### 基于H∞算法的飞机机翼结冰气动参数辨识

1. 1. 中国空气动力研究与发展中心 空气动力学国家重点实验室, 绵阳 621000;
2. 中国空气动力研究与发展中心 计算空气动力研究所, 绵阳 621000
• 收稿日期:2017-07-24 修回日期:2017-10-10 出版日期:2018-03-15 发布日期:2017-10-10
• 通讯作者: 丁娣,E-mail:dingdi1981@hotmail.com E-mail:dingdi1981@hotmail.com
• 基金资助:
国家"973"计划（2015CB755800）

### Aerodynamic parameter identification for aircraft wing icing using H∞ method

DING Di1,2, CHE Jing1,2, QIAN Weiqi1,2, WANG Qing1,2

1. 1. State Key Laboratory of Aerodynamics, China Aerodynamics Research and Development Center, Mianyang 621000, China;
2. Computational Aerodynamics Institute, China Aerodynamics Research and Development Center, Mianyang 621000, China
• Received:2017-07-24 Revised:2017-10-10 Online:2018-03-15 Published:2017-10-10
• Supported by:
National Basic Research Program of China (2015CB755800)

Abstract: H parameter identification for inflight icing detection is discussed in this paper for the development of the de-icing and anti-icing system of a large icing research prototype. First, the H method parameters are tuned and chosen.Then the method is evaluated by the airplane's longitudinal simulation data with measurement noises. The identification results show that the method can track the time-varying aerodynamic parameters in the ice accretion process, and that the maximum normalized Root Mean Square(RMS) error 11% indicates high accuracy of identification.The 81 different ice accretion processes are then identified by the H method.The results show that when the ice accretion process changes slowly and lasts long the identification accuracy is relatively poor, and that when the ice accretion time is between 100-300 s the accuracy is relatively high. The accuracy of H algorithm in the different standard deviations of measurement noises is analyzed by Monte Carlo simulation.The error and delay statistic characteristics of the three longitudinal aerodynamic derivatives show that the identification accuracy of derivatives of lift and pitching moment to angle of attack is relatively high as their mean normalized RMS errors are 1.8% and 4% respectively, and their mean delays are 3 s and 9.5 s respectively.