What it is
Anomaly detection is the automatic spotting of data that deviates from normal patterns, often before any fixed threshold would trigger. Fixed limits still have their place — the ISO 20816 severity zones are one — but they only fire once a machine is already out of band.
How it works
Models learn baseline behaviour and flag deviations even without explicit rules, which also helps reduce the flood of low-value alarms.
Where it fits
It is the foundation under predictive maintenance, performance monitoring and security monitoring alike.