军事知识服务平台
基于DBO-LSTM模型的桥梁环境温度及其效应预测方法研究
A Hybrid DBO-LSTM Model for Predicting Bridge Ambient Temperature and Its Thermal Response
摘要 针对传统健康监测系统数据处理复杂及有限元方法在桥梁环境温度效应预测中存在的计算量大、实时性差等问题,提出一种基于蜣螂优化算法(DBO)优化长短期记忆网络(LSTM)超参数的混合模型(DBO-LSTM),用于桥梁环境温度及温度效应的预测.该方法通过DBO算法对LSTM的隐藏层节点数、初始学习率及L2正则化系数三项关键超参数进行全局优化,提升了模型在长序列非线性特征学习过程中的预测精度和稳定性.基于成绵立交主线桥为期一年的温度与温度效应监测数据,对环境温度、伸缩缝位移、主梁梁底应变和墩顶偏位等多项指标进行预测实验,结果显示所建模型在各类预测项中均表现出良好的拟合能力和趋势捕捉能力.决定系数R²最低为0.966,最高达0.995,平均绝对百分比误差(MAPE)在0.02至0.38之间,预测值与实测值在数值及趋势上高度一致.
Abstract:
To address the limitations of traditional bridge health monitoring systems and finite element methods,such as heavy computational demands and poor real-time performance,a hybrid prediction model based on Dung Beetle Optimization and Long Short-Term Memory networks(DBO-LSTM)is proposed for forecasting bridge ambient temperature and temperature-induced effects.The model utilizes the DBO algorithm to optimize three critical hyperparameters of the LSTM network:the number of hidden units,initial learning rate,and L2 regularization coefficient.This optimization enhances the model's ability to learn complex nonlinear dependencies in long-term monitoring data.Using one year of real-time monitoring data from the Chengmian Interchange mainline bridge,the model was applied to predict multiple indicators,including ambient temperature,expansion joint displacement,bottom beam strain,and pier top displacement.The results demonstrate high predictive performance,with coefficients of determination(R²)ranging from 0.966 to 0.995 and mean absolute percentage errors(MAPE)between 0.02 and 0.38.The predicted values closely match the actual measurements in both magnitude and trend.
刊 名 结构工程师 ISTIC
年,卷(期): 2026, 42(4)
分类号: U446.2
机标分类号: TP391.1 TP183 U491.14
在线出版日期: 2026年9月21日
基金项目: HT-GLQ2024-0055:四川交投科研联合项目,