Anomalies find you, not the other way round.
Thresholds, drifts and downtime detected automatically on every tag, with the context to decide and the action to step in. With a minimum duration and a cooldown, because an alarm that always rings is not an alarm.
Not every alert is the same.
Threshold
The most direct condition: a value leaves the limits you set. It is written in MXL, for example temperature greater than 80.
Trend
The drift before the failure: consumption creeping up, a cycle time stretching out, a vibration changing shape.
Prediction
Models flag abnormal behaviour before it becomes downtime, combining vibration, temperature and current draw.
The alert, with the context to decide.
Why it does not ring constantly.
Minimum duration. A condition has to persist for the seconds you set before it becomes an alert. Isolated spikes wake nobody up.
Cooldown. Between one notification and the next sits the interval you choose, so the same problem does not pile onto whoever is already handling it.
Escalation. The alert goes to the operator. If the condition does not clear, it escalates. The notification channel today is email.
Before and after the alert.
The rest, in short.
How is a condition defined?
In MXL, Meddle's expression language: comparisons, AND and OR over numbers, strings and booleans. For example temperature greater than 80 and status equal to active.
How do you avoid useless alerts?
With a minimum duration before it fires and a cooldown between notifications. Both are configured per alert.
Which channels do notifications arrive on?
Today by email, with sender, recipients, subject and body configurable on each alert.
Does an anomaly trigger an action on its own?
No. It flags and suggests. Writing to the machines goes through the agents, with your approval.
Want the anomalies to find you?
We connect one line and configure the first alerts on your own tags, in a guided demo.
Book a demo