基于AR模型参数向量的灰度神经网络故障诊断方法

该文首先论述了以频域能量积分为诊断向量、以各种训练网络为诊断网络的方法在应用中存在的问题，然后提出了一种以AR模型参数作为旋转机械故障诊断向量、以灰色关联度神经网络作为诊断网络的新方法，并描述了诊断系统的具体实现过程。最后通过对10种典型旋转机械故障的仿真诊断比较了该文方法和以往方法的诊断效果。方法分析及仿真结果表明，该文方法比大多数以频域能量积分为诊断向量、以各种训练网络为诊断网络的方法有很多的优越性。

A Fault Diagnosis Method Based on AR Parameter Fault Vector and Gray Neural Network

The problems of the traditional methods in which the energy integrals in frequency domain are taken as fault vector and trainable artificial net works are used as diagnosis network in practice use are talked about. And then a new rotary mechanism fault diagnosis method is proposed. AR parameters gotten from vibration signal data series are treated as the fault vector and gray neural network is introduced to detect the fault. The realization of this diagnosis system is described concretely. The diagnosis results of these methods are compared. Simulation shows that this method has advantages of simpler calculating steps and higher diagnosis validity.