AI-driven digital twin for autonomous web tension control in Roll-to-Roll manufacturing system.

Abstract

While the roll-to-roll manufacturing process plays a key role in high-throughput and cost-effective production, precise web tension control remains a critical challenge due to the dynamic interaction of web materials and roller mechanics. To address these challenges, this study proposes an AI-driven digital twin framework for autonomous web tension control optimization. The proposed method integrates Bayesian optimization with Gaussian process modeling to efficiently explore and adjust proportional and integral control parameters. While the operation of the roll-to-roll system is managed through a real-time client-server communication, system responses are designed to be iteratively refined by the proposed surrogate model. Experimental validation on an actual roll-to-roll manufacturing system demonstrates that the optimized control strategy significantly reduces tension variation and improves system stability. The proposed method highlights the potential of AI-integrated digital twins in autonomous manufacturing, which can offer a scalable solution for a variety of industrial applications.