基于神经网络的非线性气动弹性系统辨识

IDENTIFICATION OF NONLINEAR AEROELASTIC SYSTEMS BASED ON NEURAL NETWORK

由于气动弹性系统的非线性和不确定性的存在,传统的辨识方法在工程中难以满足。针对这种情况提出了一种模糊小波神经网络(FWNN)辨识方法。首先,采用区间2型模糊逻辑系统和小波神经网络结合构建FWNN网络结构,能够较好地逼近具有不确定性的非线性AE系统;然后,考虑到辨识的快速性和准确性,系统采用一组模糊IF-THEN规则,对模糊后件采用单隐层小波神经网络结构;参数学习采用基于Lyapunov稳定性的滑模学习算法,保证系统存在参数不确定的情况下,辨识误差能更快地收敛。最后,对结构非线性二元翼段进行仿真研究,验证了该模型的有效性。

Because of the nonlinearity and uncertainty of the aeroelastic system, the traditional identification methodis difficult to meet in engineering. In this paper, a fuzzy wavelet neural network （ FWNN） identification method is proposed. Firstly, the FWNN network is constructed by the combination of interval 2 fuzzy logic system and wavelet neural network, which can approach the nonlinear AE system with uncertainties. Then, considering the fastness and accuracy of identification, the system adopts a set of fuzzy IF-THEN rules, and a single hidden layer wavelet neural network structure is used for the fuzzy consequent parts. Parameter learning is based on the Lyapunov stability of the sliding mode learning algorithm to ensure the existence of the parameters of the system uncertainty, the identification error can be faster convergence. Finally, the simulation of the nonlinear binary wing section is carried out to verify the effectiveness of the model.