Preliminary results from a subsonic high-angle-of-attack flush airdata sensing (HI-FADS) system - Design, calibration, algorithm development, and flight test evaluation

Kriging-algorithm-based Aerodynamic Model for Flush Airdata System

“…The Four-port model (2,3,4,5) & (6,7,8,9) According to the results of the three-port model, it is impossible to accurately predict both angle of attack and angle of slide for this blended wing body configuration, hence two four-port model were studied and compared. Model 1 is based on Ports 2,3,4,5 and Model 2 is based on Ports 6,7,8,9. In table 7-10, the prediction errors were compared.…”

“…Model 1 is based on Ports 2,3,4,5 and Model 2 is based on Ports 6,7,8,9. In table 7-10, the prediction errors were compared. Both the Kriging method and the neural network method illustrate high accuracy in the prediction of attack angle, slide angle and Mach number, however, Model 2 beats Model 1 in all predictions in both methods.…”

“…The hyperparameters are selected according to the maxima of the likelihood of y, which is obtained by maximizing the concentrated log likelihood function [14] [15]given by (6) The estimated process variance is given by (7) The local behavior of the random field dominates the spatial interpolations, hence it is critical to select proper models for the covariance structures wherein at least one member reflects the actual local behavior of the spatially varying quantity under study [3]. Proper tuning method is important, since inadequate tuning of the hyperparameters can lead to a poor representation of the true function, and result in a wrong prediction.…”

“…This flush airdata system is designed for the flight control at the high angle of attack (which is called Hi-FADS), since at a high angle of attack maneuvering, the lift is very sensitive to the change of angle of attack, hence any little errors in prediction of angle of attack may cause fatal accident. Though this method satisfied the requirement of the Hi-FADS, its disadvantage is obvious; because each prediction requires several iterative steps, particularly, the iterative process may fail to converge, which results in a prediction failure [6] [7].…”

A Modified Triples Algorithm for Flush Air Data Systems That Allows a Variety of Pressure Port Configurations

“…5 In order to overcome these disadvantages, the Triples Algorithm was developed. 6,7 The Triples Algorithm uses a potential flow pressure model, a cruciform pressure port configuration with the vertical ports on the line of symmetry, and a strategic set of differences to eliminate the pressures from the state vector, allowing a direct solution of the angle of attack and the sideslip angle. The potential flow model is then used to compute the freestream pressure, total pressure, and the Mach number.…”