Localized Generative Design of Ship Hulls with Diffusion Models and High-Fidelity Assessment
Abstract
Practical ship design requires localized refinement and performance evaluation. This paper presents a reward-directed diffusion framework for localized hull optimization. Parametric masking during dataset generation restricts geometric changes to selected variables, including the bulbous bow and forebody, while preserving full-hull continuity. Surrogate-assisted sequential Monte Carlo sampling targets reduced predicted calm-water resistance, with stability retained through training data and displacement matched during screening. An automated viscous free-surface CFD pipeline, validated against measurements for two benchmark container ships, evaluates a generated variant and GenHull E, a hull refined by a naval architect within the same representation. At matched KRISO Container Ship (KCS) displacement, the generated variant has comparable total resistance at Fn=0.195 but 45-50% higher resistance at Fn=0.250 and $0.260$, missed by empirical screening. GenHull E achieves 6-8% lower total resistance up to Fn=0.240 but higher resistance above Fn=0.250 as opposing friction and pressure trends reverse the gains. Aggregating resistance across this crossover can conceal tradeoffs that emerge when friction and pressure components are evaluated separately. The assessment reveals an inability to represent a faired bulbous bow and a reward lacking viscous pressure and stability terms. These findings show how high-fidelity assessment informs improved representations and objectives for physically realizable generative designs.
Related
- ScaffoldForge: Bioinspired generative design of 3D-printed bone scaffolds via conditional diffusion for large segmental bone defect reconstruction
- Beyond Distributional Fidelity: Causal-Penalized Diffusion for Synthetic Tabular Data
- Generative SDEs with Trainable Diffusion and Density-Based Losses
- Some remarks on the connection between generative diffusion and optimal transport
- SurgeGen: A Hybrid Generative Diffusion Framework for Storm Surge Scenario Synthesis
BibTeX
@misc{keramati2026localized,
title = {{Localized Generative Design of Ship Hulls with Diffusion Models and High-Fidelity Assessment}},
author = {Hadi Keramati and Mohammed A. Hannan and Patrick Kirchen and Sina Tousi and Joonas Riekkinen and Jim Pyra and Rajeev K. Jaiman},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.12870},
eprint = {2610.12870},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.12870},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Localized Generative Design of Ship Hulls with Diffusion Models and High-Fidelity Assessment
AU - Hadi Keramati
AU - Mohammed A. Hannan
AU - Patrick Kirchen
AU - Sina Tousi
AU - Joonas Riekkinen
AU - Jim Pyra
AU - Rajeev K. Jaiman
PY - 2026
DA - 2026/10/09
PB - arXiv
DO - 10.48550/arxiv.2610.12870
UR - https://arxiv.org/abs/2610.12870
AB - Practical ship design requires localized refinement and performance evaluation. This paper presents a reward-directed diffusion framework for localized hull optimization. Parametric masking during dataset generation restricts geometric changes to selected variables, including the bulbous bow and forebody, while preserving full-hull continuity. Surrogate-assisted sequential Monte Carlo sampling targets reduced predicted calm-water resistance, with stability retained through training data and displacement matched during screening. An automated viscous free-surface CFD pipeline, validated against measurements for two benchmark container ships, evaluates a generated variant and GenHull E, a hull refined by a naval architect within the same representation. At matched KRISO Container Ship (KCS) displacement, the generated variant has comparable total resistance at F_n=0.195 but 45-50% higher resistance at F_n=0.250 and $0.260$, missed by empirical screening. GenHull E achieves 6-8% lower total resistance up to F_n=0.240 but higher resistance above F_n=0.250 as opposing friction and pressure trends reverse the gains. Aggregating resistance across this crossover can conceal tradeoffs that emerge when friction and pressure components are evaluated separately. The assessment reveals an inability to represent a faired bulbous bow and a reward lacking viscous pressure and stability terms. These findings show how high-fidelity assessment informs improved representations and objectives for physically realizable generative designs.
ER -