類神經網路技術在齊平式大氣資料量測系統上之應用

An Application of Neural Network for Flush Air Data Sensing System

張志強(Chih-Chiang Chang) ； 張谷川(Ku-Chuan Chang)

《中國航空太空學會學刊》 35卷2期 (2003/06) Pp. 159-162

https://doi.org/10.6124/TAASRC.200306_35(2).05

摘要

本文目的是以數值模擬方式評估並建立類神經連網路FADS (Flush Air Data Sensing System)系統，以準確量測飛行體之攻角與側滑角。其方法是藉著以鼻錐流場數值模擬取得不同飛行條件下之鼻錐附近靜壓孔壓力值，配合局部攻角與側滑角之氣動力模式，建立以爲赫數、局部攻角（或側滑角）與攻角是最（△α與△β）之校正模式資料庫，最後以類神經網路求得校正模式資料庫之最佳近似，完成攻角與側滑角之估算。本系統經由模擬評估已經找出有效靜壓孔位置，而且可以確定雖然校正模式資料庫具有非線性及非唯一之特性，類神經網路FADS系統仍然可以求出最佳近似，因此在未來進一步進行地面與飛行試驗以建立校正模式資料庫時，應可以對試驗數據的雜訊與紊度有較佳之耐受度。

關鍵字

齊平式大氣資料感測系統 ； 類神經網路

並列摘要

The flush air data sensing (FADS) system concept was developed to circumvent many of the difficulties with intrusive air data systems. Rohloff et al. have proposed a neural network (NN) approach to develop a FADS estimation algorithm in order to overcome the inherent stability problem, hardship in both calibration and implementation of the existing FADS systems.
Assuming that free stream Mach number and flight height are known, a NN-FADS scheme is designed in this paper for estimating free stream angle of attack and side slip angle based on the measured pressure information on a vehicle's radome (eight ports). The authors apply CFD to creating a group of pressure data for different flight conditions at these positions. Since a pressure model is necessary to relate the pressure measurements to air data quantities (produced by CFD in present study), the model that associates Local angle of attack and local angle of side slip to local pressure is formed according to potential flow and modified Newtonian flow theory. Then, a neural network is trained to correlate this model with the pressure database, and establishes a correction model that transfers local angle of attack (or local angle of side slip) to free stream angle of attack (or free stream angle of side slip). Further extension of this NN-FADS system is still under investigation.

並列關鍵字

Flush air data sensing system ； Neural network