Skip to content

07 Features

Predictive maintenance

A failure almost never arrives without warning. It arrives without anyone watching the signal that had been announcing it for weeks.

Anomalies Drift Condition-aware thresholds Actual hours

Illustrative figures.

How it works

Watch the ratio, not the value

Calendar maintenance services machines that are fine and misses the ones about to break, because the calendar does not look at the machine. Run-to-failure costs less in parts and far more in downtime.

History comes first

Drift is recognised against normal behaviour, and normal behaviour is observed rather than declared. This rung does not work without the first one: that is why the ladder is a ladder.

Watch the ratio, not the value

Absolute consumption moves with load and says nothing. What drifts is the ratio between quantities: energy per unit produced, draw at equal flow, temperature differential at equal duty.

The threshold follows the conditions

A fixed threshold fires all summer and stays quiet all winter. A threshold computed on current conditions reports the deviation, which is the thing you care about.

The alert reaches whoever decides

A notification with the value, the machine and how long the drift has been running. Without the "how long", the alert becomes one more thing to ignore.

What comes out

What comes out

No model invents the warning: how much notice you get depends on the signal and on the length of the history it was built on.

It does not work without the first rung

Where it is used

In every industry, no exceptions

The protocol changes, the way of working does not. It is why the same measurement layer sits under plants that have nothing in common.

How long has that machine
been drifting?

Tell us which quantities you already record: from there we can see whether the history is enough to catch a drift.