Estimating Aerodynamic Coefficients from Uncertain Data of D-SEND Aircraft with Gaussian Process Regression

Abstract

When simulating airflow we assume an ideal situation, however, flight test data includes measurement noise when actually conducted. Therefore, it is difficult to compare simulation data with flight test data without considering uncertainty. First, we applied the Noisy Input Gaussian Process (NIGP), which can utilize uncertain inputs to estimate aerodynamic coefficients with confidence intervals to an aircraft's simulation data. This enabled us to verify the effectiveness of NIGP. We then applied NIGP to the aircraft's real flight data and compared them with aerodynamic tables based on wind tunnel testing and CFD. We found that the lift coefficient estimated using the flight test data did not contradict that obtained using simulation data, while the drag coefficient estimated using the flight test data was smaller than that obtained using the simulation data.