Underground structures play an increasingly important role in transportation networks and urban areas. Thus, ensuring their structural integrity is essential for safety and operational efficiency. Among the Structural Health Monitoring (SHM) methods already proposed for this type of structure, only a few studies propose vibration-based analyses. Furthermore, data-driven monitoring of infrastructure networks would require the installation of several sensors on each structure, which may be prohibitively expensive for local administrations. The lack of sufficiently large and comprehensive datasets can be addressed through Population-Based Structural Health Monitoring (PBSHM). The PBSHM approach, recently proposed for bridges, wind turbines and aircraft, adopts transfer-learning algorithms to share damage-state knowledge among similar structures and establish a large-scale monitoring system when only a few data are available. This study investigates the potential extension of knowledge sharing to underground structures, such as metro tunnels, by analysing feasible features and damage identification strategies and exploiting the numerical results of two dynamic finite element simulations to provide a domain adaptation case study.