基于数据复杂网络理论的系统故障检测方法-化工学报.pdf

第 65 卷 第 11 期 化 工 学 报 Vol.65 No.11 November 2014 2014 年 11 月 CIESC Journal ·4503 · 基于数据复杂网络理论的系统故障检测方法 陈雨，韩永明，王尊，耿志强 （北京化工大学信息科学与技术学院，北京 100029 ） 摘要：化工过程系统结构的大型化和复杂性，通过单独的机理模型进行故障检测已越来越困难。提出一种基于数 据复杂网络理论的过程故障检测方法，利用偏相关系数确定复杂变量间的邻接矩阵，生成过程系统数据变量之间 的网络模型，从网络拓扑结构出发，计算系统复杂网络的特征参数，通过对故障模型与非故障模型之间网络特征 参数的差异判断系统是否发生故障，进而找到故障点。以 TE 过程为应用对象，验证了该方法的有效性。 关键词：数据驱动；复杂网络；偏相关系数；故障检测 DOI ：10.3969/j.issn.0438-1157.2014.11.041 中图分类号：TP 39 文献标志码：A 文章编号：0438—1157 （2014 ）11—4503—06 System fault detection based on data-driven and complex networks theory CHEN Yu, HAN Yongming, WANG Zun, GENG Zhiqiang （School of Information Science and Technology ，Beijing University of Chemical Technology ，Beijing 100029，China） Abstract: Since system structures are more and more complex, it’s hard to detect the faults only through physical and chemical mechanism model. In order to overcome these difficulties, a new method based on data-driven and complex network theory is proposed using partial correlation coefficient to determine adjacency matrix of variables, then define the network model. Based on the network topology structure, the characteristics of complex system network are calculated. According to the changing of parameters of network, it can be determined whether a system has failure by contrasting the diffidence between the fault model and normal model. At last, the proposed method is applied to TE process to test its validity. Key words: data driven; complex networks; partial correlation coefficient; fault detection 引 言