A Data Driven Method to Detect First-Out Alarms Based on Alarm Occurrence Events

A first-out alarm is an alarm that comes first in a group of highly correlated alarms. To achieve alarm reduction, it is highly recommended in industrial alarm management standards to only present first-out alarms to operators, whereas consequential alarms should be silenced or suppressed using first-out display or logic alarm suppression in ANSI/ISA-18.2. However, to collect such groups of correlated alarms and identify first-out alarms from them usually relies on proficient knowledge and experience, and thus is time and resource intensive. This work proposes a data driven method to detect first-out alarms from historical alarm data. Two main steps are involved: 1) consequential alarms are detected for each unique alarm based on alarm occurrence events; 2) first-out rules are produced and pruned based on redundant alarm analysis. The effectiveness of the proposed method is demonstrated by an industrial case study involving alarm data collected from a real industrial facility. The detected first-out rules can be used as candidates to design first-out display or logic alarm suppression strategies in practice.