基于深度学习的破碎带盾构施工沉降预测分析

为进一步提高复杂地层条件下盾构沉降预测的准确性，以广州地铁7号线1期工程谢村站-钟村站区间盾构工程为依托，针对破碎带盾构隧道沉降控制难题，提出基于深度学习的人工智能预测模型。通过分析开挖面破碎带分布规律，确定将破碎带面积比作为地层特性参数。采用相关系数矩阵分析不同施工参数与破碎带面积比的相关性，确定采用刀盘转矩代表破碎带面积比实时描述地层分布特性。以刀盘转矩、盾尾间隙与注浆量作为输入值，地面沉降作为输出值训练深度学习模型，并利用训练后的深度学习模型进行沉降预测分析。通过分析预测结果与沉降实测值的对比验证预测模型的有效性。

Ground Settlement Prediction of Shield Tunneling in Fractured  Zone Based on Deep Learning Method

The settlement prediction precision of shield tunneling in complex ground should be improved. Hence, a settlement prediction model based on deep learning method is proposed by taking the shield tunneling project of Xiecun Station Zhongcun Station Section on Guangzhou Metro Line No. 7 for example.Firstly, the distribution law of fractured zone on tunneling face is analyzed, and the characteristics of the fractured zone is described by the ratio of fractured zone. And then the correlation between tunneling parameters and area ratios of fractured zone is analyzed by correlation coefficient matrix, and the ground distribution characteristics are described by cutterhead torque. Finally, the deep learning model is well trained by taking the cutterhead torque, shield tail gap and grouting amount as input values and the ground settlement as output value, and the settlement is predicted by the trained model. The prediction effectiveness of the model is verified by comparing the prediction results with the actual settlement values.