We are pleased to announce that the paper we co-authored with ENGYS, titled “Interactive Reduced Order Models for Ship Hull Design and Optimization,” and presented at COMPIT 2025, is now available for download on our website.
The 23rd Conference on Computer Applications and Information Technology in the Maritime Industries (COMPIT 2025) took place in the beautiful setting of Pontignano, Italy, on 7–8 October 2025. This international event brought together experts and innovators from across the maritime sector to discuss the latest developments in simulation, automation, artificial intelligence, and digitalization for marine design and operations.
Our contribution to this year’s conference focused on developing a proof-of-concept framework for ship hull design and optimization that blends open-source computational fluid dynamics (CFD) with state-of-the-art data-driven modeling techniques. The study demonstrates how Reduced Order Models (ROMs) can dramatically accelerate the hydrodynamic evaluation of ship designs while maintaining high accuracy. The method we proposed builds on a database of CFD simulations parameterized through Radial Basis Function (RBF) mesh morphing, a technique that enables smooth and flexible shape modifications of the hull geometry. These simulations were processed using the Proper Orthogonal Decomposition (POD) method to construct ROMs capable of predicting resistance, pressure distribution, and free surface elevation in real time.
Unlike traditional CFD-based design workflows that require significant computational resources, our reduced-order modeling approach allows designers and engineers to explore the design space interactively. Through a dedicated PyVista-based dashboard, users can visualize the effects of shape variations instantly, adjusting RBF parameters to observe immediate updates to hull performance predictions. The framework’s efficiency and speed make it especially well-suited for the early stages of ship design, when design decisions must be informed quickly and iteratively. Validation results show that the ROMs achieve an impressive maximum prediction error below 3.5% for resistance estimates at design points not included in the training dataset. This level of accuracy highlights the potential of data-driven surrogates to support fast and reliable hydrodynamic evaluations. The entire process is further streamlined by Python-based automation, which manages the generation of design points, CFD simulations, and data processing steps with minimal manual intervention.
Beyond demonstrating computational efficiency, this work represents an important step toward building Digital Twins for ships — virtual replicas that can continuously learn from data and assist in performance monitoring, optimization, and decision-making throughout a vessel’s lifecycle. By integrating open-source CFD software with advanced numerical modeling, the framework offers both flexibility and cost-effectiveness, enabling scalable studies without proprietary licensing constraints.
You can now download the full paper and explore how this interactive, automated approach is helping redefine the future of simulation-based ship design.