Toward an evolvable digital twin: a knowledge-driven architecture for optical manufacturing

Digital twins demonstrate significant potential for addressing optical manufacturing challenges associated with high precision demands and empirical dependency, yet existing models lack dynamic and deep knowledge integration. This paper proposes a plant-growth-inspired bionic digital twin model featuring a multi-vascular knowledge network for computable causal reasoning and a dual-loop (growth and metabolic) cognitive architecture to support knowledge evolution and real-time decision-making. Validated through continuous polishing cases, the model enhances surface form error convergence, parameter optimization, and sustained cognitive adaptability through continuous knowledge updates. This research provides a pathway toward adaptive, self-evolving digital twins for intelligent optical manufacturing applications.