Applying Machine Learning in CFD to Study the Impact of Thermal Characteristics on the Aerodynamic Characteristics of an Airfoil

A computational fluid dynamic (CFD) and machine learning approach is used to investigate heat transfer on NASA airfoils of type NACA 0012. Several different models have been developed to examine the effect of laminar flow, Spalart flow, and Allmaras flow on the NACA 0012 airfoil under varying aerodynamic conditions. Temperature conditions at high and low temperatures are discussed in this article for different airfoil modes, which are porous mode and non-porous mode. Specific parameters included permeability of 11.36 x 10-10 m2, porosity of 0.64, an inertia coefficient of 0.37, and a temperature range between 200 K and 400 K. The study revealed that a temperature increase can significantly increase lift-to-drag. Additionally, employing both a porous state and temperature differentials further contributes to enhancing the lift-to-drag coefficient. The neural network also successfully predicted outcomes when adjusting the temperature, particularly in scenarios with a greater number of cases. Nevertheless, this study assessed the accuracy of the system using a SMOTER model. It has been shown that the MSE, MAE, and R for the best performance validation of the testing case were 0.000314, 0.0008, and 0.998960, respectively, at K = 3. However, the study shows that epoch values greater than 2000 increase computational time and cost without improving accuracy. This indicates that the SMOTER model can be used to classify the testing case accurately; however, higher epoch values are not necessary for optimal performance.