Hide/Show Apps

Experimental design and statistical modeling for efficient wind tunnel testing

Savaş, Özgün
Wind tunnel testing is an essential procedure to measure the aerodynamic forces and moments on an air vehicle. In this thesis, a method is presented to perform such test in an efficient way. First, an experimental design process is carried out before the testing in order to cover the flight regime as fine as possible with the least possible number of tests. After the determination of the test matrix and conducting the wind-on tests, the modeling of the output data is the next step. On that matter, since the vehicle of interest is an agile missile which may fly at high angles of attack and in a relatively broader Mach regime, the aerodynamic data is highly nonlinear. Therefore, two nonlinear modeling techniques are presented and compared with each other. First technique is the widely used Artificial Neural Networks (ANN) and the second one is relatively lesser known but a powerful modeling algorithm Multivariate Adaptive Regression Splines (MARS). After the data is modeled using both approaches, their statistical metrics are compared and the models are integrated into a high fidelity 6DoF equations of motion model. Since all this effort is to improve the simulation accuracy and performance, 6DoF simulations are performed using several flight scenarios that cover the flight regime of the missile. Afterwards, obtained trajectories and flight parameters are compared. Finally, various flight conditions are produced and the models are evaluated in batch mode to see their performances in terms of computational speed.