基于融合算法的基坑支护桩变形预测及影响因素分析  

Deformation Prediction and Influencing Factor Analysis of Foundation Pit Support Piles Based on Fusion Algorithm

作　　者：夏水龙 XIA Shuilong(China National Building Materials(Zhejiang)Survey and Design Co.,Ltd.)

机构地区：[1]中建材(浙江)勘测设计有限公司

出　　处：《广东建材》2025年第3期105-108,共4页Guangdong Building Materials

摘　　要：城市化进程加快使基坑工程面临复杂状况,准确预测基坑支护桩变形及分析影响因素极为关键。本文基于LSTM、XGBoost和Stacking集成算法对基坑支护桩变形进行预测,在桩径0.5 m时,LSTM算法预测支护桩水平与竖向位移和实测值最大误差分别为3.8mm和2.3mm,XGBoost算法对应误差为2.1mm和2.3mm,Stacking算法仅为0.7mm和0.7mm;桩间距1.5m时也有类似规律。可见Stacking算法在不同桩径和桩间距位移预测中更贴近实测值。此外,桩径从0.5 m增大到0.8m时,水平和竖向最大变形量分别减小39.4%与37.6%,桩间距增加时,水平与竖向位移分别增加73.4%与37.4%,这些规律为工程设计、施工和管理提供了科学依据。The acceleration of urbanization has made foundation pit engineering face complex situations,and accurately predicting the deformation of foundation pit support piles and analyzing the influencing factors are extremely crucial.This article is based on the integrated algorithm of LSTM,XGBoost,and Stacking to predict the deformation of foundation pit support piles.When the pile diameter is 0.5 m,the LSTM algorithm predicts the maximum errors of horizontal and vertical displacement of the support piles and the measured values,which are 3.8 mm and 2.3 mm,respectively.The XGBoost algorithm corresponds to errors of 2.1 mm and 2.3 mm,while the Stacking algorithm only predicts errors of 0.7 mm and 0.7 mm;There is a similar pattern when the distance between piles is 1.5m.It can be seen that the Stacking algorithm is closer to the measured values in predicting displacement with different pile diameters and pile spacing.In addition,when the pile diameter increases from 0.5 m to 0.8 m,the maximum horizontal and vertical deformations decrease by 39.4%and 37.6%,respectively.When the pile spacing increases,the horizontal and vertical displacements increase by 73.4%and 37.4%,respectively.These laws provide scientific basis for engineering design,construction,and management.

关 键 词：基坑支护桩 变形预测 LSTM算法 XGBoost算法 Stacking算法 

分 类 号：TU7[建筑科学—建筑技术科学]