论文部分内容阅读
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。
The research on on-line identification of aircraft icing aerodynamic characteristics can not only be used to analyze the effect of icing on the aerodynamic characteristics of aircraft, but also has important meanings for on-line identification of icing. In recent years, Kalman filter and H_∞ algorithm are widely used in on-line identification of aircraft icing, both of which have the characteristics of high reliability and fast convergence, but they are not enough to evaluate the reliability and accuracy of the algorithm in noisy environment. In this paper, in order to meet the demand of on-line identification of icing, this paper discusses the application of extended Kalman filter and H_∞ algorithm as online icing identification algorithm. First of all, by studying NASA otter icing research aircraft, the extended Kalman filter and H_∞ algorithm are used to identify the pitch direction derivative of the otter icing after icing. The pitch of the aircraft can be quickly estimated by considering the gusting disturbance and the measured noise Direction of the three stability and control derivatives, and identify the results and reference values and found that both algorithms can quickly converge within 2s reference value near, and the state of the filter and the simulation results agree well, indicating The algorithm has high reliability and fast convergence and possesses the ability of on-line detection of icing. On this basis, the simulation data of statistical characteristics of different measurement noise are used to evaluate the effect of measurement noise on the identification accuracy of the two algorithms. It is found that with the increase of the standard deviation of measurement noise, the accuracy of the Extended Kalman Filter identification results is obviously reduced, However, the identification accuracy of H_∞ algorithm changes little, which shows that the accuracy of EKF depends on the accuracy of prior information of noise, while H_∞ algorithm does not depend on the prior information of noise. Even if the data quality is poor, H_∞ algorithm Can get the accuracy of identification results.