Machine learning (ML), one of the AI techniques, has been used in geotechnical engineering for over three decades, resulting in more than 600 peer-reviewed papers. However, AI applications in geotechnical engineering are significantly lagging compared with other fields. One of the reasons for the lagging is that hyperparameters used in many AI techniques need physical meaning in geotechnical applications. This paper focuses on widening the applications of ML in predicting tunneling-induced short- and long-term ground settlement and optimizing ML architectures considering their interpretability and ability to provide physically consistent results. Informed by the underlying physics knowledge, tunneling-induced ground settlement is divided into long-term and short-term settlements since different mechanisms and influencing parameters contribute to these two deformation types. Based on the above considerations, this paper introduces a physics-informed ensemble machine learning (PIML) framework to strengthen the connection between ML techniques and physics theories, followed by identifying/utilizing different sets of parameters for effectively predicting short- and long-term tunneling-induced settlements, respectively. Together with in situ observations and experimental lab results, parameters obtained from physics equations are set as inputs for the ML models. Results show that the proposed PIML framework effectively predicts tunneling-induced ground movements, with a predicting accuracy above 0.8. Additionally, parametric studies of variable significance and comparisons among different ML designs reveal that in situ observed dynamic parameters, for instance tunnel face and monitoring points (DTM), gap parameter, and tunnel depth, are essential in predicting tunneling-induced short-term settlement, while predicting long-term settlements largely depends on features, such as tunnel depth, volume compressibility, and excess pore pressure, derived from physics theories.

Synthetic Aperture Radar (SAR) interferometry is a formidable technique to monitor surface deformation with a millimeter detection resolution. This study applies the Persistent Scatter-Interferometric Synthetic Aperture Radar (PSInSARTM) technique to measure ground subsidence related to a twin-tunnel excavation in downtown Los Angeles, USA. The PSInSARTM technique is suitable for urban settings because urban areas have strong reflectors. The twin tunnels in downtown Los Angeles were excavated beneath a densely urbanized area with variable overburden depths. In practice, tunneling-induced ground settlement is dominantly vertical. The vertical deformation rate in this study is derived by combining Line of Sight (LOS) deformation velocities obtained from SAR images from both ascending and descending satellite orbits. Local and uneven settlements up to approximately 12 mm/year along the tunnel alignment are observed within the allowable threshold. No severe damages to aboveground structures were reported. Furthermore, ground movements mapped one year before tunnel construction indicate that no concentrated ground settlements pre-existed. A Machine Learning (ML)-based permutation feature importance method is used for a parametric study to identify dominant factors associated with the twin-tunneling induced uneven ground subsidence. Six parameters are selected to conduct the parametric study, including overburden thickness, i.e., the thickness of artificial fill and alluvium soils above the tunnel springline, the distance between the two tunnel centerlines, the depth to the tunnel springline, building height, the distance to the tunnel, and groundwater level. Results of the parametric analysis indicate that overburden thickness, i.e., the thickness of artificial fill and alluvium soils above the tunnel springline, is the dominant contributing factor, followed by the distance between tunnel centerlines, depth to the tunnel springline, and building height. Two parameters, the distance to the tunnel, and the groundwater level, play lesser essential roles than others. In addition, the geological profile provides comprehension of unevenly distributed ground settlements, which are geologically sensitive and more concentrated in areas with thick artificial fill and alluvium soils, low tunnel depth, and high groundwater levels.