Alarm rationalisation - from 400 standing alarms to the ten that matter
Walk into the control room or log on to the head end of many existing buildings and the alarm list tells the same story: hundreds of active alarms, most of them weeks or months old, and an operator who has learned to ignore the lot. When a real problem arrives - a chiller trip, a frost risk, a plant room flood - it lands in the middle of the noise.
Alarm rationalisation is the work of turning that list back into something people can act on. This guide covers why alarm floods happen, the tools available to fix them and a practical plan for cleaning up an existing system. The 400 in the title is illustrative; the principles apply whether a site has forty standing alarms or four thousand.
Why alarm floods happen
Very few systems are designed to flood. They get there gradually:
- Every point gets an alarm at commissioning - because it is easier to add alarms than to decide which ones matter.
- Limits are set too tight - a space temperature alarm a degree from setpoint will trigger every warm afternoon.
- No delays - a momentary spike raises and clears an alarm dozens of times a day.
- Faults are never fixed - a failed sensor alarms for months because the repair never gets raised.
- Building changes - plant is removed or spaces repurposed, but the alarms stay.
- Nobody owns the list - alarms go to a screen that no one is responsible for watching.
The result is alarm fatigue. People stop reading alarms, and the system loses the one job it most needs to do: telling the right person, quickly, that something needs attention.
What good looks like
The process industries have long-established published guidance on alarm management, notably EEMUA 191 and the IEC 62682 standard. Building services are different - the consequences are usually lower and the operators are often not full-time - but the core ideas transfer well:
- every alarm should need a response from someone;
- every alarm should be understandable without specialist knowledge of the strategy;
- the number of alarms should be manageable by the people receiving them;
- each alarm should reach the person who can act on it.
Those four tests are a useful filter for every alarm on the system.
Priorities and categories
A clear priority scheme is the foundation. Many sites use three or four levels, for example:
- Critical - life safety, plant protection or business-critical failures that need action now, at any hour.
- Urgent - failures that will affect comfort or plant if not dealt with the same day.
- Routine - faults to fix at the next convenient opportunity.
- Information - events worth logging but not worth interrupting anyone for.
Alongside priority, categories help route alarms: plant failures, environmental limits, maintenance alerts, communication faults and so on. Agree the scheme with the client, write it down and apply it consistently. The scheme itself matters less than everyone using the same one.
Delays, hysteresis and shelving
Three simple tools remove a large share of nuisance alarms without hiding real problems.
- Delays - an alarm condition must persist for a set time before the alarm is raised. A space temperature high alarm with a 15-minute delay ignores a door left open briefly; a pump failure alarm might need only a few seconds.
- Hysteresis (deadband) - once raised, an alarm must move back past its limit by a margin before it clears. This stops alarms chattering on and off around a limit.
- Conditional alarms - an alarm only applies when it makes sense. A low supply temperature alarm should not raise while the air handling unit is off; a space temperature alarm may not matter out of hours.
Shelving is the temporary suppression of a known alarm - for example, while a repair is awaited. It must be controlled: shelved alarms should be listed, have an owner and a reason, and come back automatically after a set time. Uncontrolled shelving is just another way of ignoring alarms.
Routing by role
An alarm that goes to everyone goes to no one. Decide who needs which alarms:
- the facilities manager - critical and urgent alarms affecting the building and its occupants;
- the service provider - plant and controls faults they are contracted to fix;
- security or out-of-hours staff - critical alarms at night and weekends, with clear instructions;
- specialists - for example, catering, laboratory or data hall alarms to the teams responsible.
Check that alarm routing still matches the people actually on site. Phone numbers and email addresses for people who left years ago are a common finding.
Monthly alarm KPIs
What gets measured gets managed. A short monthly alarm report, as part of the service report, keeps the list under control. Useful measures include:
- total alarms raised in the month, by priority;
- standing alarms at the end of the month, and how long the oldest have been active;
- the top ten alarm sources - usually a small number of points generate a large share of the alarms;
- chattering alarms - points that raised and cleared many times;
- shelved alarms, with their owners and ages;
- response times for critical and urgent alarms, where the system records acknowledgement.
For a sense of scale, EEMUA 191 is commonly cited as suggesting that, in normal operation, an operator should see no more than around one new alarm every ten minutes, and it treats more than about ten alarms in ten minutes as a flood. Those figures were developed for staffed process control rooms and do not translate directly to most buildings, where nobody may be watching the screen full time - so agree targets with each client based on how the site is actually staffed. Tracking the trend month to month matters more than hitting a particular number.
A worked clean-up plan
The following sequence is one way to tackle an existing system over a few visits. The example is fictional, but the shape is typical.
Step 1 - Measure. Export a month of alarm history. Count alarms by point, by priority and by time of day. In a typical flooded system, a small number of sources dominate the list.
Step 2 - Fix the top ten. For each of the top ten sources, find the cause. Is it a real fault that needs a repair? A limit set too tight? A missing delay? A sensor that has failed? Fix the cause, not the symptom - suppressing an alarm from a failed sensor without raising the repair simply hides the fault.
Step 3 - Clear the dead wood. Remove alarms for plant that no longer exists, and disable alarms that fail the four tests above. Record what was removed and why, and agree it with the client.
Step 4 - Apply the priority scheme. Go through the remaining alarms and give each a priority and category. This is the step that takes time, and it is worth doing with the facilities team.
Step 5 - Add delays, hysteresis and conditions where nuisance alarms remain.
Step 6 - Set up routing by role and test it - including out of hours.
Step 7 - Report monthly. Add the alarm KPIs to the service report and review them at each visit.
The aim is not an empty alarm list. It is a list where every alarm means something, reaches the right person and gets dealt with.
Where this fits
Alarm load is one of the areas of a BMS health check and a core part of every seasonal service visit. Many standing alarms trace back to faults that the rules in Fault detection on a BMS find, and to setup problems covered in Commissioning a BMS. Fixing the causes often brings energy benefits too - see Where BMS tuning saves energy.
Related tools and training: the BMS fundamentals and Commissioning good practice courses are in tecTraining, and the engineering reference and rules of thumb are in tecTools. On sites running tecGfx, alarms are followed from first report to acknowledgement alongside work orders and reports, which makes the monthly figures easier to gather.
Published guidance mentioned: EEMUA 191 Alarm systems; IEC 62682 Management of alarm systems for the process industries. Both are written for the process industries and are described here in our own words - refer to the documents for their content.