Accurate solutions to real-time complex nuclear engineering problems require high computational resources. Artificial Intelligence proven effective in various engineering problems is a promising method for analysis of heat transfer problems in nuclear engineering. This research uses a deep learning approach of artificial intelligence to study the heat transfer behavior of the lithium film flow on the plasma-facing side of the Tokamak divertor. A Multi-Domain Physics Informed Neural Network (MD-PINN) model was developed to solve the steady-state convection-diffusion partial differential equation leveraging automatic differentiation to compute the loss function from the residuals of the equation. The model was validated with analytical solutions for 1D and 2D heat conduction problems of two layer two materials. The model was improved through parametric analysis for convergence rate and accuracy. Steady-state temperature distribution was obtained using the MD-PINN model and compared with the reference results. A good agreement of the results shows the capability of the MD-PINN as an alternative to the numerical simulation.

Artificial Intelligence;heat transfer;diverter;Multi-Domain Physics Informed Neural Network;Differential Equations