Precise control of zone temperatures is critical for the efficient operation of Reheat furnaces and maintaining product quality. This study combines experimental data with machine learning, specifically Long Short-Term Memory (LSTM) models, to predict furnace temperatures over several hours. A Chemical Reactor Network (CRN) [1] is incorporated into the model to simulate fuel transitions, particularly from natural gas to hydrogen. Both LSTM and CRN models are integrated within a Digital Twin, enabling real-time monitoring, inefficiency detection, and fuel optimization. The CRN simulates combustion dynamics, using simplified reactors to represent furnace flow and combustion processes under various fuel conditions. The Digital Twin framework consists of three core components: Computational Fluid Dynamics (CFD), Machine Learning (LSTM), and CRN. The LSTM model predicts temperature changes, as shown in Figure 1, and burner firing control to ensure each zone meets its target temperatures. Meanwhile, the CRN uses high-fidelity CFD data, as shown in Figure 2, as a baseline to model the furnace’s response to fuel composition changes, dividing it into distinct zones modeled as independent reactors. This dual modeling approach allows comprehensive control and prediction of thermal behavior, helping manage the impact of hydrogen fuel on combustion and product quality. By integrating machine learning with CRN in a Digital Twin, this approach aims to optimize furnace operations and supports the transition to sustainable fuels. It enhances efficiency, predicts temperature profiles, and facilitates decarbonization through the shift from natural gas to hydrogen.