Definition
A score reflecting the estimated accuracy of an autonomous inspection result based on the environmental conditions during data collection, allowing organizations to weight inspection results by the confidence they deserve.
The Measurement Gap
Autonomous inspection accuracy degrades under poor lighting, precipitation, temperature extremes, surface moisture, and platform vibration. But no metric flags when an inspection was conducted under conditions that reduce detection reliability. A drone inspection of a bridge deck during rain gets the same confidence weight as one conducted in clear conditions. When environmental conditions reduce detection accuracy, the inspection may create false assurance rather than genuine coverage.
Decision Supported:
Should we treat this inspection result with full confidence, or were conditions during collection bad enough to flag it for verification or re-inspection?
Formula
Plain language: For each autonomous inspection event, score environmental conditions against validated accuracy curves. Each condition factor (lighting, precipitation, temperature, surface moisture, vibration) produces a confidence multiplier between 0 and 1. Multiply them for the overall EICI.
Notation:
EICI_j = PRODUCT(CF_k_j) for k in {lighting, precipitation, temperature, surface_moisture, vibration}
Fleet average: EICI_avg = SUM(EICI_j) / N_inspections
Variables:
| Variable | Description | Unit |
|---|---|---|
| CF_lighting | Confidence multiplier based on ambient light level | 0 to 1 |
| CF_precipitation | Confidence multiplier based on precipitation intensity | 0 to 1 |
| CF_temperature | Confidence multiplier based on temperature extremes | 0 to 1 |
| CF_surface_moisture | Confidence multiplier based on surface condensation/moisture | 0 to 1 |
| CF_vibration | Confidence multiplier based on platform vibration during inspection | 0 to 1 |
Sign Convention: Higher is better. 1.0 = perfect conditions. Multiplied factors mean poor conditions in any dimension degrade the score.
Proposed formula, pending validation pilots.