飞机结冰在线辨识方法研究

开展飞机结冰气动特性在线辨识研究,不仅可以用于分析结冰对飞机气动特性的影响,而且对于飞机结冰在线识别具有重要的意义。近年来卡尔曼滤波和 H ∞算法在飞机结冰在线辨识中应用较多,二者均具有可靠性高、收敛快等特点,但对于噪声环境下算法的可靠性和精度评估还不够充分。本文针对飞机结冰在线辨识需求,探讨了扩展卡尔曼滤波和 H ∞算法作为结冰在线辨识算法的应用。首先通过 NASA 双水獭结冰研究飞机算例,利用扩展卡尔曼滤波和 H ∞算法,辨识双水獭飞机结冰后的俯仰方向导数,通过考虑阵风扰动和测量噪声后的仿真数据快速估计该飞机俯仰方向上的三个稳定和控制导数,并将辨识结果与参考值对比,发现两种算法均能在2s 之内快速收敛到参考值附近,且滤波得到的状态量与仿真数据吻合较好,说明算法可靠性高且收敛快,具备飞机结冰在线探测的能力。在此基础上利用不同测量噪声统计特性的仿真数据,评估测量噪声对两种算法辨识精度的影响,经分析发现随着测量噪声标准差取值增大,扩展卡尔曼滤波辨识结果精度明显降低,而 H ∞算法的辨识精度变化较小,说明扩展卡尔曼滤波辨识精度依赖于噪声先验信息的准确性,而 H ∞算法不依赖于噪声先验信息,即使数据质量较差,H ∞算法也能得到精度相当的辨识结果。

Research on online identification methods for aircraft icing

Icing and related dynamics play important roles in aviation safety.The research on online estimation of aerodynamic parameters of icing aircrafts not only helps to reveal the influence on aerodynamic characteristics of icing,but also improves the capability of icing online identification.Lately Kalman filter and H ∞ algorithm are apply applied on online identification of aircraft icing for their high reliability and fast convergence,even though the assessment of their characteristics under noisy environment are not sufficiently understood.This paper discusses the applications of the extended Kalman filter (EKF)and H ∞ algorithm in the aircraft icing online identification.The methods are validated and evaluated based on the NASA Twin Otter icing research airplane. The three stability and control derivatives in pitch direction are quickly estimated by both two methods from simulation data with gust disturbance and measurement noise.Both two methods estimate these parameters with high precision in 2 seconds,comparing with the reference values.And the state variables filtered by the two methods are consistent with the simulated dynamic process.The results validate the constringency and effectivity of the two methods,together with their potential abilities of applying in the aircraft icing online detection. The identification accuracies of these two methods are evaluated with the simulation data under different measurement noises.With increasing standard deviation in the noise applied in the simulation,the accuracy of EKF method may deteriorate considerably while the accuracy of H ∞algorithm remains the same level.These results lie on the fact that for EKF the precision rely strongly on precise prior information.On the other hand,H ∞ exhibits better performance under the same circumstances.Its precision is not affected sensitively by bigger standard deviation of measurement noises,which shows more potential gains for online application without high quality prior information.