基于神经网络的机翼气动参数预测仿真研究

Simulation of Airfoil Plunging Aerodynamic Parameter Prediction Based on Neural Network

飞机机翼颤振是气动弹性力学中最重要的问题之一。为抑制机翼的颤振,首先需要对气动力参数进行计算或预测,而常规的方法存在稳定性差的问题。鉴于飞机的气动力参数的非线性及迟滞效应,提出应用RBF神经网络建模方法,缩短机翼在作正弦运动时的参数计算时间。利用CFD软件和历史计算数据,建立RBF神经网络模型,并对预测结果进行对比研究。结果表明,运用改进方法可预测出给定参数范围之内的升力系数和阻力系数,且误差很小,证明了上述方法在气动力分析预测领域的可行性,可为飞机机翼气动特性优化提供参考。

Flutter is one of the most important problems of Aeroelasticity. To suppress flutter, aerodynamic pa- rameters should be calculated or predicted first. However, lots of problems exist in the traditional methods. In con- sideration of the nonlinearity and retardation effect, the modeling approach of RBF neural network can be used to re- duce the parameter calculating time of the airfoil sinusoidal motion. Using CFD former data we build RBF neural net- work models. The result of the prediction has been studied and contrasted. The lift coefficients and drag coefficients have been predicted in given range, while the error is fairly small. The result proves the feasibility of the usage of this modeling method in the field of analyzing and predicting aerodynamic parameters.