<p>Web tension control plays a crucial role in maintaining product quality and operational efficiency in roll-to-roll (R2R) systems. Precise tuning of control parameters ensures stable web tension, minimizes defects, and enhances overall system efficiency. However, traditional tuning methods still require considerable time and effort when addressing the nonlinear and time-varying dynamics of R2R systems, particularly under changing operating conditions. To overcome these challenges, an AI-driven approach was developed to optimize control parameters based on key control performance metrics. The proposed approach leverages a deep neural network to efficiently model complex relationships among control parameters of the R2R system. The trained models were integrated via a weighted cost function, and a grid search optimization was employed to identify an optimal combination of control gains. To evaluate the practical impact of AI models on tension control, experimental validation was conducted using an actual R2R system. The optimized combination of control parameters, predicted by the deep neural network, was implemented, and the performance metrics—such as time constant, overshoot, and settling time—were compared against real machine data across various operating speeds. This experimental verification confirms both the compatibility and effectiveness of our AI-driven approach, underscoring its significant impact on enhancing tension control in R2R systems. Furthermore, the proposed methodology paves the way for integration into a digital twin framework of the R2R system, presenting a significant step toward autonomous manufacturing.</p>