Open the model up โ see exactly why each asset was scored the way it was.
Vibration telemetry shows an accelerating fault signature. 85% probability of failure within 4 days.
Each factorโs signed contribution to the score. Bars to the right pushed the score up; green bars are counter-evidence the model weighed down. The contributions sum to 85%.
The Failure Prediction model scored HVAC Rooftop Unit B at 85%, driven mainly by โVibration +3.2ฯ over baselineโ and โBearing temp trending up 6 wksโ. Together the top factors account for 88% of the score.
Auto-generate a Critical predictive work order and dispatch a technician within 48h.
The smallest set of factors that, if resolved, flips this to low risk.
Global importances reflect the model overall; the contribution chart above is local to this one prediction.
Telemetry, RTLS pings, maintenance & custody logs streamed from the asset.
โบEngineered signals โ vibration trend (ฯ), temperature trend and more.
โบFailure Prediction v3.5.0 (LSTM), retrained 8d ago.
โบ85% critical signal with the drivers shown above.
Every score is fully traceable end-to-end โ from the raw sensor reading to the engineered feature, the model version that ran, and the factors it weighed. That audit trail is what makes Access Genie AI explainable by design.