支持向量回归机在飞机气动力建模中的应用  

Application of Support Vector Regression for Aerodynamic Modeling

作　　者：吴辰[1] 姚宏[2] 彭兴钊[1] 杜军[1] 

机构地区：[1]空军工程大学航空航天工程学院,陕西西安710038 [2]空军工程大学理学院,陕西西安710051

出　　处：《计算机仿真》2013年第10期128-132,共5页Computer Simulation

基　　金：航空科学基金(2011139601)

摘　　要：研究准确的气动力模型是研究飞机机动特性与控制的基础。针对气动参数辨识精度和速度难以大幅度提高的问题,提出了一种支持向量回归机的飞机气动力建模方法。利用其良好的泛化能力,在小样本条件下建立了准确的气动力模型,并通过人工鱼群算法优化其中三个待定参数C,ε,σ,增强了全局最优性。在飞机纵向通道气动力建模仿真中,与极大似然法、神经网络进行了对比研究。仿真结果表明:经人工鱼群算法优化的支持向量回归机,克服了极大似然法、神经网络中的局部最优和小样本学习能力不足等问题,同时不需要确定初值,证明是一种有效的气动力准确建模方法。Accurate aerodynamic models are necessary for researches on simulation and control of maneuvers. To solve the problem that it is difficult to improve the accuracy and speed of aerodynamic parameter identification, a method of aerodynamic modeling based on support vector regression （SVR） was introduced in this paper. By using of the good generalization ability of SVR, an accurate aerodynamic model was built in the condition of small sample, and three undetermined parameters of SVR were optimized by artificial fish swarm algorithm （AFSA）, in order to im- prove global optimality of it. The comparisons with maximum likelihood estimation （MLE） and neural network （NN） were made in the aerodynamic modeling simulation of vertical channel. The simulation results show that the SVR op- timized by AFSA overcomes the local optimality and deficient small sample learning ability compared to MLE and NN, besides it doesn＇t need to determine the initial value. Therefore, the method is proved to be effective for aerody- namic modeling.

关 键 词：支持向量回归机 气动力建模 核函数 在线辨识 人工鱼群算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]