To advance sustainable underground development, current practices must evolve to the next level: the Digital Twin (DT). As a real-time single source of truth, a DT consolidates timely and accurate information about infrastructure conditions, enabling optimized design and proactive maintenance of underground structures. We propose a new framework for the reconstruction of DTs, focusing on automation in component detection, digital model reconstruction, condition mapping, and quantification and numerical analysis to estimate its bearing capacity. Tunnel point clouds are converted into images and processed by the Segment Anything Model (SAM) [1] for effective zero-shot segmentation, outperforming supervised learning [2]. Geometric parameters were then extracted from previous results for reconstructing DTs, deformation, and displacement monitoring. The fine-tuned SAM was also used for defect detection and recorded in the digital model, providing automated assessments of defect severity. The second step is the analysis of the design. We demonstrate how the seamless integration of digital and high-fidelity numerical models updated with the detected condition. Higher-order numerical models integrated with digital design tools ensure accuracy while offering superior computational efficiency [3]. We will model detected damage by material stiffness degradation according to the specified damage evolution response. For robust implementation, a computational strategy with sub-stepping integration at the local level will be combined with error control at the global level. The proposed approach is efficient, accurate and robust, and as such has potential to transform geotechnical project planning through efficient, computationally enhanced decision-making. REFERENCES [1] A. Kirillov, et al. "Segment anything." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2023. [2] Y. Zehao, J. Ninic et al. "Sam-based instance segmentation models for the automation of structural damage detection." Advanced Engineering Informatics 62 (2024): 102826. [3] J. Ninić, et al. "BIM-to-IGA: A fully automatic design-through-analysis workflow for segmented tunnel linings." Advanced Engineering Informatics 46 (2020): 101137.