Maintenance-oriented Digital Twin for Underground Infrastructure with Sensing, Machine Learning, BIM and Simulation

As buried pipelines face increasing maintenance challenges due to ageing, climate change, and infrastructure deterioration, failures can lead to severe economic losses and public safety risks. The M-Twin4US project aims to develop a novel, maintenance-oriented Digital Twin (DT) platform tailored for buried pipelines, with the flexibility to expand to other underground infrastructure. M-Twin4US will integrate advanced sensing technologies, Ground-Penetrating Radar (GPR) and closed-circuit television (CCTV), with state-of-the-art machine learning methods like Multi-task Transformers and Segment Anything Models (SAM). The project will address critical challenges in soil-pipe interaction analysis, GPR response prediction, and condition assessment by coupling digital modelling, machine learning, numerical modelling (hydro-mechanical and electromagnetic modelling) and real-scale testing. The project will develop a customisable surrogate model to predict future pipeline performance and offer an AI-aided toolkit for stakeholders to prioritise actions such as rehabilitation, excavation, or new construction. The project will be implemented in collaboration with key industry partners and will contribute to transforming underground infrastructure maintenance from a reactive to a proactive paradigm, supporting sustainability, safety, and resilience goals.