公路隧道多断面围岩变形的动态时序预测  

Time-Series Prediction of Multi-Section Surrounding Rock Deformation in Highway Tunnels

作　　者：谷玉梁 蒋鸿彬 葛圣泰 周庆琨 Gu Yuliang;Jiang Hongbin;Ge Shengtai;Zhou Qingkun(China First Highway Engineering Co.,Ltd.,Fuzhou 350007,China;College of Civil Engineering,Zhejiang University of Technology,Hangzhou 310023,China;Command Headquarters for the Reconstruction Project of National Highway 235 Taishun Section,Wenzhou 325500,Zhejiang,China)

机构地区：[1]中交一公局集团有限公司,福州350007 [2]浙江工业大学土木工程学院,杭州310023 [3]235国道泰顺段改建工程指挥部,浙江温州325500

出　　处：《科技通报》2025年第4期94-100,122,共8页Bulletin of Science and Technology

摘　　要：为克服公路隧道开挖过程中传统预警方法的局限性,本文提出一种基于双向长短期记忆网络(bidirectional long short-term memory,BiLSTM)的隧道围岩变形动态时序更新模型。随着隧道掘进的推进,模型通过不断结合新断面数据,动态调整权重,从而使其能够适应不断变化的地质条件。在围岩变形预测中,这一动态更新过程有助于逐步提高对围岩位移和围岩稳定时间预测的准确性。依托浙江温州市小燕隧道进行了方法的验证与应用,并对隧道左洞共计23组断面的围岩稳定性进行合理可靠的预测分析,验证其预测的有效性。对比分析表明,BiLSTM在时序预测和最终沉降值的预测方面优于其他深度学习模型,在长期时序预测中能够有效捕捉围岩稳定时间和位移波动。To overcome the limitations of traditional warning methods used during highway tunnel excavation,this paper proposes a dynamic time-series updating model based on the bidirectional long short-term memory(BiLSTM)network.As tunnel excavation advances,the model dynamically adjusts its weights by incorporating new sectional data,thereby adapting to the continuously changing geological conditions.This dynamic updating process helps to progressively enhance the accuracy of predictions for surrounding rock displacement and stability time.The method was validated and applied in the Xiaoyan Tunnel in Wenzhou,Zhejiang,where the stability of surrounding rock across 23sections of the left tunnel was reliably predicted and analyzed,verifying the effectiveness of the predictions.Comparative analysis shows that BiLSTM outperforms other deep learning models in both time-series prediction and final settlement estimation,effectively capturing surrounding rock stability time and displacement fluctuations during long-term predictions.

关 键 词：公路隧道 BiLSTM 变形预测 拱顶沉降 

分 类 号：U452.11[建筑科学—桥梁与隧道工程]