Estimation of aerodynamic coefficients of a non-slender delta wing under ground effect using artificial intelligence techniques

This work presents machine learning techniques to estimate the aerodynamic coefficients of a 40° swept delta wing under the ground effect. For this purpose, three different approaches including feed-forward neural network (FNN), Elman neural network (ENN) and adaptive neuro-fuzzy interference system (ANFIS) have been used. The optimal configuration of these models was compared with each other, and the best accurate prediction model was determined. In the generated machine learning models, the lift CL and drag coefficients CD of the delta wing under the ground proximity of h/c = 0.4 were predicted by using the data of actual CL and CD of the delta wing under the ground proximities of h/c = 1, 0.7, 0.55, 0.25 and 0.1. In FNN, ENN and ANFIS models, the angle of attack α and ground distance h/c were utilized as input parameters, CL and CD as output parameters, separately. Although all three models estimate the CL and CD of the delta wing under h/c = 0.4 with very high accuracy, the ENN method predicts the CL and CD with much higher accuracy than the FNN and ANFIS models. For the estimation of CL, while optimal configuration of ENN resulted in 1.0709% MAPE, 0.00595 RMSE and 0.00504 MAE, the best configurations of FNN and ANFIS end up with the results of 1.172% and 1.1028% MAPE, 0.00786 and 0.0071 RMSE, 0.00593 and 0.0054 MAE, respectively. Thus, results show that the developed FNN, ENN and ANFIS models can be accurately employed to forecast the aerodynamic coefficients of the delta wing under ground effect without the need of for many experimental measurements that causes extra time, labor and experimental costs.