摘要
结冰会对飞行安全造成极大的威胁,非常有必要建立结冰的预测模型。飞机结冰预测模型是智能结冰系统的重要组成部分之一。预测模型以冰形为研究对象。先用曲线替代翼型前缘,把附着于翼型上的冰形分离出来,再通过保角映射得到独立于翼型的冰形数学模型。接着以傅里叶级数形式展开,得到冰形的傅里叶系数。然后建立神经网络预测模型,以影响飞机结冰的大气和飞行条件参数作为神经网络的输入,以冰形的傅里叶系数作为神经网络的输出。分别以传统的五项参数(液态水含量、平均水微滴直径、结冰时间、温度和飞行速度)和新增加两项参数(相对湿度和攻角)建立网络进行训练仿真。仿真结果表明,该方法能很好地预测冰形,增加两项参数后的神经网络模型预测效果更好。
Ice accretion will be a great threat to the flight safety, so it is necessary to build the aircraft ice prediction model, which is an important part of the Smart Icing System. The ice shape is main study object of the prediction model. The airfoil leading edge was approximated by curve, so the ice shape that attached to airfoil was separated, then the mathematical model of ice shape that is independent of the airfoil was got through the conformal mapping. After that it was expanded into Fourier series, so the Fourier coefficients of the ice shape were got. Then the neural network prediction model was built. The input of the neural network was the atmospheric and flight parameters that have effects on icing. The output of the neural network was the Fourier coefficients of the ice shape. Two types of network for training and simulation were established： one with five traditional parameters （liquid water content, median volumetric diameter, freezing time, temperature and flight speed） and the other added two additionalparameters （relative humidity and angle of attack）. The simulation results show that this method can predict the ice shape well, and the prediction model with two additional parameters can predict the ice shape better.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2014年第1期221-224,229,共5页
Journal of System Simulation
基金
民用大型客机总体气动系统设计平台研制(产-36-专项-1)
关键词
飞机结冰
冰形预测
傅里叶系数
神经网络
aircraft icing
ice shape prediction
Fourier coefficient
neural network
