经验模态分解和AR模型在飞行器健康诊断中的应用

摘要

To effectively diagnose the aircraft structure components health status, a new kind of health diagnosis approach for the aircraft,based on EMD-AR model and PNN,is proposed in this paper. The advanced acoustic emission(AE) technique is used to monitor the aircraft key parts health state and get the AE information. And the AE signal is decomposed into the limited inherent modality function(IMF) by the EMD.Tben the first two IMF components are used to set up AR model,and compute the AR' parameters with the method of U-C.The auto-regressive parameters and residual variance are extracted to be the eigenvectors.The health status of the aircraft can be diagnosed with PNN health monitor. Experiments show that this method can effectively monitor the fatigue crack of the aircraft structure components. It presents a new approach to diagnose effectively health state of aircraft structure components.%为了有效地诊断飞行器的健康状况,提出了一种基于EMD-AR模型和PNN的飞行器健康诊断新方法.该方法采用EMD(Empircal Mode Decomposition,EMD)将飞行器关键部件的声发射信号进行分解,得到多个内禀模态分量(Intrinsic Mode Function,IMF),对前两个IMF分量建立AR模型,采用U-C算法对AR模型进行参数估计,以模型主要的自回归参数和残差的方差构建特征向量;运用概率神经网络(Probabilistic Neural Network,PNN)对飞行器的健康状态进行诊断.通过对莱型号真实飞行器关键结构部件的健康监测实验表明,该方法可以有效地诊断出飞行器关键结构部件的疲劳裂纹,从而证明了该方法的有效性.