摘要
目前风洞试验仅为民用飞机飞行性能提供有限数据。全飞行包线的技术支持对于民机飞行试验十分重要,需要采用数学建模和参数辨识的方法。选择合适的机器学习算法是参数辨识中最为关键的一步。支持向量机(SVM)采用结构风险最小化原理,尤其适用于小样本情形。根据A320非巡航起降阶段的几组真实数据,以及全机气动力估算的结果,使用最小二乘支持向量机建立预测模型。随后采用粒子群算法优化模型参数从而提升泛化能力。由此实现民机飞行包线的气动性能整体建模与辨识。与Ma=0.78时的实验数据相比较,PSO-LSSVM模型的预测结果吻合,是一种有效的气动数学建模方法。
The data provided by wind tunnel tests by far are not nearly sufficient when studying the aero-dynamic performances of civil airplanes, which accordingly calls for technical support by means of mathe-matical modeling and parameter identification. Support vector machine provides effective tool concerning small sample data,as it embodies structural risk minimization principle. Taking A320 for example,a pre-diction model is built up employing least square support vector machine ( LS- SVM) method, according to its whole plane aerodynamic parameters estimated by empirical formulae and several real test data at non -cruising phases. Particle Swarm Optimization (PSO) method is later applied to improve the model, which enables it better generalization properties. Thus,the overall modeling and identification of aerody-namic parameters come into being. Comparing with the test data when M a =0. 78, PSO - LSSVM model shows good predicting performance, and can be adopted as a practical way of aerodynamic calibration.
出处
《航空计算技术》
2017年第2期68-71,75,共5页
Aeronautical Computing Technique