理想薄平板气动导数的人工神经网络识别

介绍一种用BP人工神经网络方法识别理想薄平板的气动导数的方法．通过构造BP网络，比较各种因素对识别结果的影响．数据处理方式对神经网络的训练结果影响很大．隐层单元数量、训练次数、随机赋值次数、样本数量等对训练结果也有影响，但影响较小．采用这种方法识别理想薄平板的气动导数是可行的，并且具有较高的精度．

Identification of Aerodynamic Derivatives of Ideal Thin Plates with Artificial Neural Network

A BP artificial neural network (ANN) method is introduced to identify aerodynamic derivatives of ideal thin plates. An artificial neural network is constructed followed by the comparison between the prediction results influenced by some factors. The mothods of data processing have strong effects on training results, and the effects of other factors are relatively weak. The training results show that this approach is feasible with satisfactory accuracy.