基于BP神经网络的地铁车站早龄期混凝土热学参数反演分析  

Inverse Analysis of Thermal Parameters of Early-age Concrete in Subway Stations Based on BP Neural Network

作　　者：曹玉新 寇帅 霍曼琳[2] 姜永涛 王国义 李宗奇 李金武 CAO Yuxin;KOU Shuai;HUO Manlin;JIANG Yongtao;WANG Guoyi;LI Zongqi;LI Jinwu(Sinohydro Railway Construction Co.,Ltd,Beijing 100060,China;School of Civil Engineering,Lanzhou Jiaotong University,Lanzhou,Gansu 730070,China;Chengdu Construction Investment Co.,Ltd.,Chengdu,Sichuan 610218,China)

机构地区：[1]中电建铁路建设投资集团有限公司,北京100060 [2]兰州交通大学土木工程学院,甘肃兰州730070 [3]中电建成都建设投资有限公司,四川成都610218

出　　处：《水利与建筑工程学报》2024年第5期145-152,205,共9页Journal of Water Resources and Architectural Engineering

基　　金：2021年度中国电建智慧轨道交通工程研究中心定向支助计划(DJ-PTZX-2021-02)。

摘　　要：针对地铁车站混凝土结构的复杂性和地下环境的特殊性,传统的室内试验方法难以准确测得实际工程中的混凝土温度参数。为提高早龄期混凝土温度场模拟的精度,通过利用BP神经网络对混凝土热学参数反演获得精确的数值。首先确定混凝土热学参数(导热系数、比热、胶凝材料放热量)的选取范围,采用正交设计生成25组样本进行温度场模拟,得到不同工况下的温度变化数据,利用温度数据训练BP神经网络,建立混凝土热学参数与温度变化之间的非线性映射关系并反演热学参数,最后利用多项现场实测温度对反演参数进行验证。结果表明:模拟数值与实测温度的误差较小。该方法不仅提高了模拟的精度,且相比于传统试验更加经济高效。In view of the complexity of the concrete structure of the subway station and the particularity of the underground environment,the traditional indoor test method is difficult to accurately measure the concrete temperature parameters in the actual project.In order to improve the accuracy of temperature field simulation of early age concrete,the accurate value of concrete thermal parameters is obtained by using BP neural network.Firstly,the selection range of thermal parameters(thermal conductivity,specific heat,heat release of cementitious material)of concrete is determined.Secondly,25 groups of samples are generated by orthogonal design to simulate the temperature field,and the temperature change data under different working conditions are obtained.Thirdly,the BP neural network is trained by using the temperature data,and the nonlinear mapping relationship between the thermal parameters of concrete and the temperature change is established and the thermal parameters are inverted.Finally,the inversion parameters are verified by a number of field measured temperatures.The results show that the error between the simulated value and the measured temperature is small.This method not only improves the accuracy of the simulation,but also is more economical and efficient than the traditional test.

关 键 词：地铁车站 大体积混凝土 温度场 BP神经网络 数值模拟 

分 类 号：TU528[建筑科学—建筑技术科学]