The Algorithm And Application Research For Flush Air-Data Sensing System Based On The Surrogate Model

As the requirements of high-speed, mobility, agility and stealth characteristics for the modern aircraft, the FADS system plays an important role in the flight control system. In this thesis, the CFD software was used to simulate the flow over a flying wing aircraft, hence plenty of sample data was obtained. The angle of attack, angle of slide and Mach number was predicted by using Kr iging algor ithm and BP neural network algorithm. By compareing with the actual result, the Kr iging algorithm shows slightly higher precision than the BP neural network when the sample is small. It is also found that precision increases as the number of pressure points increase. For some aircrafts, it is difficult to mount a sensor on the forebody or nose, therefore the locations for the pressure sensors on the wing as well as fuselage were studied in depth. Form the comparison, it is found that slighty loss in presicion for these two layout. In addition, for the case that sensors moued on the wing, Ma(Mach number), α(angle of attack) and β(angle of sideslip) are quite sensitive to the sensors which are close to the fuselage and the leading edge; for the case that sensors mounted on the fuselage, β is not sensitive for all the sensors, while Ma and α is sensitive to the sensors near the leading edge. To construct the aerodynamic model for the FADS system, a lot of sample data need to be obtained in advance. As the pressure coefficient has less change at different attitude for same state, therefore the number of sample data can be decreased by simulat ing only once at an att itude. Based on these datum, the aerodynamic model for FADS system can be built, which the inputs are pressure coefficients, and the output is air-data. By using suchmodel, the static pressure, dynamic pressure, Ma, α and β can be predicted with tiny error.