基于机器学习的盾构刀盘斜切混凝土地连墙推力预测  

Thrust Prediction During Diagonal Cutting of Concrete Diaphragm Wall for TBM Based on Machine Learning

作　　者：张颖欣 王占生[2] 陶宇帆 史培新 贾鹏蛟 ZHANG Yingxin;WHANG Zhansheng;TAO Yufan;SHI Peixin;JIA Pengjiao(School of Rail Transportation,Soochow University,Suzhou 215131,Jiangsu China;Suzhou Rail Transit Group Co.Ltd.,Suzhou 215004,Jiangsu China;Intelligent Urban Rail Engineering Research Center of Jiangsu Province,Suzhou 215131,Jiangsu China)

机构地区：[1]苏州大学轨道交通学院,江苏苏州215131 [2]苏州市轨道交通集团有限公司,江苏苏州215004 [3]江苏省智慧城轨工程研究中心,江苏苏州215131

出　　处：《河南科学》2024年第12期1792-1799,共8页Henan Science

摘　　要：针对城市轨道交通线网建设中盾构机穿越钢筋混凝土地连墙风险控制难题,本文提出了一种直接预测盾构切墙时刀盘切墙推力的方法 .首先,通过分析实际工程与现场试验数据,揭示了盾构推力与刀盘切墙推力之间的映射关系.结合贝叶斯优化方法(BO)、双向门控循环单元(BiGRU)和注意力机制构建了盾构总推力的混合预测模型(BO-BiGRU-Attention),经验证该模型具有较高的预测精度(R^(2)=0.91).最后,结合映射关系结果与盾构总推力预测模型,实现对刀盘切墙推力的精确预测(R^(2)=0.96).本研究解决了实际工程中的刀盘切墙推力的获取难题,为盾构机穿越钢筋混凝土地连墙的掘进参数选取及结构响应控制提供了理论基础,对实际工程施工具有一定的指导作用.In the construction of urban rail transit networks,controlling the risk of shield machines crossing reinforced concrete diaphragm walls is a significant challenge.This paper proposes a direct method to predict the cutterhead thrust during wall cutting.First,the mapping relationship between shield thrust and cutterhead cutting wall thrust is revealed by analyzing actual project and test data.Then,combining Bayesian Optimization(BO),Bidirectional Gated Recurrent Units(BiGRU)and attention mehanisms,a hybrid prediction model(BO-BiGRUAttention)for TBM thrust is constructed.Validation shows that this model has high prediction accuracy(R^(2)=0.91).Furthermore,by integrating the mapping relationship and the TBM thrust prediction model,precise prediction of the cutting wall thrust is achieved(R^(2)=0.96).This study proposes a novel solution to the problem of obtaining data on cutting thrust of the cutterhead in practical engineering applications,and provides theoretical foundations for selecting tunneling parameters and controlling structural responses when TBM cross reinforced concrete diaphragm walls.This study provides practical guidance for engineering construction.

关 键 词：盾构切墙 机器学习 注意力机制 推力预测 

分 类 号：TU94[建筑科学—建筑技术科学] TP391[自动化与计算机技术—计算机应用技术]