改进FBPNN在飞机系统故障预测中的应用

针对飞机故障预测问题展开研究,提出一种改进的模糊BP神经网络(Fuzzy BP Neural Network,FBPNN)故障预测模型。在FBPNN第二层中,选取影响系统状态评估的多个因素,为各因素确定隶属度值,并利用方差-协方差法给每个因素的隶属度值赋权值,来减小传统FBPNN直接由专家经验选取已有的单因素隶属度函数或它们的演化形式带来的主观性,从而得到更满足实际要求的预测模型,最后利用MATLAB结合飞机襟缝翼系统的相关数据进行仿真研究。结果证明,改进的FBPNN较传统FBPNN具有更好的预测效果。

Application of Improved FBPNN in Aircraft System Fault Prediction

Aiming at the problem of aircraft fault prediction, an improved fuzzy BP neural network fault prediction model is proposed. In the second layer of FBPNN, the membership value is determined for each variable by combining multiple variables that affect the state evaluation of the system, and the weight of each variable’s membership is obtained by the variance-covariance method, so as to reduce the subjectivity brought by the expert experience selecting the existing single factor membership function or their evolutionary forms. The membership function satisfying more practical requirements is obtained. Finally, the simulation results of the aircraft sliver system are carried out by MATLAB. The results show that the proposed method improve the prediction accuracy.