Detection of temporal dependencies in alarm time series of industrial plants

Jens Folmer, Falk Schuricht, Birgit Vogel-Heuser

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

24 Scopus citations

Abstract

Concerning industrial plants, operators face the problem that more alarms are generated than can be physically perceived and addressed by a single operator. Such a situation is called alarm flood. The main reason for alarm floods are causally related disturbances, which either way raise an alarm, based on a single causal disturbance. These dependencies are difficult to recognize during the engineering of an AMS (Alarm Management System). However, the alarms are logged and stored as time series (historical data). Information about the alarm types and the time stamps of their occurrence can be used to analyze the time series data and thus finding dependencies between different alarms. This contribution presents an approach to find temporal dependencies between alarm events in an alarm time series. Therefore an algorithm was designed, implemented, and evaluated to detect temporal dependencies in alarm time series.

Original languageEnglish
Title of host publication19th IFAC World Congress IFAC 2014, Proceedings
EditorsEdward Boje, Xiaohua Xia
PublisherIFAC Secretariat
Pages1802-1807
Number of pages6
ISBN (Electronic)9783902823625
DOIs
StatePublished - 2014
Event19th IFAC World Congress on International Federation of Automatic Control, IFAC 2014 - Cape Town, South Africa
Duration: 24 Aug 201429 Aug 2014

Publication series

NameIFAC Proceedings Volumes (IFAC-PapersOnline)
Volume19
ISSN (Print)1474-6670

Conference

Conference19th IFAC World Congress on International Federation of Automatic Control, IFAC 2014
Country/TerritorySouth Africa
CityCape Town
Period24/08/1429/08/14

Keywords

  • Alarm systems
  • Data processing
  • Delay analysis
  • Manufacturing processes
  • Process automation
  • Timing series analysis

Fingerprint

Dive into the research topics of 'Detection of temporal dependencies in alarm time series of industrial plants'. Together they form a unique fingerprint.

Cite this