非稳定动态过程非定常气动力建模

现有的大迎角非定常气动力建模方法，通常是以一个或多个频率的稳定振动试验数据来预测稳定滞环。然而，飞机快速机动如过失速机动的过程，不可能是持续的稳定振动，而是一个非稳定的动态过程。因此，这个过程中的气动力不会达到稳定滞环，而是始终处于进入滞环的初始非稳定过程中。基于振动理论分析得出，非定常气动力的动态响应过程存在非稳定和稳定两个阶段，传统建模方法着眼于稳定阶段，而飞机的真实机动过程在非稳定阶段。设计了一种适于非线性系统辨识的激励输入，并以最小二乘支持向量机（LS-SVM）方法为例，实现了在大迎角区幅值和频率范围内任意运动的非定常气动力建模。模型训练完成后，用来预测某机翼在不同基准状态下大迎角范围内做俯仰运动时的升力系数、阻力系数和俯仰力矩系数。结果表明，不仅稳定滞环实现了准确预测，进入滞环的初始非稳定过程也得到了准确预测；此外，基准状态对气动力在初始非稳定过程中的特性存在明显影响。进一步的验证还表明，基于稳定滞环数据只能预测到稳定滞环，无法预测进入滞环的非稳定过程。

Unsteady aerodynamic modeling of unstable dynamic process

Current unsteady aerodynamic modeling methods at high angle of attack usually use stable vibration test data at multiple frequencies to predict stable hysteresis loop. However, the rapid maneuvering process of aircraft, such as post-stall maneuver, cannot be a constant and stable vibration, but an unstable dynamic process. Therefore, the aerodynamics would not reach a stable hysteresis loop, but would always be in the initial unstable process of entering the hysteresis loop. The vibration theory analysis shows that the dynamic response process of unstable aerodynamics has the unstable and stable stages. The traditional modeling method focuses on the stable stage, while the actual maneuvering process of aircraft is in the unstable stage. Based on the Least Squares Support Vector Machine (LS-SVM), an excitation input suitable for nonlinear system identification is introduced to model unsteady aerodynamic forces of any motion in the amplitude and frequency ranges at high angle of attack. After completing the model training, the method is applied to predict the lift coefficient, drag coefficient, and pitching moment coefficient of a wing at high angle of attack with different reference states in pitching motion. The results show that not only the stable hysteresis loop is accurately predicted, but also the initial unstable process of entering the hysteresis loop is accurately predicted. In addition, the results also show that the reference state has significant influence on the characteristics of aerodynamics in the initial process. Further validation also shows that modeling based on the stable hysteresis loop data can only predict the stable hysteresis loop, and cannot predict the unstable process of entering the hysteresis loop.