Insights / SCADA
Designing SCADA and alarm management sensibly
An alarm that does not lead to an action is not information, it is noise. And noise has one property: it trains the operating team to ignore alarms. That is precisely the moment in which the decisive alarm disappears.
Good alarm management is therefore not a configuration question but a question of operational organisation. The software provides the tools – priorities, escalation, filters, logs. They only become useful once it is clear who reacts to what and how long they have for it.
Alarms are a language problem
Every alarm is a statement about the state of a plant. Good alarm texts name cause and location, not just the event: “Drive A3 over-temperature” is a statement, “temperature error” is a label. From a label the operator still has to derive what to do – and it is precisely those seconds that are missing in a fault.
- Name cause and location in the alarm text.
- Include unit and limit value when they matter for interpretation.
- Do not take technical designations from the visualisation out of context.
- One alarm per event – no collective messages that obscure several causes.
Priorities that actually mean something
Alarm systems allow a high number of priorities – EVON XAMControl up to 32,767. The temptation is to prioritise every difference. In practice far fewer levels make sense, because a priority only carries information if it is also treated differently during operation.
A model that has proven itself
- Critical – safety-relevant deviation, immediate reaction, escalation to the standby team.
- Warning – deviation requiring action, dealt with in the next shift.
- Information – non-critical deviation or status message, logged.
Anything that does not fit into one of these levels belongs in the process visualisation and not in the alarm list. The place where a message appears is itself a prioritisation.
Alarm notification: the path to the operator
A priority without an escalation path stays in the visualisation. Especially at night and at weekends, notification by SMS or e-mail decides whether a fault is noticed before it causes damage. The alarm groups determine who is informed in which order – and how long the response time may be.
- Alarm groups per plant, shift and area.
- Escalation levels with clear response deadlines.
- Acknowledgement requirement and documentation of operator interventions.
- An alarm journal that keeps the history of a fault reconstructable even after weeks.
Operator logic: the interface decides
The alarm screen is the workplace during a fault. Its columns, sort criteria and filters determine how quickly an operator finds the actual problem – or how long they have to work through a list. The most important function is therefore the direct navigation to the alarmed object: one click, and the visualisation shows exactly that spot.
Just as important is the look back: what happened before? A journal with alarm history and operator interventions, combined with a trend of the affected value, often answers the question “since when” faster than any additional diagnostics.
Alarm management is part of traceability
In validated processes it does not only matter that an alarm arrived, but that the reaction is documented. Anyone who logs operator interventions automatically has the evidence that the process was under control – without having to reconstruct it afterwards. That is one of the reasons we also use XAMControl where the EMS is the focus.
How many priorities make sense?
Fewer than you think. Three to five levels are usually sufficient in practice, because only priorities that are genuinely treated differently during operation carry information.
What happens in an alarm storm?
That is the case in which prioritisation and notification at the source work: root cause elimination, grouping and time-staggered escalation prevent the operator from being overwhelmed by a hundred messages.
Can an existing alarm system be taken over?
Yes. It is usually more sensible to take over the existing structure, re-evaluate the priorities and add the escalation than to force a migration to a new system.