基于长短时记忆网络的深基坑变形安全风险预警  

Safety Risk Warning of Deep Foundation Pit Deformation Based on LSTM

作　　者：夏天 成诚 庞奇志[1] Xia Tian;Cheng Cheng;Pang Qizhi(Faculty of Engineering,China University of Geosciences,Wuhan 430074,China)

机构地区：[1]中国地质大学工程学院,湖北武汉430074

出　　处：《地球科学》2023年第10期3925-3931,共7页Earth Science

摘　　要：为了预防深基坑施工安全事故,提出了一套基于监测数据的风险预警标准,建立了基于长短时记忆网络(Long Short-Term Memory,LSTM)的深基坑变形安全风险预警模型.依托实际深基坑工程项目,将风险预警模型应用其中,对基坑各监测项目的变形量进行短期的预测,预测数据与实际数据最大误差为5.04%,最小误差为0.04%,平均相对误差为2.41%,证明该模型的预测效果良好.表明基于LSTM的深基坑变形安全风险预警模型在基坑变形预测方面有着良好的精确性和优越性,可以为基坑工程的安全性判断与风险管控提供可靠的保障.In order to prevent deep foundation pit construction safety accidents,a set of risk warning standards based on monitoring data is proposed,and a deep foundation pit deformation safety risk warning model based on long short-term memory(LSTM)was established.Relying on the actual deep foundation pit engineering project,the risk warning model is applied to it to make shortterm predictions of the deformation of each monitoring item of the foundation pit.The maximum error between the predicted data and the actual data is 5.04%,the minimum error is 0.04%,and the average relative error is 2.41%,which proves that the prediction effect of the model is good.It shows that the LSTM-based deep foundation pit deformation safety risk early warning model has good accuracy and superiority in the prediction of foundation pit deformation,and can provide a reliable guarantee for the safety judgment and risk management of foundation pit engineering.

关 键 词：深基坑 基坑安全 变形量预测 长短时记忆 风险预警 安全工程 工程地质学 

分 类 号：P642[天文地球—工程地质学]