Definition
Mean absolute error of CV defect measurements compared to manual measurements under marine conditions.
The Measurement Gap
What This Metric Measures
CVMAMC measures the mean absolute error of CV-based defect measurements (crack length, corrosion area, coating breakdown percentage, pit depth) compared to manual measurement by a qualified inspector under actual marine inspection conditions. It quantifies whether the CV system produces accurate enough measurements to support severity classification decisions in the dirty, dark, vibrating environments where marine inspections actually occur.
The Decision It Supports
Can CV defect measurements be trusted for severity classification under actual marine inspection conditions, or must manual measurement confirmation be required for every severity-critical finding? If CVMAMC shows the CV system measures a 4mm crack as 2mm because fouling partially obscures the crack tip, the severity classification based on the CV measurement would say "monitor" when the actual measurement says "replace immediately."
Why Legacy Metrics Cannot See This
Legacy inspection metrics track defect counts, inspection completion rates, and defect severity distributions. No metric validates whether the CV measurement of a specific defect is accurate against manual measurement. The CV system reports a crack length of 2mm. The inspector files this as a 2mm crack. Nobody checks whether it is actually 4mm because the CV measurement is treated as the measurement, not as an estimate requiring validation. CVMAMC requires paired CV and manual measurements on the same defects to quantify the measurement error that affects every downstream severity classification, maintenance decision, and class survey finding.
Formula
Primary Formula
CVMAMC = 1 − (1/N) × Σ |CV_measurement_i − Manual_measurement_i| / Manual_measurement_i
Variable Definitions
- CV_measurement_i = CV system's defect measurement at point i (mm for cracks, cm² for corrosion area, % for coating breakdown, mm for pit depth)
- Manual_measurement_i = qualified inspector's manual measurement using calibrated instruments at the same point
- N = number of measurement comparison points
Units: 0 to 1 scale. A value of 1.0 indicates perfect measurement accuracy. Segment by defect type (crack, corrosion, coating breakdown), equipment type, and inspection condition quality to identify where the CV system is accurate and where it falls short.
Edge Cases
Manual measurement is also subject to error: treat manual as the reference standard rather than absolute ground truth, and acknowledge the measurement uncertainty in both methods. Some defects can only be measured manually (depth gauging with ultrasonic thickness measurement) where CV provides area or length measurement but not depth for subsurface defects. Minimum sample of 50 measurement comparisons per defect type across the fleet for statistically meaningful accuracy assessment.
Proposed formula, pending validation pilots.