基于模糊规则和支持向量机的凝汽器故障诊断

Diagnoss of Condenser Faults Based on Fuzzy Rules and a Supportive Vector Machine

有向无环决策图支持向量机(DDAGSVM)算法是在支持向量机1-v-1算法基础上,引入了图论中有向无环图的思想而构建的多分类方法,它将多个支持向量机1-v-1两类分类器组合成一个带有根结点的多层的有向无环决策图来实现分类,它建立在结构风险最小化原理基础之上,能在训练样本较少的情况下得到很好的分类效果。在总结汽轮机凝汽器常见故障的基础上,建立了凝汽器典型故障集,通过模糊规则获得凝汽器故障征兆知识库,采用DDAGSVM算法对小样本情况下凝汽器设备典型故障诊断进行了研究,实例计算表明DDAGSVM算法具有较高的诊断准确率。

The algorithm of decision-directed acyclic graph supportive vector-machine(DDAGSVM) represents a multiple classification method established by introducing the directed acyclic graph ideology of graph theory on the basis of the supportive vector machine 1-v-1 algorithm.It combined two kinds of multiple supportive machine 1-v-1 classifiers into a multiple layer directed acyclic decision-making chart with root nodes to realize a classification.It is built on the minimum structure risk theory and can attain an excellent classification effectiveness under the condition of relatively less training specimens.On the basis of summarizing commonly seen faults of steam turbine condensers,a typical fault set was established and a fault omen repository,acquired through fuzzy rules.A study was conducted of the use of DDAGSVM algorithm for diagnosing typical faults of condenser equipment items under the condition of small specimens.The example calculation results show that the DDAGSVM algorithm enjoys a relatively high diagnosis accuracy.