Service Engine - Area 02
A service need, surfaced before it becomes a complaint.
Predictive alerts read two things: how this unit is running compared to its own healthy band, and how its model tends to fail in the field. When both point the same way, the pulse lights up an amber flag.
Predicted service need in 12 days
Usage pattern deviates from healthy band for this model.
confidence 0.82 · reason: cycle-time drift
What this area does
Three things it gets right.
Usage-pattern deviation
Cycle time, run-hours and duty pattern for a specific unit are checked against the healthy band derived from similar units in similar conditions.
Model failure history
The known failure curve of the model - which components tend to age at which point - shapes the prediction, not just this one unit.
Confidence, not certainty
Every alert carries a confidence score and a reason string. Dispatchers can accept, defer, or override the pulse without losing the trail.
Worked example
One alert, from signal to accepted visit.
Signal detected
day 341
Cycle-time drift 1.6x above healthy band, three weeks running.
Model context added
X200 series
Compressor fatigue known between day-330 and day-370.
Alert raised
confidence 0.82
Reason: cycle-time drift + model failure history match.
Accepted by dispatcher
visit scheduled
Passed to scheduling with a suggested 5-day window.
A note on honesty. Prediction accuracy depends on how much usage data an engine has seen for the model. Coverage is strongest for common appliance categories and AMC-heavy equipment. New models start with model-history priors, and improve as unit data lands.