An event-based plant alarm system is proposed as an alternative to the current variable-based alarm system used in the process industry. While a traditional variablebased process alarm indicates an unsafe deviation of a process variable, an event alarm will be a direct indicator of an undesired event. A conceptual framework for the design of such an alarm processor is outlined in this article. A methodology for allocation of alarms based on the information content of variables is highlighted. The annunciation philosophy for event alarms is detailed, and alarm prioritization method and operator guidance are discussed. The method uses Bayesian network and Bayesian inference to estimate probability of an event and to identify its root causes.
History
Publication title
Proceeding of the 11th World Congress on Intelligent Control and Automation
Volume
2015
Pagination
491-496
ISBN
978-1-4799-5825-2
Department/School
Australian Maritime College
Publisher
Curran Associates Inc.
Place of publication
USA
Event title
11th World Congress on Intelligent Control and Automation