The Research Of The Flush Airdata Sensing (FADS) System And It's Neural Network Algorithm

The flush air data sensing (FADS) system uses a matrix of pressure transducer on the vehicle nose to estimate air data parameters. The first part of this paper presents the principle,realization and performance of FADS system . Preliminary the vehicle's architecture, estimating algorithms and a mathematical analysis of the system stability are discussed,and then the calibration method for the position error coefficient is introduced and the flow correction terms for the angle of attack and angle of sideslip are derived from wind tunnel data. Statistical accuracy of the calibration is evaluated by comparing the wind tunnel reference conditions to the air data parameter estimated. This comparison is accomplished by applying the calibrated FADS algorithm to the sensed wind tunnel pressures and the resulting is satisfactory. The second part of the paper is focused on the research of neural network flush air data sensing system. The relationship between the pressure measurements and the air data is a complex non-linear function that is not easily described with simple aerodynamic models. Back-propagation neural network was used to substitute the aerodynamic models to describe the functionality between measured and estimated parameters. Firstly the technique of producing and compiling training data was presented in detail and the Euclidean distance was found to be the most effective way for the compilation of the training set. The architecture and a fast algorithm of the neural networks air also discussed in this part . As a result, the neural network models were shown to be stable across the entire relevant range of new input data, including both interpolation within and extrapolation outside the original domain of the training data. The resulting system of models were shown to be accurate and robust to noise in the input signals.