TY - JOUR
T1 - Improved alarm flood analysis by cluster identification and alarm assignment
AU - El Sakka, Feras
AU - Bloch, Henry
AU - Kinghorst, Jakob
AU - Pirehgalin, Mina Fahimi
AU - Fay, Alexander
AU - Vogel-Heuser, Birgit
N1 - Publisher Copyright:
© 2018 Walter de Gruyter GmbH, Berlin/Boston.
PY - 2018/8/28
Y1 - 2018/8/28
N2 - Alarm analysis aims to group similar and frequently occurring alarm sequences in alarm clusters. These clusters are so far identified by data-driven approaches based on the historical alarm data. To improve the alarm analysis, especially the handling of alarm floods, additional information can be integrated into state of the art alarm analysis approaches. This additional information consists of plant and process information, which improves the identification and detection of alarm clusters. Information such as neighborly relations of alarm sources or the heat transmission between two not directly connected alarm sources, improve the identification of alarm clusters. The handling of unknown alarms, which occur during runtime, and the assignment to known alarms and additionally to identified alarm clusters can be supported by process and plant information as well. Based on the combination of alarm analysis approaches and the evaluation of process and plant information, it seems feasible to identify causal coherent alarms within alarm floods. This improvement supports the plant operator to handle the abnormal situation by a reduction of the occurring alarms and information about the causal relations between alarms.
AB - Alarm analysis aims to group similar and frequently occurring alarm sequences in alarm clusters. These clusters are so far identified by data-driven approaches based on the historical alarm data. To improve the alarm analysis, especially the handling of alarm floods, additional information can be integrated into state of the art alarm analysis approaches. This additional information consists of plant and process information, which improves the identification and detection of alarm clusters. Information such as neighborly relations of alarm sources or the heat transmission between two not directly connected alarm sources, improve the identification of alarm clusters. The handling of unknown alarms, which occur during runtime, and the assignment to known alarms and additionally to identified alarm clusters can be supported by process and plant information as well. Based on the combination of alarm analysis approaches and the evaluation of process and plant information, it seems feasible to identify causal coherent alarms within alarm floods. This improvement supports the plant operator to handle the abnormal situation by a reduction of the occurring alarms and information about the causal relations between alarms.
KW - Alarm sequence analysis
KW - alarm cluster detection
KW - causality analysis
KW - plant information sources
KW - process information sources
UR - http://www.scopus.com/inward/record.url?scp=85051456248&partnerID=8YFLogxK
U2 - 10.1515/auto-2018-0071
DO - 10.1515/auto-2018-0071
M3 - Article
AN - SCOPUS:85051456248
SN - 0178-2312
VL - 66
SP - 647
EP - 655
JO - At-Automatisierungstechnik
JF - At-Automatisierungstechnik
IS - 8
ER -