Definition
The correlation between sensor health degradation metrics (drift, noise, dropouts) and alert system effectiveness (missed, delayed, or false alerts), measuring whether sensor degradation is undermining the alerting infrastructure.
The Measurement Gap
A bearing temperature sensor drifts 8 degrees high over six months. The alert threshold absorbs the drift. When a genuine bearing overheat occurs, the delta from the drifted baseline looks smaller than it actually is. The alert triggers late or not at all. Hot bearing detector hit rate measures wayside detections, not on-board sensor accuracy. No legacy metric tracks sensor health versus alert reliability correlation.
Decision Supported:
Is sensor drift or degradation causing missed, delayed, or false alerts? Which sensors need recalibration or replacement before they compromise the decision-making layer built on top of them?
Formula
Plain language: For each sensor, compute a health score (drift, noise, data completeness). For each alert tied to that sensor, assess whether it was timely, late, missed, or false. Compute the correlation between sensor health scores and alert quality scores across the sensor fleet.
Notation:
SHACI = rho(SH_s, AQ_s) across all sensors s
Where:
SH_s = (1 - D_s/D_max) x (1 - N_s/N_max) x DC_s
AQ_s = Timely alerts / (Timely + Late + Missed + False)
Variables:
| Variable | Description | Unit |
|---|---|---|
| D_s | Measured drift for sensor s over evaluation period | Measured parameter units |
| D_max | Maximum acceptable drift before mandatory recalibration | Same units |
| N_s | Noise level for sensor s relative to baseline | Ratio |
| DC_s | Data completeness: actual readings / expected readings | Ratio (0 to 1) |
| AQ_s | Alert quality score: timely correct alerts / total alert-worthy events | Ratio (0 to 1) |
Sign Convention: Lower absolute SHACI is better. High positive correlation (sensor health down = alert quality down) is the dangerous signal.
Proposed formula, pending validation pilots.