端到端的飞机尾涡特征参数估计  

End-to-End Aircraft Wake Feature Parameter Estimation

作　　者：潘卫军[1] 冷元飞 蒋倩兰 吴天祎 PAN Weijun;LENG Yuanfei;JIANG Qianlan;WU Tianyi(College of Air Traffic Management,China Civil Aviation Flight Academy,Guanghan 618307)

机构地区：[1]中国民航飞行学院空中交通管理学院,广汉618307

出　　处：《舰船电子工程》2022年第9期87-91,共5页Ship Electronic Engineering

基　　金：国家重点研发计划(编号:2018YFC0809500,2016YFB0502403);民航局安全能力建设项目(编号:AQ20200019);国家自然科学基金联合基金重点项目(编号:U1733203)资助。

摘　　要：为了实现繁忙机场的动态尾流间隔调整,提出了一种基于深度学习网络和相干光激光雷达技术的端到端飞机尾流特征参数估计模型。首先,基于深度可分离卷积块构建轻量化主干网络以提取径向速度场的潜在基础特征。然后针对飞机尾流左右涡旋,构建相应的特征参数预测模块。最后,使用经半自动RV方法处理的风场测量用于模型参数输出的监督学习。团队在双流国际机场安置激光雷达扫描飞机航迹线区域风场构建数据集对模型训练、验证和测试。实验结果表明,定制的参数估计模型预测环量误差低至8.42m2/s、位置定位误差低至1.59m,模型能够有效辅助管制员进行空中交通智能化管理。In order to realize dynamic wake interval adjustment in busy airports,an end-to-end aircraft wake feature parameter estimation model based on deep learning network and coherent optical LIDAR technology is proposed.First,a lightweight backbone network is constructed based on depthwise separable convolutional blocks to extract the underlying underlying features of the radial velocity field.Then,for the left and right vortices of the aircraft wake,the corresponding feature parameter prediction module is constructed.Finally,wind field measurements processed by the semi-automatic RV method are used for supervised learning of model parameter outputs.The team installed LIDAR at shuangliu international airport to scan the wind field of the aircraft track line area to build a data set for model training,validation and testing.The experimental results show that the customized parameter estimation model predicts the circulation error as low as 8.42 m~2/s and the position positioning error as low as 1.59m.The model can effectively assist the controller in intelligent management of air traffic.

关 键 词：激光雷达 尾涡 特征参数 尾流探测 

分 类 号：U8[交通运输工程]