摘要
针对环境温度对邻近铁路施工时铁路结构监测数据会产生较大的干扰,邻近施工所引起的轨道结构附加变形难以真实被反映的问题,提出了一种基于长短时记忆循环神经网络(LSTM)的轨道结构温度变形的预测方法,并根据某区间盾构下穿既有铁路线工程的实测数据对轨道结构变形结果进行了预测。通过与多层感知机(MLP)神经网络模型进行对比,研究表明:在监测数据精度有限且存在一定噪点的情况下,相比于基于多层感知机(MLP)神经网络的轨道变形预测方法,运用基于LSTM循环神经网络的预测方法预测效果更好,预测精度可达0.2 mm。
Aiming at the problem that the ambient temperature will cause great interference to the railway structure monitoring data during the construction of adjacent railways,and the additional deformation of the track structure caused by the adjacent construction is difficult to be truly reflected,a method is put forward for predicting the temperature deformation of the track structure based on a long-short-term memory loop neural network(LSTM),and the results of the track structure deformation are predicted based on the measured data of the shield tunnel undertaking the existing railway line in a certain section.By comparing with the multi-layer perceptron(MLP)neural network model,the study shows that when the accuracy of the monitoring data is limited and there is a certain amount of noise,compared with the track deformation prediction method based on the multi-layer perceptron(MLP)neural network,using the prediction method based on the LSTM recurrent neural network has a better prediction effect,and the prediction accuracy can reach 0.2 mm.
作者
姚逸行
刘建国
YAO Yihang;LIU Jianguo(The Key Laboratory of Road and Traffic Engineering,Ministry of Education,Tongji University,Shanghai 201804,China;Shanghai Key Laboratory of Rail Infrastructure Durability and System Safety,Tongji University,Shanghai 201804,China)
出处
《铁道建筑技术》
2021年第9期33-37,共5页
Railway Construction Technology
基金
上海市科委科研计划项目(19DZ1201004)。
关键词
LSTM
轨道结构
温度变形
预测方法
邻近铁路工程
LSTM
track structure
temperature deformation
prediction method
adjacent railway engineering
