Data-driven modeling and regression analysis on ship resistance of in-service performance

This study employs operational data to model ship resistance, aiming to bridge the gap between controlled experiments and real-world conditions. It comprehensively analyzes wind, waves, and currents, employing nonlinear regression and z-score filtering. The model is validated using data from three identically designed ships operating on the similar servicevoyages. Key findings reveal significant impacts of wind and waves on the added resistance, variability in resistance across different loading conditions, and discrepancies between in-service performance and model test results, especially at medium to low speeds. Calm water resistance results are reliable, varying within 5%-10% of the average, though in-service performance is generally higher, indicating a need for further research. The added resistance due to wind is significant, with variations within 5%-10%, and the transverse projected area does not always proportionally affect resistance. Head winds have a greater impact on resistance than following winds at the same speed. The analysis of added resistance due to waves shows significant, but sometimes inconsistent, transfer function coefficients, suggesting simpler model structures could be more effective. The added resistance due to current if found to typically fall within a 2-3% range, indicating that significant changes are rare and localized. For large ships, short waves dominate, with resistance increasing proportionally with the non-dimensionalized wave length. While head currents can increase resistance by up to 20% and following currents can reduce it by 5-10%, these larger changes are infrequent. Segmenting data by loading conditions, routes, and speeds improves regression analysis accuracy, though excessive segmentation reduces data diversity and reliability.