Alarm rationalisation is the work of cleaning this up so that every alarm tells an operator something real and useful. This guide describes a practical programme for engine and turbine plants, and the measures that show whether it is working.
Why alarm floods happen
Alarms tend to accumulate. Each commissioning problem, each incident report and each new piece of equipment adds a few. Few are ever removed. Settings copied from the supplier defaults, often generous and generic, remain unchanged for years.
Floods also arise from cascades: one real event, such as a trip, triggers dozens of consequential alarms. Without suppression logic, the operator sees all of them at once, at the moment when clear information matters most.
Start with an alarm philosophy
A philosophy is a short document that defines what an alarm is, what it must tell the operator, how priorities are set, how many alarms an operator should handle and who may change them. Without it, every engineer decides differently.
A common rule is that every alarm must require a response, be understandable, be relevant to the current situation and give time to act. An alarm that fails these tests is a candidate for removal or redesign.
Collect data before changing anything
Export the alarm history for a period such as three months. Count alarms per day, per hour and per operator. Identify the most frequent alarms, which often make up a large share of the total, the alarms that stay active for long periods and the ones that chatter on and off.
This analysis gives a baseline and directs effort to the few alarms that cause most of the noise. Our article on power plant performance monitoring software explains how data collection supports this.
Set priorities
Assign each alarm a priority based on the consequence of ignoring it and the time available to respond. A three-level scheme is often enough: critical, major and minor. If everything is critical, nothing is.
Review the distribution. A healthy system has only a small share of high priority alarms and a lot of low priority information that is reviewed, not sounded.
Review setpoints and deadbands
Many nuisance alarms come from setpoints that sit too close to normal operating values, or from signals that are noisy. Adjust the setpoint to give a useful warning, add deadbands and delays so that a signal crossing the limit briefly does not trigger an alarm, and filter noisy signals.
Check against the equipment manufacturer’s limits. Do not widen a setpoint to silence an alarm if that removes protection. For rotating machines, see our note on gas turbine vibration monitoring.
Handle floods with state-based logic
Suppress consequential alarms when the machine is stopped or tripped, since many alarms are meaningless in that state. Group related alarms so that a first-out indication shows the original cause. Use conditions such as running state, load and mode to enable or disable alarms.
Design these carefully so that no important alarm is suppressed in a state where it matters.
Shelving, annunciation and operator support
Operators need a controlled way to shelve alarms that are temporarily irrelevant, such as a sensor under repair, with a time limit and a record. Uncontrolled acknowledgment and muting hide problems.
Provide clear alarm text, with the cause and the expected action. A short line of guidance at the point of alarm is worth more than a manual on a shelf.
Management of change
After the cleanup, controls are needed to keep it. Every new alarm should be justified through a short form that states its purpose, priority, response and setpoint. Changes to existing alarms need approval and a log entry.
Without this discipline, the system will drift back to its previous state within a few years.
A worked example of a bad actor alarm
Consider a lube oil temperature alarm that sounds forty times a day on a standby engine because the setpoint sits just above the warm-up temperature in hot weather. Operators acknowledge it without thinking. Adjusting the setpoint with a delay, or suppressing it when the engine is stopped, removes dozens of calls a week without losing any protection.
A short list of such bad actors, usually only a dozen or so, often accounts for most of the noise. Fixing them first gives the quickest win and builds support for the larger programme.
Involving operators in the programme
Operators know which alarms are useless and which ones save the plant. Include them in reviews, ask for their list of the worst alarms and report back on changes made. Their trust is the foundation of any improvement.
Run short workshops by system, such as fuel, cooling and electrical, with operators, maintenance and engineering together.
Measuring success
Track the average number of alarms per operator per hour, the peak count in ten minutes, the share of time with more than a stated number of active alarms and the list of the top bad actors. Publish the figures monthly and review them with operations and maintenance.
Set targets and celebrate progress. Operators who see their feedback acted on are more likely to report problems early.
Who writes this
Operator workload and alarm quality are central to how our plant monitoring specialists work. Thanks to the USP&E operations team for the content. USP&E works as principal on EPC and O&M scopes, not as a broker, and highlights 150+ power projects in 45+ countries, 25,000 MW contracted since 2002, 24+ years in the field. See how our operations and maintenance services use monitoring data to run plants.
Offices: Colorado Springs (1755 Telstar Dr, Suite 300), Knoxville (800 S Gay St, Suite 700), London (162-168 Regent Street), Johannesburg (22 Kildoon Rd, Bryanston), Cape Town (31 Brickfield Road, Woodstock) and China (No. 1 Ningzhou Road, Wuxi). Regions: USA | AFRICA | UK | UAE | CHINA. ISO 9001 and ISO 45001 certified.
Questions about alarm rationalisation
How many alarms should an operator handle?
Far fewer than most plants have today. Set a target in the alarm philosophy and measure alarms per operator per hour against it.
Is it safe to remove alarms?
Yes, if each one is reviewed against the philosophy and the equipment limits. Never remove protection without engineering approval.
What is the first step?
Export three months of alarm history and find the few alarms that account for most of the noise.
How do we keep it clean?
Use a management of change process for every new or changed alarm, and review the key figures every month.
Drowning in plant alarms?
Send us a sample of your alarm history and our engineers will show where the noise comes from and how to reduce it.
Send drawings, load data or questions via the project enquiry form; the switchboard is +27 10 822 2324 and the inbox is info@uspeglobal.com. Equipment for plant upgrades is listed in the USP&E inventory.