Journal - Jin-Hyeok Kim, Myung-Il Roh, In-Chang Yeo, “An MLP (Multi-Layer Perceptron)-Based Method for Optimizing Multiple Hull Forms at Once Using Surrogate Modeling with Geometric Moments”, Ocean Engineering, Vol. 339, Part 1, 2025.11.15

Extra Form
Abstract	The hydrodynamic performance of a hull form is a crucial factor in the design process, and optimizing it typically requires computationally expensive and time-consuming CFD analyses. To overcome this issue, numerous approaches have been explored to efficiently handle hull forms characterized by high-dimensional parameter spaces and mitigate the associated computational complexity. In this study, we proposed a method that enables efficient optimization without restricting the degrees of freedom of the hull form. First, we introduced a hull form generation approach utilizing an MLP, allowing each vertex of the hull mesh to serve as a design variable and move freely. By directly incorporating hydrodynamic performance-related terms into the loss function of the MLP, our method enables effective searches for optimal solutions even in high-dimensional design spaces. Additionally, we considered design constraints, such as block coefficient (CB) and longitudinal center of buoyancy (LCB), and hull form surface quality, such as fairness, throughout the hull form generation and optimization process. To evaluate the hydrodynamic performance of hull forms efficiently, we constructed and employed a surrogate model that incorporates a dimensionality reduction technique based on geometric moments. Finally, we applied the proposed hull form optimization method to the publicly available KCS (KRISO Container Ship) hull form. The results demonstrated that our approach successfully explores hull forms that enhance hydrodynamic performance while leveraging high-dimensional design variables.
Publication Date	2025-11-15
Role	Corresponding Author
Category	SCIE

Abstract

The hydrodynamic performance of a hull form is a crucial factor in the design process, and optimizing it typically requires computationally expensive and time-consuming CFD analyses. To overcome this issue, numerous approaches have been explored to efficiently handle hull forms characterized by high-dimensional parameter spaces and mitigate the associated computational complexity. In this study, we proposed a method that enables efficient optimization without restricting the degrees of freedom of the hull form. First, we introduced a hull form generation approach utilizing an MLP, allowing each vertex of the hull mesh to serve as a design variable and move freely. By directly incorporating hydrodynamic performance-related terms into the loss function of the MLP, our method enables effective searches for optimal solutions even in high-dimensional design spaces. Additionally, we considered design constraints, such as block coefficient (CB) and longitudinal center of buoyancy (LCB), and hull form surface quality, such as fairness, throughout the hull form generation and optimization process. To evaluate the hydrodynamic performance of hull forms efficiently, we constructed and employed a surrogate model that incorporates a dimensionality reduction technique based on geometric moments. Finally, we applied the proposed hull form optimization method to the publicly available KCS (KRISO Container Ship) hull form. The results demonstrated that our approach successfully explores hull forms that enhance hydrodynamic performance while leveraging high-dimensional design variables.

Publication Date

2025-11-15

Role

Corresponding Author

Jin-Hyeok Kim, Myung-Il Roh, In-Chang Yeo, “An MLP (Multi-Layer Perceptron)-Based Method for Optimizing Multiple Hull Forms at Once Using Surrogate Modeling with Geometric Moments”, Ocean Engineering, Vol. 339, Part 1, 2025.11.15

단축키

Articles