A CLOSED-LOOP DIGITAL TWINS FRAMEWORK FOR REAL-TIME OPTIMIZATION OF SMART PRODUCTION SYSTEMS

The rapid evolution of smart manufacturing systems within the Industry 4.0 paradigm requires advanced Digital Twins solutions capable of dynamic learning, adaptive control, and realtime decision support.Conventional Digital Twins implementations, however, often rely exclusively on either physics-based modeling or data-driven analytical techniques.While physics-based models provide interpretability and stability, they lack adaptability under dynamic operating conditions, whereas purely data-driven approaches may suffer from limited generalization and robustness.To address these limitations, this paper proposes a Hybrid Digital Twins (HDT) framework designed for real-time production optimization in smart manufacturing environments.The proposed framework integrates first-principles physical modeling with artificial intelligence-based learning mechanisms, forming a closed-loop system that continuously synchronizes physical processes with their digital representation.The hybrid structure combines mechanistic equations describing process dynamics with machine learning predictors that forecast near-future system behavior, enabling proactive control adjustments.A multi-objective optimization function is formulated to simultaneously minimize energy consumption, production cycle time, and process instability while maintaining product quality.Online parameter calibration and feedback-driven optimization ensure adaptive system behavior under stochastic disturbances.The effectiveness of the proposed HDT framework is validated through simulation experiments conducted on a representative assembly-line production model.Comparative analysis with a traditional physics-based Digital Twins demonstrates significant performance improvements, including an increase in energy efficiency of approximately 12%, an enhancement in process stability of 18%, and a reduction in cycle time of 15%.The results confirm that hybrid Digital Twins enable faster convergence, improved operational resilience, and enhanced real-time control.The proposed approach provides a practical foundation for intelligent production optimization and supports the transition toward more autonomous and sustainable smart manufacturing systems.