Uncertainty Quantification Within ITTC Full-Scale Extrapolation Procedure for Powering Performance Prediction Using Regression-Based Non-Intrusive Polynomial Chaos Method

Accurate prediction of full-scale ship performance through the ITTC (International Towing Tank Conference) extrapolation procedure is crucial for ship design and energy assessment, yet it is subject to uncertainties from extrapolation coefficients, correlation factors, and force measurements. A regression-based Non-Intrusive Polynomial Chaos (NIPC) method is developed to perform a comprehensive uncertainty quantification. The analysis propagates uncertainties from extrapolation coefficients (ship form factor, hull and propeller roughness), the wetted surface area, and model-scale hydrodynamic forces (resistance, propeller thrust and torque) of the ship. Statistical metrics for resistance and self-propulsion performances are computed and validated against Monte Carlo simulations, followed by a global sensitivity analysis using Sobol indices. Results show the coefficient of variation (CV) for delivered power is approximately 4% at the designed speed, indicating a significant prediction unreliability, whereas predictions for propulsive efficiency and propeller shaft speed are more reliable (CV < 1%). The sensitivity analysis identifies hull roughness as the dominant uncertainty source for power predictions and reveals that, under the studied conditions, the prediction process for key outputs is predominantly linear. The study concludes that while the power prediction carries notable uncertainty, the procedure yields reliable estimates for propulsive efficiency, providing a quantified risk assessment framework to guide robust ship design and performance evaluation.