基于ALSTM模型的基坑周边建筑物沉降预测  

Prediction of buildings settlement around foundation pit based on ALSTM model

作　　者：余凯 吴来根 Yu Kai;Wu Laigen(Fujan Geotechnical Engineering Investigation and Research Institute Co.Ltd,Fuzhou 350108,China)

机构地区：[1]福建岩土工程勘察研究院有限公司,福州350108

出　　处：《工程勘察》2023年第3期63-67,72,共6页Geotechnical Investigation & Surveying

摘　　要：中短期基坑沉降监测序列具有非线性和数据量小的特点,导致常规预测模型很难获取准确的预测结果。针对传统模型未考虑到历史时刻沉降情况对未来沉降量具有不同影响的缺点,本文采用ALSTM(Attention LSTM)预测模型,并以某大厦基坑工程变形监测的数据为例进行验证。实验结果表明,相比LSTM、支持向量回归和BP神经网络模型,ALSTM模型能够取得更加准确的预测结果,适用于短期和中短期两种情况下的沉降变形预测。The monitoring sequence of medium and short-term foundation pit settlement has the characteristics of non-linearity and small amount of data, which makes it difficult for conventional prediction models to obtain accurate prediction results. Aiming at the disadvantage that the traditional model does not take into account the fact that the settlement at the historical moment has different effects on the future settlement, this paper adopts the ALSTM(Attention LSTM) prediction model, and takes the deformation monitoring data of a building foundation pit as an example to verify;the experimental results show that, compared with LSTM, support vector regression and BP neural network models, the ALSTM model can achieve more accurate prediction results, and is suitable for the prediction of settlement deformation in both short-term and medium-short-term conditions.

关 键 词：基坑监测 沉降预测 LSTM 注意力机制 

分 类 号：TU196[建筑科学—建筑理论]