岩土与地下工程 2026年 第卷 第07期

DOI: 10.7672 / sgjs2026070072

铁板砂地层环境下尾盾变形及神经网络预测分析研究

钟 涵¹’²’³,何 源⁴,许 超¹’²,吴诗琦¹’³,张飞雷¹

作者简介:

钟 涵,部门副总工程师,工程师,E⁃mail: zhonghancug@ 163. com

作者单位:

1. 中交第二航务工程局有限公司,湖北 武汉 430040; 2. 交通运输行业交通基础设施智能制造技术研发中心,湖北 武汉 430040; 3. 长大桥梁建设施工技术交通行业重点实验室,湖北 武汉 430040;4. 中交二航局第三工程有限公司,江苏 镇江 212021

基金项目:

∗国家自然科学基金(52379114)

摘要:

盾构掘进过程中,由于地质、盾构操作等因素,会造成盾构机尾盾整体或局部变形. 某海外大直径水下盾构在高埋深、铁板砂地层中纠偏掘进时,出现盾构机尾盾局部变形,盾构无法继续掘进的情况. 经过尾盾变形原因分析,盾构机尾盾开孔取芯探查和矫正施工,尾盾局部变形恢复,实现盾构脱困. 复推后对尾盾变形原因及采取的控制措施进行分析,总结出盾构机尾盾局部最大变形量与以下因素有关:地层参数、盾构姿态、分区压力、贯入度、仿行刀开启行程、径向孔放砂量. 利用神经网络预测模型对已有 25 组尾盾变形样本进行训练,得到每环掘进完成后尾盾最大变形量预测模型,并通过另外 5 组尾盾变形数据验证了模型准确性,用于后续盾构掘进尾盾变形预测.经过实践论证,该神经网络模型预测变形值与实测值误差在 10%以内且预测值比实测值偏大,可用于盾构在高埋深、铁板砂地层尾盾变形预测.

English:

In the process of shield tunneling, due to factors such as geology and shield operation, theoverall or local deformation of the shield tail will be caused. When an overseas large⁃diameter underwatershield was excavated in high buried and iron slab sand stratum, the local deformation of the shield tail ofthe shield machine occurred, and the shield cannot continue the excavation. After the analysis of thecauses of the deformation of the shield tail , the hole coring exploration and correction construction of theshield tail , the local deformation of the shield tail was restored, and the shield was freed from the trap.The causes of shield tail deformation and the control measures were analyzed. It is concluded that thelocal maximum deformation of the shield tail is related to the following factors such as formationparameters, shield attitude, partition pressure, penetration, imitation cutter opening stroke, and radialhole sand discharge. The artificial neural network prediction model is used to train 25 groups of shield taildeformation samples, and the prediction model of the maximum deformation of the shield tail after eachring excavation was obtained. The accuracy of the model was verified by the other five groups of shield tail deformation data, which is used for the prediction of shield tail deformation in subsequent shieldtunneling. Through practical demonstration, the error between the predicted deformation value and themeasured value of the BP neural network model is within 10%, and the predicted value is larger than themeasured value, which can be used to predict the deformation of the shield tail of the shield in the highburied depth and the iron plate sand stratum.