Digital Twin Modelling via Integration of Simulation and Data-Driven Methods

Digital twins -high-fidelity digital counterparts of physical assets -are increasingly used to solve real-world problems across industries.Building a high-quality digital twin requires an integrated stack spanning IoT, data processing, modelling & simulation, 3D visualisation, and networking, with the modelling layer pivotal.Yet widely adopted modelling practices remain limited.We propose a digital twin modelling method that combines simulation and data-driven modelling, selecting among three integration strategies by goal: (i) accuracy enhancement via calibration, assimilation, and hybridisation; (ii) execution efficiency via surrogate or reduced-order models; and (iii) decision optimisation via simulation-in-the-loop using learned response surfaces.We formalise selection criteria and workflows for each strategy and show their composition within a single methodology.A smart farm case study demonstrates improved predictive accuracy, reduced runtime, and support for operational optimisation, illustrating practical value for purpose-built digital twins.