多智能方法及其融合系统在化工过程故障诊断中的应用  

APPLICATION OF MULTIPLE INTELLIGENT METHODS AND FUSION SYSTEM IN FAILURE DIAGNOSIS OF CHEMICAL PROCESS

作　　者：黄承武[1] 奚旦立[1] 王菁辉[2] 

机构地区：[1]东华大学环境科学与工程学院,上海市200051 [2]中国石化集团洛阳石油化工工程公司工程研究院,河南省洛阳市471003

出　　处：《炼油技术与工程》2007年第6期40-43,共4页Petroleum Refinery Engineering

摘　　要：简要介绍了决策树、神经网络、K近邻等五种智能方法,提出了一种将这五种智能方法相融合的故障诊断系统,并将其应用在催化裂化装置上。通过对这五种智能方法及其融合系统的性能考察表明,融合策略发挥了各方法的优点,回避了它们的缺点,提升了系统的整体性能,其召回率为100％,精确度为98％。Five intelligent methods, such as decision tree, neural network, K-nearest neighbor, etc. are briefly introduced, and the failure diagnosis system with fusion of these five methods is applied in a FCC unit. The performances of these five methods and their fusion system indicate that the disadvantages of these five methods are avoided while gaining their advantages in the fusion system. The recall and precision of the fusion system are 100% and 98%, respectively.

关 键 词：化工过程 融合 故障诊断 决策树 神经网络 K近邻 贝叶斯 支持向量机 

分 类 号：TQ02[化学工程]