Interactive Reduced Order Models for Ship Hull Design and Optimization Coupling an Open-Source CFD Tool with Advanced RBF Mesh Morphing

Overview

This technical paper presents a proof-of-concept numerical procedure for ship hull design and optimization using Reduced Order Models (ROMs) derived from computational fluid dynamics simulations. The approach combines open-source CFD tools with Radial Basis Function (RBF) mesh morphing techniques to efficiently explore the design space of hull geometries.

The methodology enables rapid prediction of hydrodynamic performance by constructing ROMs from a database of CFD simulations generated through a Design of Experiments (DOE) study. This approach significantly reduces computational costs while maintaining strong predictive accuracy, supporting more efficient simulation-driven ship design workflows.

Abstract

We present the development of the proof-of-concept (PoC) for an effective and efficient numerical procedure conceived to design and optimize ship hulls by leveraging reduced order models (ROMs) gained processing a database of computational fluid dynamics (CFD) fields relevant to the marine sector that is created by performing a design of experiments (DOE) study parametrizing the CFD case using radial basis functions (RBF) mesh morphing. The ROMs, that can be interactively used to get real-time calculations, demonstrated a maximum percentage error on prediction of less than 3.5% for resistance estimates at design points outside the original DOE set, indicating strong predictive accuracy. The resulting ROMs serve as a foundational step toward building Digital Twins of ships.

What This Paper Covers

• Reduced Order Models (ROMs) for ship hull design and optimization
• Coupling open-source CFD simulations with RBF mesh morphing
• Design of Experiments (DOE) strategies for generating CFD databases
• Proper Orthogonal Decomposition (POD) for model reduction
• Real-time prediction of hull resistance and CFD fields
• Interactive visualization of simulation results through a dashboard