Accelerations and added-resistance predictions using machine learning for planing hulls

While vertical motions and added resistance in displacement ships can generally be predicted following a linear pattern based on the sea spectrum using the RAO method, high-speed vessels operating in irregular waves exhibit nonlinear seakeeping that requires alternative semi-empirical predictive approaches at early design phases. These nonlinearities stem from vessel geometry, speed, and wave spectrum characteristics. This study performs the prediction of vertical acceleration and added resistance statistics in planing hulls using five machine learning methods, including neural networks. The dataset includes experimental series in irregular waves conducted by Fridsma, extended cases from Brown, and further expansions by Zarnick and Turner. These data-driven models outperform traditional empirical methods, such as those by Savitsky and Brown, particularly under high-speed conditions where nonlinear effects dominate. Machine learning emerges as a robust tool for preliminary design assessments, offering enhanced prediction capability without requiring explicit L/B input, thereby increasing model generalizability.