Enhancing ship hydrodynamic performance via machine learning-driven CFD parametric optimization

Jie Liu

doi:10.1080/19942060.2025.2565801

Enhancing ship hydrodynamic performance via machine learning-driven CFD parametric optimization

2025-12-31

0

PRE

AI

DOI：10.1080/19942060.2025.2565801

Original

Original request for help

Save

Abstract

En 中文

In order to improve the efficiency of ship hydrodynamic optimization and reduce resistance, this study employed machine learning methods to predict resistance. In response to the limitations of machine learning's generalization ability on high-dimensional sparse data, an IXGB-MAML hybrid model based on improved eXtreme Gradient Boosting (XGBoost) and model-agnostic meta-learning strategy methods was proposed. The predictive ability of the hybrid model in regression tasks was tested based on standard functions. Subsequently, the proposed hybrid method was used to train a resistance prediction model based on computational fluid dynamics (CFD) simulation data, and its prediction accuracy was verified through error analysis and performance evaluation. Therefore, a ship form optimization design framework is constructed by integrating the prediction, optimization, and deformation modules. The prediction results of machine learning were used to guide the optimization search to achieve a nonlinear optimization of ship form. The results indicate that the hybrid method performs excellently in resistance prediction, and the resistance of the optimized ship form in both calm water and waves is reduced, thereby demonstrating good hydrodynamic potential. The IXGB-MAML method can effectively reduce the computational cost, and is helpful in promoting the application of intelligent optimization methods in ship design.

Keywords:

machine learning

multi-objective optimization

hydrodynamic

ship form

CFD

Enhancing ship hydrodynamic performance via machine learning-driven CFD parametric optimization

Abstract

Journal

Organization