摘要
高速公路软土具有高压缩性、流变性和显著的时间效应,使得其位移发展呈现非线性,难以实时分区级别的特征提取,导致监测结果出现误差。为此,提出高速公路软土地基侧向位移实时监测算法。采用融合动态时间规整(Dynamic Time Warping,DTW)与地质约束的K-means动态分区方法,通过双驱动距离指标实现地基分区智能化划分。建立滑动窗口机制自适应更新分区;基于分区结果构建分区级多元线性回归模型,结合Elastic-Net算法提取侧向位移主导因子,为地基侧向位移监测模型提供差异化特征输入;构建基于遗忘递推最小二乘法(Forgotten Recursive Least Squares,FFRLS)的地基侧向位移动态监测模型,以特征输入为核心驱动变量,通过模型初始化、动态参数更新及误差反馈优化,实现地基侧向位移分区自适应的实时监测。实验结果表明:研究算法可剔除与地基位移无关的变量,筛选出主导因子,且该算法能逐点递推预测并同步输出准确监测值,K5+200测点侧向位移监测结果均方根误差(Root Mean Square Error,RSME)与平均绝对误差(Mean Absolute Error,MAE)仅为0.35、0.28 mm,K12+500测点侧向位移监测结果RMSE与MAE仅为0.42、0.34 mm。
Soft soil of highways has high compressibility,rheology and significant time-dependent effects,which makes its displacement development nonlinear and difficult to extract features at the real-time zonal level,resulting in errors in monitoring results.Therefore,a real-time monitoring algorithm for lateral displacement of soft soil foundation on expressways was proposed.The K-means dynamic partitioning method that integrates dynamic time warping(DTW)and geological constraints was adopted to achieve intelligent partitioning of foundation zones through dual driving distance indicators.A sliding window mechanism was established to adaptively update partitions.Based on the partition results,a partition level multiple linear regression model was constructed,and combined with the Elastic Net algorithm to extract the dominant factor of lateral displacement,providing differentiated feature inputs for the foundation lateral displacement monitoring model.A dynamic monitoring model was built for lateral displacement of foundations based on Forgotten Recursive Least Squares(FFRLS)method,with feature input as the core driving variable.Through model initialization,dynamic parameter updates,and error feedback optimization,real-time monitoring of adaptive zoning of lateral displacement of foundation was achieved.Experiments showed that the research algorithm can eliminate the variables irrelevant to the foundation displacement,screen out the dominant factors,and recursively predict point-by-point and synchronously output the accurate monitoring values.The RMSE and MAE of the lateral displacement monitoring of the K5+200 measurement point were only 0.35 mm and 0.28 mm,respectively.The RMSE and MAE at the K12+500 measurement point were only 0.42 mm and 0.34 mm,respectively.
作者
李仰印
王蕾
LI Yangyin;WANG Lei(Shandong Jiaotong University,Jinan 250023,China;Zhong Zi HuaKe Traffic Construction Technology Co.Ltd,Beijing 100089,China)
出处
《国外电子测量技术》
2025年第5期56-62,共7页
Foreign Electronic Measurement Technology
关键词
软土地基
动态时间规整
实时监测
侧向位移监测
soft soil foundation
dynamic time warping
real-time monitoring
lateral displacement monitoring
