基于多分类SVM的水泵故障诊断的研究

Fault Diagnosis of Pump Based-on Multi-class Support Vector Machines

为了预测水泵在运行中的故障,提高水泵运行的安全性和经济性,在分析故障诊断的基本原理的基础上,建立了水泵运行故障知识库,并运用有向无环决策图多分类算法,对水泵故障样本进行训练,建立了基于支持向量机的离心式水泵运行故障诊断模型(DAGSVM)。通过实例验证表明,该模型能有效的诊断水泵的常见故障,为水泵故障诊断提供了一种新方法。

In order to forecast the faults of pump operation and enhance the security and efficiency, based on the analysis of the basic principle of fault diagnosis based on SVM, the fault knowledge database of pump operation is set up. Meanwhile, the multi-class algorithm of Decision Directed Acyclic Graph （DDAG） is applied for the training of pump faults samples, and the fault diagnosis model of centrifugal pump operation based on SVM is set up. In the end, the precision of the model is verified through an engineering example. The result of diagnosis shows that it is simple, practical and can diagnose the faults effectively.