旋翼/机身气动干扰的神经网络模型

Neural Network Model of Aerodynamic Interaction Effects of Rotor Wake on Fuselage for Helicopter

依据直升机空气动力学理论建立起来的旋翼 /机身气动干扰模型 ,至今尚不能达到满意的准确度 ,而且它的计算工作量相当大。文中采用多层前向通道神经网络建立旋翼 /机身气动干扰模型 ,探索一种新的建模方法。在试验结果的基础上 ,对机身的空气动力随前进比和桨盘载荷变化情况进行了分析 ,然后提出了一个二输入 /三输出的旋翼 /机身气动干扰神经网络模型。网络训练样本直接来源于试验数据。训练好的干扰模型清楚地反映了机身阻力与前进比 μ和桨盘载荷p的变化都有着密切关系 ;对于机身升力和俯仰力矩 ,桨盘载荷 p的变化起着主导作用。给定状态点的实际测量与采用神经网络模型计算结果的比较 ,进一步验证了旋翼 /机身气动干扰神经网络模型的合理性与有效性。

The existing model of aerodynamic interaction between helicopter fuselage and rotor is developed in the light of helicopter aerodynamics theory. It is still not satisfactory and it requires a great deal of calculation. This paper explores a novel modeling method for the interaction effect of rotor wake on fuselage by way to adopting multilayer feedfoward neural networks. Based on the experimental results of aerodynamic interaction effect made by us last year, including the aerodynamic forces/moments of fuselage changes in accordance with advance ratio ( μ ) and disk loading ( p ), a two input/three output neural network model for calculating the aerodynamic interaction effects of rotor wake on fuselage is proposed. A set of training sample come directly from the experimental data. The trained network model makes the conclusion clear: for the drag of the fuselage the effect of rotor airflow is closely in connection with advance ratio μ and disk loading p , while for the thrust and pitch moment of the fuselage the disk loading plays a more important role. The reasonability and the effectiveness of the neural network model is further testified with the comparison between the experimental measurement of the given test state and the calculated results of the neural network model. The trained neural network model can be directly applied to aerodynamics research and real time simulation of helicopter.