贝叶斯正则化BP网络在机翼载荷分析中的应用

采用贝叶斯正则化与BP网络相结合的方法,构造了一个分析飞机机翼载荷的三层BP网络。贝叶斯正则化方法提高了BP神经网络的泛化能力,且能考虑非特征化的非线性影响。使用某型飞机对称机动和滚转机动的试飞实测数据作为载荷激励来训练神经网络,并用这个训练完的网络预测了该飞机机翼的飞行载荷。最后将神经网络预测结果与实测结果进行了比较,结果表明该方法能够准确地实现飞机机翼载荷预测,对新机研制和飞行试验有较高的参考价值。

Application of BP Neural Networks Based on Bayesian Regularization to Aircraft Wing Loads Analysis

This paper establishes three layers BP neural networks model for analysis the aircraft wing flight loads with the method of combining the Bayesian regularization and BP neural networks,this neural networks can account for uncharacterized nonlinear effects while improving the capability to generalize.Neural networks were trained with one aircraft's flight data for symmetries and rolls maneuvering for load excitation,and used this trained networks for forecasting the aircraft wing flight loads.Finally,compare...