Neural network-based analysis of thermally driven convection in supercritical fluid for thermal systems

Supercritical fluids (SCFs) are widely used in many thermal and energy systems like cryogenic cooling systems because of their superior heat transfer characteristics and thermodynamic efficiency. However, their thermophysical properties vary rapidly near the critical point, which influences thermally driven convection and heat transfer. Therefore, understanding these variations is important for the analysis and designing of cryogenic cooling systems. Hence, the present study employs an innovative Machine Learning approach to analyze thermally driven convective heat propagation of a non-polar SCF flowing along a heated vertical plate. The Redlich–Kwong equation of state is employed to evaluate the coefficient of thermal expansivity (β). The variations of density, specific heat capacity, thermal diffusivity, and heat transfer coefficient are investigated in the subcritical, near-critical, and supercritical regions. The Multi-Layer Feed-Forward Neural Network (MLFFNN) is used to solve the nonlinear coupled partial differential equations governing the flow and thermal fields. The velocity and temperature profiles are discussed, and the effect of temperature on the Prandtl Number (Pr), Rayleigh Number (Ra), and Nusselt Number (Nu) is presented in a graph. Nu was calculated and plotted against Ra. The results show that the velocity profile initially increases and then decreases, whereas the temperature profile decreases continuously with increasing horizontal distance from the heated plate. The critical region exhibits the most significant variations in thermophysical properties resulting in higher Pr, Ra, and Nu, implying enhanced thermally driven convection and heat transfer. These observations indicate that the critical region governs the thermal behavior of cryogenic cooling systems involving SCFs. The obtained curves of SCFs and the properties predicted by the trained MLFFNN are similar to that evaluated by the traditional method.