A Numerical Study for the Prediction of Unmanned Aerial Vehicle Aerodynamic Performance Based on Dihedral and Tip-Twist Angles of the Wing

Abstract

AbstractUAV technology has developed so rapidly to help humans solve problems around them. One of the important criteria in the development of a UAV is its aerodynamic performance. The aerodynamic performance of a UAV is largely determined by the geometry of its wings. This study focuses on variations in wing geometry, especially dihedral and tip-twist to predict the aerodynamic performance of the UAV, each with \(\max \left( {C_L /C_D } \right)\) and \(\left. {C_D } \right|_{\alpha = 0}\). The Artificial Neural Network (ANN) method is used in this study to get predictions of each aerodynamic performance. ANN was chosen because of its superiority in approaching very complex relations between several variables. ANN network structure has up to two hidden layers while each hidden layer has 2 to 10 neurons selected in this research to get the smallest Mean Square Error value. Mean Square Error (MSE) is the difference between the aerodynamic performance target value and the predicted value. Prediction value is obtained from the training results of a network configuration based on predictor and target values as network input and output. Factors analyzed, namely dihedral angle and tip-twist, described their relationship to each of the \(\max \left( {C_L /C_D } \right)\) and \(\left. {C_D } \right|_{\alpha = 0}\) using a network structure of 2-5-7-1 and 2-4-9-1. Both networks yield the smallest MSe \(8.4757 \times 10^{ - 7}\) and \(1.952 \times 10^{ - 8}\) respectfully. The graph of the relationship of factors to changes in the value of \(\max \left( {C_L /C_D } \right)\) and \(\left. {C_D } \right|_{\alpha = 0}\) shows that the tip-twist angle gives a greater contribution than the dihedral. Comparison between prediction and validation simulation shows the relative error smaller than 5%.KeywordsWing geometryDihedralTip-twistAerodynamic performanceNeural networkMSE