Abstract
This work presents a research-centric perspective on digital twinning of physical processes as it applies to efforts in closed-loop research with self-driving laboratories, as informed by efforts at Natural Resources Canada. Efforts at CanmetMATERIALs are exploring how multi-physics simulations can be coupled with automated experimentation within modular, traceable workflows that iteratively update models to guide experimental decisions in optimization and exploration activities. Emerging AI-enabled approaches, such as surrogate modeling, and generative methods, are being examined as supporting tools rather than replacements for physics-based models.
In this work we focus on the overview and implementation of a proposed architecture, demonstrated using a simulation-driven case study on the generative design of 3D printed CO2 capture structures. This case study illustrates how digital twins can be integrated into the structured workflows to support the exploration of complex process-structure-property relationships.