Nobody Reads the Alarm List: Designing Priority Levels, Suppression and Reset Archiving in SCADA

Five reasons alarm lists stop being read

• No priority levels: a fault that stops the machine and a piece of reference information appear in the same colour with the same tone, so operators cannot tell what to handle first
• Chattering repeats: a value crossing back and forth around a threshold regenerates the same alarm dozens of times within minutes and fills the list
• One fault triggers a cascade: when an upstream machine stops, downstream stations emit dozens of derived alarms and the first-out cause gets buried
• No suppression strategy: during a stop, maintenance work or start-up, normal-running conditions keep evaluating and keep producing alarms that are irrelevant at that moment
• Nobody owns closeout: nothing is acknowledged and no action is recorded, so unclosed alarms accumulate and the list never gets cleared

What these five share is not that the list is long, but that nobody can pick the one item worth reading out of it. That is the problem alarm rationalization has to solve.

Five design actions you can implement

Rationalization is not about deleting alarm points. Delete them and the abnormal conditions still happen, only nobody notices. The work is to give every alarm a priority level, a trigger condition, a suppression rule and an owner, so it appears only when it deserves to be seen.

• Define priority levels: fix the colour, audible tone, notification target and response time per level, and keep the on-screen representation identical across the plant
• Deadband and delay: add a deadband around each threshold and set separate on-delay and off-delay durations so chattering is filtered out before it reaches the list
• First-out alarm and cascade suppression: define parent-child relationships between alarms, suppress downstream derived alarms while an upstream unit is stopped, and keep only the first-out event
• Reset and archiving rules: record all four states — raised, acknowledged, returned to normal, closed — and write them to a historian so each alarm becomes a searchable event
• Owner-based closeout: assign a first responder per level, and review unclosed counts and dwell time daily rather than total occurrences

How to set the priority criteria

Three or four levels is a practical range. Adding levels makes the criteria harder to apply on the floor, and the usual result is that every alarm ends up registered at the same level.

A clear way to draw the lines is three questions: does the equipment stop, is product quality affected, is safety involved. If two levels lead to the same human action, they do not need to be separate levels.

The table above is a starting template, not a universal standard. The criteria have to be checked line by line against your equipment and process conditions, together with the site team.

Suppression and notification have to be designed together

Suppression and notification are two sides of the same design. Adding notification channels without suppression means the phones ring too, and they end up on silent. The more channels there are, the more thoroughly they get ignored.

• Suppress by operating mode: mask alarms by group during a stop, maintenance work or start-up, and show clearly on screen that a mask is active
• Tier the channels: on-screen display covers all levels, audible tones cover the two upper levels, email and mobile push cover critical only, narrowing at each step
• Escalate on dwell time: raise only the alarms that stay unclosed past the agreed time, rather than escalating on occurrence counts
• Every mask needs an expiry: temporary suppression always carries an expiry time and reverts automatically, so no permanent mask survives that nobody remembers

Using the archive for monthly event review

Three figures in the alarm history are worth reviewing every month: the alarm items leading the list by occurrence count, the average time to close, and repeat counts that suggest chattering. On many sites, acting on the ten most frequent items alone changes the length of the list noticeably.

The review is not a one-off. Thresholds and levels should be readjusted quarterly, and that operating rule belongs in the technical agreement at acceptance. We have handled alarm engineering on a project of nearly 30,000 alarm points. Those figures are evidence from a specific job; they are not the size we assume for every site, and they are not a tag promise copied from product literature onto every contract. The scope covered priority levels, reset archiving, event management and multi-channel notification, and we run the same approach to alarm priority and event management on pump station group control SCADA projects.