In this paper, we propose a data-driven framework to control morphing airfoils in the subsonic flight regime, considering high Reynolds numbers with an efficient and safe way to reach a shape with improved values of the aerodynamic coefficients. The online solution is based on a data-driven controller combined with a surrogate model and a multi-gradient descent algorithm. Without full knowledge of the aerodynamic parameters (lift, drag, and pitching moment coefficients), the learning framework searches for an airfoil shape that minimizes a metric of performance associated to drag, lift, and pitching moment coefficients. The solution uses online data to improve the accuracy of the predictions of the aerodynamic coefficients provided by the surrogate model along the trajectory. The optimization framework focuses on subtle airfoil deformations to assure a smooth trajectory between the initial and the final shape. Finally, the efficacy and the robustness of our proposed solution was shown in numerical examples, resulting in a significant reduction in the prediction error.

中文翻译：

---

高雷诺数流中变形翼型的数据驱动控制器和多梯度搜索算法

在本文中，我们提出了一种数据驱动的框架来控制亚音速飞行状态下的变形翼型，考虑高雷诺数，并以有效且安全的方式达到具有改进的空气动力系数值的形状。在线解决方案基于数据驱动控制器以及代理模型和多梯度下降算法。在不完全了解空气动力学参数（升力、阻力和俯仰力矩系数）的情况下，学习框架会搜索最小化与阻力、升力和俯仰力矩系数相关的性能指标的机翼形状。该解决方案使用在线数据来提高替代模型沿轨迹提供的空气动力学系数的预测准确性。优化框架侧重于细微的机翼变形，以确保初始形状和最终形状之间的平滑轨迹。最后，我们提出的解决方案的有效性和鲁棒性在数值示例中得到了证明，从而显着降低了预测误差。