Establishment of a Digital Twin Model to Predict and Analyze Greenhouse Gas Emission and Transport in Turbulent Flames from Lagrangian Viewpoint

A data-driven digital twin model is developed for rapid prediction of greenhouse gas emissions such as CO₂ during flame combustion, and then the interactions between combustion states and vortices are elucidated using Lagrangian analysis method. First, numerical solutions are obtained from Reynolds Averaged Navier-Stokes (RANS) simulations and used to train a nested U-shaped neural network. By encoding and decoding the characteristics of the mixed flow field, the flame temperature, velocity and emission concentration are predicted. In addition, the accuracy of the prediction is discussed through three quantitative metrics. The analyzed results demonstrate the effectiveness and accuracy of the method on the current dataset. Finally, the transport and mixing processes of CO₂ are analyzed based on the predicted data and the coherent structure is identified from Lagrangian viewpoint. Importantly, the interaction of the flame and the flow structures is characterized, and the correlations are evaluated by the coherence ratio and mixing parameters. As a conclusion, the coupling of neural network and Lagrangian analysis allows for predictive modeling of turbulent flames, visualization of internal processes, what-if analysis, and control of greenhouse gas emissions.