基于LSTM模型的高原桥梁温度场预测研究  

Research on Temperature Field Prediction of Plateau Bridges Based on LSTM Model

作　　者：张云必 ZHANG Yunbi(JSTI Group Co.,Ltd.,Nanjing 210019,China)

机构地区：[1]苏交科集团股份有限公司,南京210019

出　　处：《现代交通技术》2023年第6期67-71,89,共6页Modern Transportation Technology

摘　　要：基于LSTM(long short-term memory,长短期记忆)神经网络深度学习模型,通过自适应矩估计算法建立桥位处的环境温度、地表太阳辐射量(包括直接辐射和散射辐射)、平均风速和风向等环境参数与箱梁温度场之间的映射关系,并通过贝叶斯算法优化LSTM模型中的超参数,使均方差降至最低,从而获得箱梁温度场预测模型。预测值与实测值的对比分析表明,建立的LSTM神经网络深度学习模型能够根据环境参数准确预测高原环境下连续刚构桥箱梁的温度场,且模型具有较强的泛化能力。In the research,a prediction model for the box girder temperature field is established using an LSTM neural network deep learning model.The model is built based on an adaptive matrix estimation algorithm,which maps environmental parameters,such as environmental temperature at the bridge location,ground surface solar radiation(including direct radiation and scattered radiation),average wind speed,and wind direction,to the box girder temperature field.Bayesian algorithm is used to optimize the hyperparameters of the LSTM model,reducing the mean squared error to a minimum,thus obtaining the prediction model for the box girder temperature field.Comparative analysis between predicted values and measured values shows that the established LSTM neural network deep learning model can accurately predict the box girder temperature field for continuous rigid frame bridges in the plateau environment based on environmental parameters,and the model exhibits strong generalization capabilities.

关 键 词：高原环境 连续刚构桥 LSTM模型 贝叶斯算法 箱梁温度场 

分 类 号：U441+.5[建筑科学—桥梁与隧道工程] U448.23[交通运输工程—道路与铁道工程]