ShipNet: A Geometric Deep Learning Surrogate for Real-Time Ship Hydrodynamics
Researchers have developed ShipNet, a geometric deep learning model designed to predict ship hydrodynamics in real-time. This surrogate model uses hull geometry and speed to approximate pressure distributions and wave patterns, offering a significant speedup over traditional computational fluid dynamics methods. ShipNet achieved high accuracy on a held-out test set, predicting hull pressure with an R^2 of 0.98 and wave fields with an R^2 of 0.91, with inference taking approximately 0.15 seconds per case. AI
IMPACT Accelerates ship design by providing rapid, accurate hydrodynamic predictions, enabling more extensive parametric exploration.