神经网络在风洞流场马赫数辨识中的应用研究

风洞流场的多维性、复杂性以及马赫数的不可直接测量、马赫数控制一直是风洞控制的难点和重点。笔者在研究FL26Y风洞流场特性的基础上,应用神经网络技术,研究马赫数在线辨识问题,为高精度风洞流场马赫数控制提供技术支持。首先介绍神经网络马赫数辨识原理,然后介绍神经网络拓扑结构的设计,并构造神经网络的马赫数辨识模型(NNI)。最后通过软件实现及仿真研究动态数据优化、软件滤波以及动量系数对网络学习性能的影响,进一步验证神经网络辨识器在实时性、自适应性、以及辨识精度等方面的优越性。

Research of application of neural network in Mach number identification of wind tunnel flow field

Flow field of wind tunnel is multidimensional and complex. Mach number cannot be measured directly. Because of these, Mach number control is always the emphasis and difficult question in wind tunnel control system. Based on studying the characteristics of FL26Y wind-tunnel, on-line Mach number identification using neural network is studied, the purpose of that is to help to improve the precise of Mach number control system of wind-tunnel. First the principle of Ma identification based on neural network (NNI) is introduced. Then the topology structure of NNI is designed, and the model of NNI (Neural Network Identification) is developed. At last we use simulation experimentation to research a set of problem, such as the influence of dynamic data optimization, the software filter and the influence of momentum to the parameter of NN. All of these indicate the advantage of NNI in real-time, adaptation and precise.