Transforming the maintenance of underground infrastructure through Digital Twins: State of the art and outlook

Abstract

Underground infrastructure, designed to last for decades, play a vital role in urban life. Its maintenance and upkeep have significant societal values and contribute to sustainability. Simulation and modelling along with digitisation and virtualisation as key technologies in the context of Industry 4.0 have fundamentally transformed delivery of engineering projects. With climate change pose pressing challenges on the physical environment of human dwelling, infrastructure resilience has been strategised as a sustainable development goal. Inception of Industry 5.0 has been formulated surrounding this need incentivising to build and maintain with sustainability, resilience and human-centric as core values. Digital twin (DT) paradigm facilitated by a series of cross-disciplinary technologies has emerged to play a role toward these targets. Along with data management and artificial intelligence (AI), DTs present transforming potentials from a technology-focus approach to a value-focus approach. This paper conducts a systematic review on generation and applications of DTs for underground infrastructure maintenance, highlighting the multi-physics, multi-scale, and interdisciplinary characteristics of underground infrastructure. After examining challenges and opportunities for underground infrastructure, and thoroughly reviewing the existing definitions and maturity levels of DT in Section 1, a DT framework for maintenance is established under the “Descriptive-Reflective-Predictive-Prescriptive” maturity model featuring progressive function requirements, forming 2 Descriptive twin, 2.1 BIM multi-system modelling, 2.1.1 BIM for underground infrastructure, 2.1.2 Geo- and structural model, 2.2 Functions and interoperability of digitalisation, 2.3 BIM for maintenance, 3 Reflective twin, 3.1 Sensing technologies, 3.2 Data-driven interpretation, 3.3 Data fusion and augmentation, 3.4 Interoperability and semantics, 4 Predictive twin, 4.1 Deterioration modelling, 4.1.1 Physics-based methods, 4.1.2 Machine learning methods, 4.2 Multi-physics simulation, 4.2.1 Interoperability and computational scalability, 4.2.2 Case studies, 4.3 Towards interpretable and reliable predictive twin, 5 Prescriptive twin of the paper. Section 2 focuses on reviewing information modelling techniques for creating a descriptive twin, essentially answering the question of “what is it”. Section 3 explores sensing technologies and data-driven analytics to develop a reflective twin, addressing applications requiring the knowledge of “what is happening”. Section 4 examines methods of deterioration modelling and multi-physics simulations to establish a predictive twin, providing insights into “what will happen”. Section 5 investigates approaches to intervention decision-making through service-oriented and multi-criteria frameworks, providing outlook into a prescriptive twin that addresses “what should be done”. Finally, a brief summary and some prospects requiring further investigations are presented.