This is the authors' abstract. We don't add key points for this paper.
Traditional prediction methods for predicting ship resistance, such as computational fluid dynamics (CFD), are computationally costly and inefficient, severely limiting ship form optimization efficiency.To address this bottleneck, this paper proposes a novel physics-informed dynamic regularization extreme gradient boosting (PIDR-XGBoost) model, which is further coupled with the Harris hawk optimization (HHO) algorithm to realize high-precision resistance prediction and efficient optimization. Firstly, high-reliability sample data were generated via CFD. Subsequently, a calm water resistance prediction model was constructed based on the PIDR-XGBoost method. This model innovatively embeds a dual-exponential dynamic regularization module, which adjusts the regularization strength in real time according to the local sample density and iterative fitting residuals, fundamentally solving the over-constraint/under-constraint dilemma. Meanwhile, a hydrodynamic mechanism-based physical deviation penalty term is introduced into the loss function, thereby addressing the deficiency of physical interpretability and the tendency to deviate from actual hydrodynamic laws inherent in conventional machine learning models. PIDR-XGBoost model reduces the root mean square error (RMSE) by 36.920% and the mean absolute error (MAE) by 58.473%, while increasing the determination coefficient (R²) by 2.793% compared with the traditional XGBoost model, providing high-precision and physically interpretable prediction support for subsequent ship form optimization. Finally, the CFD validation of the optimized ship form demonstrates that the total calm water resistance is reduced by 4.358%, with the error between the PIDR-XGBoost predicted value and the CFD validation value as low as 0.404%. The proposed integrated framework of PIDR-XGBoost and HHO breaks through the limitations of high computational cost, and can be directly applied to hull optimization and ship energy consumption assessment, exhibiting significant engineering application value.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: