基于循环神经网络的盾构隧道#br# 引发地面最大沉降预测

伴随着计算机技术的快速发展，机器学习等新兴算法正在被越来越多地运用于预测隧道掘进引发的地面最大沉降。在隧道施工过程中，由盾构机和地面监测点位采集的数据具有很强的序列化特征，而传统的机器学习算法对序列数据的处理存在一定的局限性。循环神经网络(RNN)具有极强的对时序型数据的处理能力，在视频识别、语音翻译等领域有着广泛的应用。采用两种RNN模型(LSTM、GRU)和传统的BP神经网络模型，以地质参数、几何参数和盾构机参数作为输入，对隧道施工过程中引发的地面最大沉降进行预测分析。结果显示，RNN对隧道沉降的预测结果优于传统的BP神经网络模型，并且RNN在连续未知区段的预测结果比BPNN更加稳定。

Prediction of maximum ground settlement induced by #br# shield tunneling based on recurrent neural network

With the rapid development of computer technology, emerging algorithms such as machine learning are being increasingly used to predict the maximum ground settlement induced by tunneling. In the process of tunnel construction, the data collected by shield machines and ground monitoring points have strong serialization characteristics, and the traditional machine learning algorithm has certain limitations on the processing of such data. Recurrent neural network (RNN) has extremely strong ability to process time-series data, and has a wide range of applications in video recognition, speech translation and other fields. Taking geological parameters, geometric parameters and shield machine parameters as inputs, the prediction efficiency of two RNN models (LSTM, GRU) and a traditional BPNN model on the maximum ground settlement caused by the tunneling was investigated. The results show that the prediction result of RNN for tunnel settlement is better than that of traditional BPNN model, and the prediction result of RNN in continuous unknown sections is more stable than that of BPNN.