Alarm Flood Analysis by Hierarchical Clustering of the Probabilistic Dependency between Alarms

Iris Weis, Jakob Kinghorst, Thomas Kroger, Mina Fahimi Pirehgalin, Birgit Vogel-Heuser

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

7 Scopus citations

Abstract

Pattern detection in alarm data aroused great interest in research activities in recent years. Reducing alarm floods, which arise of causal dependencies of equipment and their alarms in automated production systems, is aimed to decrease alarm rates and aggregate information to valuable notifications for the operator. However, common alarm flood analysis often lacks robustness against random occurring alarms or interfering alarm patterns, which disturb the known structure of sequences. Therefore, this contribution introduce a preprocessing step, calculating the dependencies of alarms probabilistically. This approach meet the fuzziness of alarm systems regarding precision in time domain and interfering alarm signals. The results, based on two different industrial data sets, reveal high accurateness of the clusters defined by the proposed method. Alarm patterns can be detected even though the sequences are interrupted and interfered by further alarms or further causal dependent alarm floods.

Original languageEnglish
Title of host publicationProceedings - IEEE 16th International Conference on Industrial Informatics, INDIN 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages227-232
Number of pages6
ISBN (Electronic)9781538648292
DOIs
StatePublished - 24 Sep 2018
Event16th IEEE International Conference on Industrial Informatics, INDIN 2018 - Porto, Portugal
Duration: 18 Jul 201820 Jul 2018

Publication series

NameProceedings - IEEE 16th International Conference on Industrial Informatics, INDIN 2018

Conference

Conference16th IEEE International Conference on Industrial Informatics, INDIN 2018
Country/TerritoryPortugal
CityPorto
Period18/07/1820/07/18

Keywords

  • Alarm pattern detection
  • Data-driven methods
  • Hierarchical clustering

Fingerprint

Dive into the research topics of 'Alarm Flood Analysis by Hierarchical Clustering of the Probabilistic Dependency between Alarms'. Together they form a unique fingerprint.

Cite this