Framework for a Digital Twin System in Corrosion Prediction

ABSTRACT Accurate corrosion prediction in complex automotive assemblies remains challenging due to fragmented data sources, simplified geometry representations and limited integration between simulation, physical testing and AI‐based analysis. Existing approaches typically address isolated aspects of corrosion behaviour without a unified system architecture. This paper presents a surface‐centric digital twin framework for corrosion prediction that integrates digital product data, multiphysics simulation results (CFD/FEM), algorithmic surface classification and physical validation data into a single source of truth. All corrosion‐relevant information is mapped consistently onto a high‐resolution triangulated product surface, enabling traceable data flows and modular interaction between system components. The applicability of the framework is demonstrated in two automotive case studies, where AI‐based surface classification is transformed into geometric entities and embedded into electrolyte simulations, significantly improving simulation fidelity for adhesive‐sealed gaps. Furthermore, the model’s possibility of predicting corrosion on geometries based on the available data shows the functionality of the digital twin. The proposed modular architecture therefore supports transparency, reproducibility and iterative design decisions, enabling early‐stage corrosion assessment and more efficient corrosion protection development. This modular architecture approach, as well as the surface‐centred digital twin for corrosion prediction, both offer a novel approach that improves corrosion prediction abilities.