基于可视图网络节点重要性度量的离心泵振动故障诊断方法

Method for Diagnosing the Vibration Fault of a Centrifugal Pump Based on the Visual Graph Network Node Importance Measure

针对离心泵振动信号的非线性非平稳的特征,提出了一种基于可视图构建复杂网络(简称可视图网络)节点重要性度量的离心泵故障诊断方法。采用可视图的方法构建网络,提取网络中的特征参数,对离心泵的正常、不对中、不平衡和基础松动4种状态进行分析得到,可视图网络比相关系数网络能提取更为准确的网络信息,更能准确的对离心泵的故障进行诊断分析。通过对网络中重要节点的度量,得到网络节点重要性的综合评价结果,采用中间中心度(BC)指标进行故障诊断,诊断正确率能达到98.7%,与其它指标相比更适合故障诊断。研究结果表明,基于可视图网络节点重要性度量的方法对离心泵振动故障能进行较为准确的诊断。

In the light of the nonlinear and unsteady characteristics of the vibration signals from a centrifugal pump,proposed was a method for diagnosing a fault of a centrifugal pump based on the visual graph network node importance measure( shortly referred to as visual graph network). A network was built by using the visual graph method and its characteristic parameters were extracted from the network. An analysis of the four kinds of network of the centrifugal pump,i. e. normal,non-aligned,non-balanced and loosened in the foundation,came to a conclusion that more accurate network information can be extracted from the visual graph network than from the relevant coefficient network and the former can be used to more accurately diagnose and analyze any fault of the centrifugal pump. By measuring the important nodes of the network,the authors obtained comprehensive evaluation results of the importance of the nodes in the network. The BC( betweenness centrality) index was used to diagnose fault,the correct percentage of a diagnosis can be up to 98. 7%,more applicable for fault diagnosis compared with other indexes. It has been found that the method based on the visual graph and network node importance measure can diagnose more accurately any vibration fault of a centrifugal pump.