A redistribution of stress is caused by the excavation of a new tunnel within the soil surrounding the existing shield tunnel structure. This stress redistribution results in alterations to the soil and water loads around the existing tunnel structure, ultimately impacting its stress and deformation. The deformation response of existing tunnels is primarily focused on in most existing studies, with limited research conducted on the coupling system of new tunnel excavation, soil stress state, and deformation and stress of existing tunnels. This paper proposed an innovative methodology to predict deformation in existing tunnel structures and estimate additional stress during the excavation of a new shield tunnel. The reliability of the suggested calculation method was verified using measured internal force and convergence data of the structures. The results demonstrated that the Long Short-Term Memory (LSTM) model could more effectively establish the temporal correlation between the excavation parameters of new tunnels and the deformation of existing tunnels. Compared to the Recurrent Neural Network (RNN) model, the R of the LSTM model increased by 11.81% in the prediction of the tunnel waist and 9.38% in the prediction of the tunnel crown. The Light gradient boosting machine (LightGBM) algorithm was used to analyze the crucial factors of deformation of the existing tunnel structures during shield tunneling, and the results showed that the relative position and the average earth pressure were the essential factors. Furthermore, a calculation method for the additional stress of existing tunnel structures was proposed based on the load structure model. The method assumed that the additional load was distributed rectangularly at the top and bottom of the existing tunnel, and triangularly at the waist of the tunnel. The comparison between the calculated results and the measured results indicated that the phased characteristics of the internal force changes in the existing tunnel structure during the construction of a new tunnel could be accurately reflected by this method.

中文翻译：

---

基于 LSTM 的变形预测，用于估计相邻盾构隧道施工引起的现有隧道结构的附加应力

在现有盾构隧道结构周围的土壤中开挖新隧道会导致应力重新分布。这种应力重新分布导致现有隧道结构周围的土壤和水荷载发生变化，最终影响其应力和变形。现有研究大多集中于既有隧道的变形响应，对新隧道开挖、土体应力状态、既有隧道变形与应力耦合系统的研究较少。本文提出了一种创新方法来预测现有隧道结构的变形并估算新盾构隧道开挖过程中的附加应力。利用结构的实测内力和收敛数据验证了所建议计算方法的可靠性。结果表明，长短期记忆（LSTM）模型可以更有效地建立新隧道开挖参数与现有隧道变形之间的时间相关性。与循环神经网络（RNN）模型相比，LSTM模型在隧道腰部预测中的R提高了11.81%，在隧道拱顶预测中的R提高了9.38%。采用光梯度增强机（LightGBM）算法对盾构掘进过程中既有隧道结构变形的关键因素进行分析，结果表明相对位置和平均土压力是影响因素。进一步提出了一种基于荷载结构模型的既有隧道结构附加应力计算方法。该方法假设附加荷载在现有隧道的顶部和底部呈矩形分布，在隧道腰部呈三角形分布。计算结果与实测结果对比表明，该方法能够准确反映新隧道施工过程中既有隧道结构内力变化的阶段性特征。