Abstract

In the digital era, with growing volumes of infrastructure, environmental, and operational data, there is a need for next-generation tools that can seamlessly integrate data and models to support intelligent, real-time decision-making. This research develops and deploys technologies that will lead to prescriptive digital twins (DTs) for infrastructure, optimising design, construction, and operational processes.

The work focuses on the automatic reconstruction of DTs in two areas: i) modelling and assessment for effective and optimised design; and ii) reconnection of existing assets through machine learning-based information retrieval, digital model generation, condition mapping, and computational analysis for structural capacity assessment.

For the predictive phase, digital data and high-fidelity numerical models are integrated through parametric and surrogate modelling to enable real-time design evaluation and control. In modelling part particularly focuses on meshless methods as an effective means of integrating digital and numerical models, eliminating approximations caused by meshing. Our group is developing and advancing several such approaches, including Isogeometric Analysis (IGA), CutFEM, Singular Boundary Method (SBM), and other hybrid techniques to address soil–structure interaction problems in excavation modelling and elastodynamics.

For the maintenance phase, zero-shot segmentation and defect detection methods enable automated reconstruction of semantically rich structural models, supporting condition-based performance assessment and prioritised maintenance. The proposed approaches enhance automation, accuracy, and resilience, paving the way for prescriptive digital twins that transform infrastructure project planning, operation, and asset management.