In this study, a new mechanical approach, discontinuum analysis, was applied to the structural examination of mountain tunnel lining, which is different from the conventional approach based on continuum mechanics. A series of behaviors of mountain tunnel lining, from a normal state to the whole structure collapse accompanied by cracks, peeling/flaking, and spalling, and the development of the destruction, were simulated using the distinct element method. Block distinct element method analysis and granular distinct element method analysis were performed to reproduce the results by full-scale loading experiments of mountain tunnel linings under conditions where axial force and bending moment occurred, and conditions where the influence of bending moment was dominant. As a result, the reproducibility of structural separation such as cracks, peeling/flaking, and the whole lining structure collapse and its development was examined. The usefulness of the method in evaluating structural separation, the destruction of the entire lining structure and its development was discussed, and issues in its applicability were also clarified.

 The support design and the excavation plan of the underground structure such as mountain tunnels and underground power plants, are updated based on the results of measurements of ground deformation and observations of the tunnel face. Since these design changes are sometimes examined through numerical analysis, it is necessary to consider various uncertainties such as initial and boundary conditions, geological structures, and mechanical properties of rocks. An ensemble-based data assimilation (EBDA) is studied to achieve a highly credible prediction of the ground deformation in the construction cycle. An excavation analysis method using EBDA was presented and applied to the problem of tunneling in the ground assuming elastic model. In addition, numerical experiments simulating a triaxial compression test were conducted to clarify issues in the application of EBDA to the prediction of deformation behavior of elasto-plastic model.

 Undersea tunnels are in extraordinary conditions for maintenance in that they are affected by high water pressure. In particular, it is difficult to evaluate the soundness of the bedrock, including the effect of the grout injected because the back side of the tunnel lining is difficult to monitor. In this study, we conducted data analysis and numerical simulation of the measured pore water pressure for the purpose of constructing a method to grasp the state of the bedrock and the grout. As a result of data analysis, the change of the pore water pressure turned out to be affected by tide. Numerical simulation was performed under multiple parameter settings. The parameters that explain the current state of the bedrock inspected from the measured pore water pressure and the fluctuations in pore pressure that occur when the function of the grout changed were determined.

 Video Dust Meter (VDM) is a newly developed video camera specialized for detecting floating dust, designed to monitor tunnel dust levels in construction. This study has developed an object detector using deep learning to automatically detect and count dust particles from video captured by the VDM. Laboratory experiments simulating dust-filled environment in tunnel construction were conducted to prepare image datasets for deep learning. YOLO v5 was used for object detection. Utilizing the developed YOLO dust particle detector, the VDM video of the laboratory dust experiment was analyzed to count the dust particles, and the results were compared to the reference measurements obtained by the Digital Dust Meter (LD-5R). As a result, the number of dust particles measured with the VDM and YOLO detector correlates highly with the measurements obtained by the reference Digital Dust Meter. The calibration coefficient between the two methods corresponds approximately to the ratio of the assessed air volumes by each method. It implies that the VDM can estimate the dust levels obtained by the Digital Dust Meter.