Multiobjective Optimization of a Single Slotted Flap Using Artificial Neural Network and Metaheuristic Algorithms

This research aims to improve the aerodynamic performance of the high-lift device used during landing in a general aviation aircraft. Our approach involved conducting a multiobjective aerodynamic optimization of the single slotted flap on the aircraft's wing. The objective was to simultaneously maximize the lift and drag coefficients while minimizing the moment coefficient. To achieve this, we performed several simulations with varying gap and overlap values, resulting in a primary data set. The simulations were conducted using the k-epsilon turbulence model to simulate the incompressible turbulent flow around the NACA-23012 airfoil at a Reynolds number of Rec=3.6x106. However, due to the extensive computational resources and time required for solving, analyzing the data posed significant challenges. To address this, we utilized a surrogate model to generate data, reducing both cost and time. During the machine learning process, an artificial neural network (ANN) was trained on the initial data set. To solve this type of problem, we employed the multiobjective genetic algorithm (MOGA) as the most suitable algorithm. By utilizing MOGA, we were able to determine the optimal configuration for the single slotted flap. The training of computational fluid dynamics data in an ANN significantly reduced computational time and cost, enabling the generation of a continuous response surface in less than one minute. Additionally, the multiobjective genetic algorithm efficiently produced the Pareto front in just 45 s. Finally, based on the design requirements of the proposed configurations, the best option was selected. Subsequently, an aerodynamic analysis of the final configuration was conducted to evaluate its performance.