural network/concrete box girder/temperature gradient分类
交通工程引用本文复制引用
王凯,张勇,刘建磊,何旭辉,蔡陈之,黄石基..基于BP神经网络的混凝土箱梁最大温度梯度预测[J].铁道科学与工程学报,2024,21(2):837-850,14.基金项目
中国铁道科学研究院集团有限公司基金资助项目(2022YJ174) （2022YJ174）
