ShipNet: A Geometric Deep Learning Surrogate for Real-Time Ship Hydrodynamics

2026-06-13 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Engineering & Applied Sciences, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

ShipNet, a novel geometric deep-learning surrogate, has been developed to predict ship hydrodynamic performance, specifically hull-surface pressure distributions and far-field free-surface wave patterns, directly from hull geometry and speed. This model employs a regularized dynamic graph convolutional backbone operating on hull point clouds, coupled with a multi-head decoder for simultaneous near-body pressure and free-surface elevation outputs. The network was trained on 420 inviscid free-surface simulations, generated using a potential-flow panel method for two parent yacht hulls, each parameterized into 70 variants and evaluated at three distinct speeds. On a geometry-held-out test set, ShipNet achieved an R^2=0.98 for hull pressure and R^2=0.91 for wave fields. Crucially, inference requires approximately 0.15s per case, demonstrating over a 550x speedup compared to the potential-flow solver on conventional hardware. Current limitations include restricted geometry and speed ranges, alongside the use of inviscid training data.

Key takeaway

For naval architects and marine engineers evaluating new hull designs, ShipNet offers a significant opportunity to accelerate early-stage parametric exploration. You can utilize this geometric deep learning surrogate to obtain rapid hydrodynamic performance predictions, including hull pressure and wave patterns, at a 550x speedup over traditional potential-flow solvers. This enables far more design iterations within tight deadlines, though be mindful of its current limitations regarding geometry and speed ranges, and its reliance on inviscid data.

Key insights

ShipNet uses geometric deep learning to rapidly predict ship hydrodynamics from hull geometry, achieving 550x speedup over traditional methods.

Principles

• Geometric deep learning can surrogate complex fluid dynamics.
• Graph convolutions effectively process hull point clouds.
• Multi-head decoders enable simultaneous multi-output prediction.

Method

ShipNet employs a regularized dynamic graph convolutional backbone on hull point clouds, with a multi-head decoder predicting pressure coefficients and 2D wave elevation maps using a composite loss.

In practice

• Accelerate ship design exploration with rapid hydrodynamic predictions.
• Integrate geometric deep learning for real-time simulation surrogates.
• Reduce computational cost for parametric studies.

Topics

• Ship Hydrodynamics
• Geometric Deep Learning
• Surrogate Models
• Computational Fluid Dynamics
• Graph Neural Networks
• Hull Design Optimization

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• AI-powered design comes to high-speed racing — AI surrogate models, like GIST, can drastically accelerate complex physics simulations while maintaining high accuracy.
• Advanced Flood Prediction with Physics-Guided Deep Learning: Combining UNet, FNO, and SAR/Optical Imagery — Embedding hydrodynamic principles into deep learning models significantly improves flood prediction accuracy and physical consistency.
• A Geometry-Aware Triplane Field Network for Vehicle Aerodynamic Prediction — GTF-Net combines triplane representations with explicit geometric cues for accurate vehicle aerodynamic prediction.
• BALLAST: Bayesian Active Learning with Look-ahead Amendment for Sea-drifter Trajectories under Spatio-Temporal Vector Fields — BALLAST optimizes Lagrangian observer placement by simulating future trajectories using sampled vector fields.
• Mesh Graph Neural Network Framework for Accelerating Finite Element Simulation for Arbitrary Geometries — Mesh Graph Networks effectively generalize finite element analysis predictions across arbitrary structural geometries by encoding relative features.
• Physics-informed convolutional neural networks for fluid flow through porous media — Physics-informed CNNs accurately predict fluid flow in porous media, significantly accelerating traditional LBM simulations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.