(16b) Alarm Management for Improved Process Safety, Reliability, and Product Quality | AIChE

(16b) Alarm Management for Improved Process Safety, Reliability, and Product Quality

Authors 

Soroush, M. - Presenter, Drexel university
Cadet, O. - Presenter, Process Control & Logistics, Air Liquide
Arbogast, J. - Presenter, Process Control & Logistics, Air Liquide
Seider, W. D. - Presenter, Risk Management and Decision Center, Wharton School,University of Pennsylvania
Oktem, U. - Presenter, Risk Management and Decision Center, Wharton School,University of Pennsylvania
Pariyani, A. - Presenter, University of Pennsylvania


Most chemical plants have
hundreds of variables that monitor the dynamics of their processes.  When the
variables move out of their normal operating ranges, alarms are triggered to notify
the human operators.  The frequency and selectivity of alarms influence the
performance of operators, which in turn, impacts the plant performance and
reliability.  This work introduces techniques to perform alarm management
that improves the reliability of plants.  To ensure the relevance and
specificity of alarms, particularly during abnormal situations, the work
focused on providing operators with actionable alarm information.  Bad
Actors
(BAs), which are alarms that deteriorate the performance of alarm
systems, are identified and resolved.  Typically, BAs are poorly-configured alarms
that have the highest alarm counts, often contributing to periods of high alarm
frequency (referred to as alarm floods).  In this work, they include
alarms that experience the most-critical abnormal events (when variables
cross their Emergency Shutdown thresholds) and/or lengthy departures from their
normal operating ranges ? jeopardizing plant safety, reliability, and product quality. 
The latter are important for many plants operated unmanned over nights/weekends.

Several months of alarm data
were analyzed for an industrial air-separation unit operated by Air Liquide.  The
BAs were identified and assigned different grade levels.  Several Key
Performance Indicators (KPIs) were developed to evaluate the performance of
alarm systems.  The methodologies developed in this work improve upon the
existing industry-standard techniques, which focus only on the alarm counts and
not on the criticality of abnormal situations.  Results show that the average frequency
of alarms was well under manageable limits, with the alarm system in the flood
condition just 1% of the total time.  However, 45% of the total alarms during
the study period occurred during alarm floods.  These alarm floods are referred
to as near-misses, which often distract operators, but don't necessarily
result in trips or accidents.