基于深度学习的亚声速机翼升力线理论的改进  

The Improvement of Lifting Line Theory in Subsonic Wings Based on Deep Learning

作　　者：乔诗展 顾文喆 冯雨森 QIAO Shi-zhan;GU Wen-zhe;FENG Yu-sen(School of Astronautics,Northwestern Polytechnical University,Xi'an Shanxi 710072,China;School of Civil Aviation,Northwestern Polytechnical University,Xi'an Shanxi 710072,China;School of Computer Science,Northwestern Polytechnical University,Xi'an Shanxi 710072,China)

机构地区：[1]西北工业大学航天学院,陕西西安710072 [2]西北工业大学民航学院,陕西西安710072 [3]西北工业大学计算机学院,陕西西安710072

出　　处：《计算机仿真》2022年第10期31-33,103,共4页Computer Simulation

摘　　要：针对计算流体力学及升力面理论解算效率低、算力需求大的问题,提出基于深度学习的改进亚声速机翼升力线理论。通过灰度化、阈值化等处理,提取机翼的轮廓。将图像的轮廓通过YOLO系列神经网络使用原创数据集进行角点检测,获得机翼控制点坐标。分别使用控制点坐标及机翼轮廓通过升力线和升力面理论计算机翼的气动参数,并使用多层感知机修正升力线与升力面的误差,获得精度较高的改进升力线模型。结果表明采用上述算法可有效提升升力线理论的准确率,并保证解算效率。Aiming at the low efficiency and the large demand for the computing power of computational fluid dynamics and lifting surface theory, an improved subsonic wing lifting line theory based on deep learning is proposed. The contour of the wing was extracted through grayscale and thresholding. The corner points of the image were detected by the YOLO series neural network using the original data set to obtain the wing control point coordinates. The control point coordinates and the wing profile were then used to calculate the aerodynamic parameters of the wing through the lifting line and lifting surface theory respectively, and the multi-layer perceptron was used to correct the error between the lifting line and the lifting surface to obtain an improved lifting line model with higher accuracy. The results show that the use of this algorithm can effectively improve the accuracy of the lift line theory and ensure the efficiency of the solution.

关 键 词：深度学习 气动分析 亚声速机翼 升力线理论 升力面理论 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计]