基于深度学习的盾构姿态预测及纠偏研究

Study on Shield Attitude Prediction and Deflection Correction Based on Deep Learning

以苏州某在建隧道工程为研究背景,基于机器学习技术提出一种盾构姿态预测模型和纠偏方法。首先通过卷积神经网络挖掘盾构姿态数据的空间特征,然后通过双向长短期记忆神经网络挖掘数据的时序特征,紧接着通过注意力机制挖掘重要的时间特征信息。在预测结果的基础上,引入Apriori算法对盾构数据的关联规则提取,并提出盾构姿态纠偏方法。实验结果表明该文提出的盾构姿态预测模型具有较好的泛化能力,且相较于选取的3种基准模型,得到的均方根误差和平均绝对误差值最小,具有更高的预测精度。基于姿态理论控制模型,构建多环姿态控制模型,实现对姿态调整获取参数建议值,为智能化姿态控制提供参考依据。

Taking a tunnel project under construction in Suzhou as the research background,this paper proposes a shield attitude prediction model and correction method based on the machine learning technology.Firstly,the spatial features of shield posture data were mined through a convolutional neural network.Then,the temporal features of data were mined through a bidirectional long short-term memory neural network.Afterwards,the important temporal feature information was mined through the attention mechanism.On the basis of the prediction results,the Apriori algorithm is introduced to extract the association rules of shield data,and the shield attitude correction method is proposed.Experiments show that the proposed prediction model in this paper has good generalizability.Compared to the three selec-ted baseline models,it achieves the smallest root mean square error and mean absolute error values,indicating higher prediction accuracy.Based on the attitude theory control model,a multi-loop attitude control model is constructed to obtain parameter suggestions for attitude adjustment,which provides a theoretical reference for intelligent attitude control.