引用本文:方欢,张源,吴其林.一种基于结构因果关系和日志变化挖掘的BPMSs故障诊断方法[J].控制理论与应用,2018,35(8):1167~1176.[点击复制]
FANG Huan,ZHANG Yuan,WU Qi-lin.A log induced change mining method for fault diagnosis using structure causality in BPMSs[J].Control Theory and Technology,2018,35(8):1167~1176.[点击复制]
一种基于结构因果关系和日志变化挖掘的BPMSs故障诊断方法
A log induced change mining method for fault diagnosis using structure causality in BPMSs
摘要点击 2460  全文点击 1164    
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2018.17091
  2018,35(8):1167-1176
中文关键词  行为轮廓  变化挖掘  故障检测与定位  Petri网  结构因果关系
英文关键词  behavioral profiles  change mining  fault detection and diagnosis  Petri nets  structural causality
基金项目  Supported by the National Natural Science Foundation of China (61472003, 61272153, 61340003, 61402011, 61572035), the Natural Science Foundation of Anhui Province (1608085QF149), the Youth Talents Support Project in Universities of Anhui Province (gxyqZD2018038) and the Science & Technology Foundation of Huainan City (2016A23).
作者单位
方欢* 安徽理工大学 
张源 安徽理工大学 
吴其林 巢湖学院 
中文摘要
      业务流程管理系统存在可以改变系统行为的潜在故障, 因此研究定位系统中故障发生的最小结构变化区域是 十分必要的, 它对提高业务系统的鲁棒性具有重要意义. 本文提出了一种日志诱导下的变化挖掘方法, 即最小结构故障 域识别方法(minimal structure fault region identification, MSFRI), 该方法通过系统的行为变化来定位故障发生的结构因 果关系. 进一步, 针对合理的自由选择业务流程Petri网系统, 形式化定义了6种典型变化模式, 这些变化模式为故障的结 构因果关系变化挖掘提供理论基础. 本文所提出的故障定位方法通过识别业务流程Petri网系统的行为变化, 实现具有最 少库所和变迁数目的故障区域定位, 有助于实现系统更加复杂的变化挖掘. 本文工作的主要创新之处在于从结构因果关 系的角度出发, 借助系统行为变化挖掘实现定位业务系统中的潜在故障.
英文摘要
      In business process management systems(BPMSs), impending faults could change the behavior of the system; thus, it is necessary to investigate the methods that identify the minimum change region of the system due to faults, which is of great importance for the robustness of BPMSs. In this paper, we propose a log induced change mining method, named the minimal structure fault region identification (MSFRI) method, which has the primary goal to identify the structural causality for given behavior changes. The MSFRI method is applied in free-choice net Petri nets systems, and six characteristic change patterns are formalized, which provide structural causality foundations for change mining. The method locates the fault regions that deduces the behavior changes, which have the smallest number of places and transitions in a Petri net system. The identification of these regions is helpful for change mining. Our novel MSFRI method provides a structural perspective for analyzing the causality for behavior changes, such as an impending fault in BPMSs.