Research on Model-less Diagnosis of Discrete Event Systems

2015 Fiscal Year Final Research Report

• Project/Area Number: 25330011
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Theory of informatics
• Research Institution: Japan Advanced Institute of Science and Technology
• Principal Investigator: Hiraishi Kunihiko 北陸先端科学技術大学院大学, 情報科学研究科, 教授 (40251970)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: 離散事象システム / 故障診断 / 異常検知 / イベントログ

Outline of Final Research Achievements

Model-base diagnosis is an existing technique for diagnosis of discrete event systems, but it is hard to be applied to large concurrent systems due to state space explosion. In this research, we have developed a new diagnosis method for discrete event systems, called model-less diagnosis, that does not use any system models. The proposed method consists of two phases: (i)learning probabilistic models that approximate system behavior from given system logs under various situations (learning phase), and (ii) computing characteristic values of event sequences based on the probabilistic model in the normal situation and identifying system failures using the characteristic values. Effectiveness of the method is validated using computer simulation and real data,

Free Research Field

システム科学