Operating Condition-Adjusted RUL Accuracy (OCRA)

GMI-MAR-019Maritime IndustriesIntegration-Required

Definition

Accuracy of RUL predictions when operating conditions change versus when they remain stable.

The Measurement Gap

What This Metric Measures

OCRA measures the accuracy of remaining useful life predictions when the equipment's operating conditions during the RUL window differ from the conditions at prediction time, compared to accuracy when conditions remain stable. It groups RUL predictions into stable-condition and changed-condition cohorts and compares prediction error rates between them, quantifying exactly how much operating condition variability degrades the model.

The Decision It Supports

Can the chief engineer trust a 2,000-hour RUL estimate when the vessel is about to change fuel type, trade route, or loading condition, or should the estimate be discounted when operating conditions are expected to change? A turbocharger RUL predicted during a low-sulfur distillate fuel leg may be accurate for continued distillate operation but dangerously optimistic when the vessel switches to heavy fuel oil, which changes exhaust temperatures, fouling rates, and turbocharger loading. OCRA quantifies this discount factor.

Why Legacy Metrics Cannot See This

MTBF measures historical failure intervals. RUL accuracy is typically reported as a fleet-wide average. Neither metric accounts for future operating condition changes that invalidate the prediction basis. A PdM system can show 90% RUL accuracy on its dashboard (computed across all predictions) while being 50% accurate on predictions where operating conditions changed after the prediction was issued. OCRA separates these populations and reveals the true reliability of RUL estimates under the variable conditions that are normal in maritime operations.

Formula

Primary Formula

OCRA = MAPE_stable / MAPE_changed

Where:
MAPE = (1/N) x SUM(|RUL_predicted - RUL_actual| / RUL_actual)

Variable Definitions

  • MAPE_stable = mean absolute percentage error of RUL predictions where operating conditions remained stable during the prediction window
  • MAPE_changed = mean absolute percentage error of RUL predictions where operating conditions changed materially during the prediction window
  • RUL_predicted = predicted remaining running hours to failure or maintenance need at prediction time
  • RUL_actual = actual running hours from prediction to failure or maintenance need
  • Material condition change = fuel type change (always qualifies), average load change exceeding 20%, or trade route change producing more than 10 degrees C mean sea temperature difference

Units: Ratio. A value of 1.0 means condition changes do not affect accuracy. Values below 1.0 indicate predictions under changed conditions are less accurate (expected). The gap between 1.0 and actual OCRA quantifies the degradation.

Edge Cases

Equipment that failed within the RUL window for reasons unrelated to the predicted failure mode should be excluded. Minimum sample of 20 RUL predictions per group (stable and changed) per equipment type is required for statistical validity.

Proposed formula, pending validation pilots.

At a Glance

Unit
Ratio
Trend Direction
Toward 1.0
Reporting Frequency
Rolling semi-annually
Category
Predictive Maintenance and Deferral Risk
← Fleet Model Accuracy Segmentation (FMAS)Cargo Fitness Inspection Coverage Index (CFICI) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment