English

应用科学学报 ›› 2016, Vol. 34 ›› Issue (4): 469-474.doi: 10.3969/j.issn.0255-8297.2016.04.012

• 控制与系统 • 上一篇    

基于SVM的直升机飞行状态识别

熊邦书1, 刘雨1, 莫燕1, 黄建萍2, 李新民2   

  1. 1. 南昌航空大学图像处理与模式识别江西省重点实验室, 南昌 330063;
    2. 中国直升机设计研究院直升机旋翼动力学国防科技重点实验室, 江西 景德镇 333001
  • 收稿日期:2015-11-10 修回日期:2016-02-02 出版日期:2016-07-30 发布日期:2016-07-30
  • 作者简介:熊邦书,教授,研究方向:模式识别、智能信号处理和图形处理,E-mail:xiongbs@126.com
  • 基金资助:

    国家自然科学基金（No.61462063）；航空科学基金（No.20135756010）；江西省高等学校科技落地项目基金（No.KJLD13058）资助

Recognition of Helicopter Flight Condition Based onSupport Vector Machine

XIONG Bang-shu1, LIU Yu1, MO Yan1, HUANG Jian-ping2, LI Xin-min2   

  1. 1. Key Laboratory of Image Processing and Pattern Recognition, of Jiangxi ProvinceNanchang Hangkong University, Nanchang 330063, China;
    2. National Key Laboratory of Rotorcraft Aerodynamics, China Helicopter Research andDevelopment Institute, Jingdezhen 333001, Jiangxi Province, China
  • Received:2015-11-10 Revised:2016-02-02 Online:2016-07-30 Published:2016-07-30

PDF

514

被引次数

3

摘要：

针对直升机飞行状态识别训练样本数据少而导致识别率不高的问题，提出一种基于支持向量机(support vector machine，SVM)的直升机飞行状态识别方法.首先利用限幅、去野点和均值滤波对飞行数据进行去噪，用最小二乘法对飞行数据进行直线拟合获取变化率，并根据线性相关性提取状态特征参数，以减少数据冗余；然后根据特征参数将飞行状态分为10小类，对每一小类进行SVM分类器设计以提高识别效率；最后利用训练样本训练每个SVM分类器，用训练好的SVM分类器识别直升机全起落飞行状态.通过某型直升机实飞数据进行飞行状态识别实验，并将所提出的方法与RBF神经网络法进行对比，所得结果表明该方法在小样本情况下的识别率有明显提高，可为直升机故障诊断和寿命预测提供依据.

关键词: 最小二乘法, 飞行状态识别, 支持向量机, 线性相关性, 小样本

Abstract:

To solve the problem of low recognition rate due to insufficient training samples, a flight condition recognition method based on SVM is proposed. The flight data first undergo denoising by clipping, outlier removal, and averaging. The changing rate of flight data is obtained with least square line fitting. Redundancy in the data is reduced based on the characteristic parameters extracted using linear correlation. The flight condition is classified into ten categories according to the characteristic parameters. An SVM classifier is designed for each category to improve identification efficiency. Finally, every SVM classifier is trained with training samples, and all flight conditions of the helicopter are identified by the trained SVM classifier. Actual flight experiments show that, compared with the RBF neural network method, the proposed method can improve performance under a small sample condition. It provides a reference for helicopter fault diagnoses and life prediction.

Key words: support vector machine (SVM), flight condition recognition, least squaremethod, linear dependence, small sample

中图分类号: