Multi-Objective Optimization of the Aerodynamic Efficiency of Naca 0012 Airfoil for High Reynolds Number Using Cfd and Ann-Based Genetic Algorithm

Abstract

CFD simulations have become a handy and versatile tool that is frequently utilized to understand the phenomena of fluid flow over complex shapes. One of these often-studied shapes is the airfoil, a shape that has vast applications in industry. In recent years, the optimization of aerodynamic shapes has emerged as an essential component in the field of aircraft design. The primary aim of this study is to present a comprehensive understanding of the process of characterizing airfoil attributes by employing multi-objective optimization techniques. Initially, CFD analysis had been performed on a NACA 0012 airfoil using Ansys Fluent considering a range of higher Reynolds numbers, 1×106 ≤ Re ≤ 6×106 and angles of attack, -20⁰ ≤α≤+20⁰. A Matlab programme has been designed to apply a Genetic Algorithm combined with an Artificial Neural Network (ANN) for the purpose of optimization in order to enhance its lift-to-drag ratio. Additionally, conventional machine learning prediction models i.e., AdaBoost, Random Forest, and KNN, were implemented to forecast the efficiency of the airfoil using Python. This analysis provided significant findings, including the lift coefficient, drag coefficient, stalling position, and pressure coefficient. The outcomes of this analysis hold considerable importance in evaluating the efficiency of the airfoil. A notable phenomenon becomes evident at the stalling position of the airfoil, characterized by a sudden alteration in efficiency. The framework using an ANN-based genetic algorithm can produce significant optimal airfoil characteristics with 78% efficiency compared to CFD analysis. The optimization cases were ranked using the TOPSIS-based MCDM tool, revealing that the most favourable combination was observed at an angle of attack of α = -3.1057°. This approach offers notable advantages, including strong robustness, efficient convergence, rapid calculation speed, and significant accuracy in prediction when compared to traditional machine learning methods.

Keywords: Aerodynamic performance, Pitching moment coefficient, Genetic algorithm, TOPSIS, Optuna, Multi-objective optimization