Prediction of the further mechanical behaviors is vitally important for tunnel engineering to prevent disasters and maintain stability. It is a challenge for most existing researches to couple multiple influence factors. This study aims to develop a novel Load-Temporal (LT) model to predict the further mechanical behaviors of structure using machine learning method, which considers the effect of both historical performance and external loads. As a case study, the developed model is employed in an underwater shield tunnel, in which a Structural Health Monitoring System (SHMS) is installed. Based on the monitoring data obtained from SHMS, plenty of data experiments are conducted to develop model and determine the optimal parameters. Also, the comparison analysis is adopted to indicate the prediction accuracy of proposed model is higher than that of the classical models. The predicted ability of LT model is discussed via experiments of different time scale in further. As promising applications, LT model is used to predict the mechanical behaviors under various boundary conditions, based on which to determine the dangerous states and the structural performance under these conditions.

预测进一步的力学行为对于隧道工程预防灾害和保持稳定至关重要。耦合多个影响因素是大多数现有研究的挑战。本研究旨在开发一种新的负载时间 (LT) 模型，以使用机器学习方法预测结构的进一步力学行为，该模型考虑了历史性能和外部负载的影响。作为案例研究，开发的模型用于安装了结构健康监测系统 (SHMS) 的水下盾构隧道。基于从SHMS获得的监测数据，进行了大量的数据实验，以建立模型并确定最佳参数。还，通过对比分析表明，所提模型的预测精度高于经典模型。通过不同时间尺度的实验进一步讨论了LT模型的预测能力。作为有前景的应用，LT 模型用于预测各种边界条件下的力学行为，以此确定这些条件下的危险状态和结构性能。