Abstract:

Starting from the challenges posed by extreme environments in deep underground engineering, this presentation reveals the time constraints and cognitive limitations associated with the “golden hour” in tunneling. To address the “small data” challenges arising from significant geological variability and sparse monitoring, this presentation emphasizes the necessity of incorporating physics and knowledge of complex systems. A rapid analysis approach that seamlessly integrates physics, knowledge and data should be developed. On this basis, the presentation clarifies the pathway for transitioning from digital twins to engineering intelligent agents. We propose to develop tunnel intelligent agents equipped with five core capabilities: “self-sensing, self-decision-making, self-control, self-organization, and self-learning.” These technologies have been implemented and validated in projects such as the Ehan High-speed Grand Canyon Tunnel. Finally, the presentation highlights the need to establish an autonomous “machine-environment” collaborative system based on a System of Systems (SOS) architecture. This would enable a paradigm shift from a “human-machine-environment” framework with minimal human assistance to a fully integrated “machine-environment” framework characterized by unmanned operations.