基于ARMA神经网络的泵机组故障诊断

Fault Diagnosis of Engine Pump Machine Based on ARMA and Neural Network

泵机组是保障油库正常作业的主要装备,也是一个有机的整体,采集的信号往往以几种故障状态的形式表现出来,因此对其进行故障诊断非常复杂。通过ARMA(时间序列分析法)提取振动信号中的振型参数、阻尼比、振型系数,然后利用BP神经网络进行模式识别。实验结果证明,利用ARMA与BP神经网络结合的方法,可以达到很好的识别效果。

Pump machine is the main equipment that support task of oil port.It is a organic whole.The signal gathered usually includes several fault forms.The fault diagnose to it is very difficult.The character vectors,such as mode parameter,damping ratio and vibration mode coefficient,were drawn out from vibration signal with ARMA（Analysis of Time Series）.The pattern recognition was made with BP neural network.The experimental result proves that the recognition effect of the method combining ARMA and BP neural network is good.