Digital Twin-Driven Adaptive Multi-Point Heat Variation Tracking and Predictive Control System Using Edge AI

Summary

Accurate thermal regulation in industrial systems is often challenged by limited sensing coverage and dynamic operating conditions. This study proposes a Digital Twin-driven Adaptive Multi-Point Heat Variation Tracking and Predictive Control System using Edge AI. The framework integrates distributed temperature sensing, thermal-state reconstruction, predictive analytics, and reinforcement learning (RL)-based control for real-time thermal management. A digital twin continuously synchronizes with the physical process to estimate the complete thermal field, while an RL controller optimizes control actions without manual tuning. Simulation results demonstrate superior performance compared with a conventional PID controller, achieving energy efficiency improvements from 78.2% to 89.5% under low-load conditions and from 68.9% to 85.4% under dynamic operating conditions, representing gains of 14.5–23.9%. The Thermal Uniformity Index also improved from 0.82 to 0.93 under low-load conditions and from 0.70 to 0.87 under dynamic conditions, corresponding to improvements of 13.4–24.3%. In addition, the developed system achieved lower temperature tracking errors, improved disturbance rejection, and enhanced thermal stability. These results demonstrate the effectiveness of combining digital twins, Edge AI, and reinforcement learning to achieve intelligent, scalable, and energy-efficient industrial thermal management.

Index Terms

Browse additional articles authored by one or more contributors to this paper.