基于Bayes和LSTM神经网络模型的基坑变形值预测研究  

Research on Prediction of Deformation Value of the Foundation Pit Based On Bayesian and LSTM Neural Network Models

作　　者：曹玉江 Cao Yujiang(China Railway 18th Bureau Group Municipal Engineering Co.,Lid.,Tianjin 300222,China)

机构地区：[1]中铁十八局集团市政工程有限公司,天津300222

出　　处：《市政技术》2024年第11期119-126,共8页Journal of Municipal Technology

摘　　要：为了提高基坑变形预测的准确性和可靠性,提出了一种基于贝叶斯方法(Bayes)和长短期记忆(Long Short-Term Memory,LSTM)神经网络的复合模型,并结合杭州市文一西路改造工程现场监测数据,比较了Bayes-LSTM模型与其他预测模型在大跨度基坑上方的地表沉降与水平位移数据预测误差。研究结果表明:与LSTM模型和支持向量机(SVM)模型相比,Bayes-LSTM模型对基坑上方地表沉降的预测精度分别提高了1.0和1.26,证明了Bayes-LSTM模型在地表沉降预测方面表现出较高的预测精度和泛化能力。该研究成果可为大跨度基坑施工安全管理提供决策与支持。In order to improve the accuracy and reliability of pit deformation prediction,a composite model was proposed based on Bayesian method(Bayes) and long short-term memory(LSTM) neural network in this paper.Combined with the site monitoring data of Wenyi West Road renovation project,the prediction error of surface settlement and horizontal displacement data above the large span pit was compared by Bayes-LSTM model with other prediction models.The results show that the prediction accuracy has been improved by Bayes-LSTM model by 1.0 and 1.26 respectively compared with the LSTM model and the SVM(support vector machine) model,which is proved to have high prediction accuracy and generalization ability in the prediction of surface settlement.The study provides decision support for the safety management of large-span foundation pit construction.

关 键 词：基坑沉降 贝叶斯网络 LSTM神经网络 预测模型 

分 类 号：TU941[建筑科学—建筑技术科学] TU433