Structural Twin Prediction Accuracy (STPA)

GMI-MAR-026Maritime IndustriesInfrastructure-Dependent

Definition

Accuracy of structural twin predictions compared to actual thickness measurements.

The Measurement Gap

What This Metric Measures

STPA measures the accuracy of structural digital twin predictions when compared to actual physical measurements from thickness gauging, tank surveys, and close-up inspections at the next survey event. The metric quantifies how well the twin's structural model predicts real-world plate thickness, corrosion rates, and fatigue life by comparing predicted values to measured values at survey checkpoints.

The Decision It Supports

Can the organization use structural twin predictions to justify extending survey intervals or deferring dry-dock, or is the twin's structural model insufficiently validated against physical reality? Classification societies are beginning to define acceptable twin accuracy thresholds for condition-based class, and STPA provides the metric they will need to evaluate whether a twin is trustworthy enough to replace physical survey frequency.

Why Legacy Metrics Cannot See This

Legacy structural monitoring tracks plate thickness measurements at each survey and computes corrosion rates between surveys. These metrics describe what happened. They do not evaluate whether the digital twin's prediction of what would happen was accurate. An organization might defer a physical survey based on twin predictions showing adequate plate thickness, but without STPA, nobody validates whether those predictions were correct at the last survey. The twin could be systematically overestimating plate thickness by 10%, and the error would remain invisible until a surveyor measures a plate that the twin predicted was safe but is actually approaching minimum thickness. STPA closes this validation gap by requiring every survey event to serve as a twin accuracy checkpoint.

Formula

Primary Formula

STPA = 1 − (1/N) × Σ |Predicted_i − Measured_i| / Measured_i

Variable Definitions

  • Predicted_i = structural twin's predicted value for measurement point i (mm thickness, mm/year corrosion rate, or fatigue life estimate)
  • Measured_i = actual physical measurement at point i from ultrasonic thickness gauging or close-up inspection
  • N = number of measurement comparison points across the survey event

Units: 0 to 1. A value of 1.0 indicates perfect prediction. Measure per structural area: bottom plating, side shell, ballast tanks, and deck structures should each carry their own STPA value.

Edge Cases

Localized corrosion not captured by the twin's uniform-rate model represents the most dangerous error. Measure STPA separately for uniform areas and localized corrosion hotspots. Surveys where no physical measurement was taken (visual-only) cannot produce STPA: flag as validation gaps. Minimum sample of 20 measurement comparison points per survey event required.

Confidence Grading (required alongside every STPA value):

Grade Basis Interpretation
A: Empirical 5+ observed events in measurement window provide direct tail observation Measurement grounded in observed outcomes; highest confidence
B: Hybrid 1-4 observed events plus Monte Carlo simulation with Bayesian updating Blends observation and simulation; moderate confidence
C: Simulated Zero observed events; derived entirely from Monte Carlo simulation Modeled, not observed; treat as indicative, not validated

A STPA of any value at Grade C is a simulation that agrees with itself. It tells you the model is internally consistent but not that it matches reality. Grade C scores should trigger proactive stress-testing against historical near-miss events and cross-asset analogues rather than acceptance as calibrated.

Proposed formula, pending validation pilots.

At a Glance

Unit
Score (0 to 1)
Trend Direction
Toward 1.0
Reporting Frequency
Per survey with physical measurements
Category
Digital Twin Fidelity & Model Validation
← Biofouling Assessment Calibration (BAC)AI Dry-Dock Scope Completeness (ADDSC) →

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