Hull Optimization for Fast and Efficient Semi-Planing Craft using RANS CFD combined with the Volume of Fluid Method

Details

Supervisors

Winckelmans, Grégoire

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil mécanicien, à finalité spécialisée

Abstract

enThis thesis presents a fully automated computational fluid dynamics (CFD) optimization pipeline developed for the hull form of a fast semi-planing electric yacht (Lumen Yachts, Lwl = 7m, Fr = 0.796, corresponding to twice the hull speed). The pipeline integrates parametric hull generation through Grasshopper and Rhino.Compute, fluid domain construction in FreeCAD and SpaceClaim, mesh generation and flow simulation in ANSYS Meshing and Fluent, and result extraction via Python scripts. It constitutes an optimization tool: any geometric parameter can be targeted and the entire evaluation chain runs without manual intervention. The hull is a 7-metre chined semi-planing craft, operating in a regime where both wave-making and dynamic lift are significant. The flow is modelled using Reynolds-Averaged Navier-Stokes (RANS) equations with the realizable k-ε turbulence model and scalable wall functions, combined with the Volume of Fluid (VOF) method for free-surface treatment. A mesh sensitivity study confirms that the 2.7% drag deviation between the selected and finest meshes is small compared to the drag variations between hull variants, placing the selected mesh in a suitable operating range for comparative hull evaluation. A one-dimensional Sobol exploration over the stern chine position, the parameter identified as the most influential on drag, reveals a flat optimal zone Chine_Y ∈ [0.058, 0.080] where the mean drag is 6.7% below the base design. Within this zone, drag differences between hull variants fall below the numerical noise floor, so no single point can be identified as the unique optimum. The nominal minimum (Chine_Y = 0.071, D = 1359.8N, −7.7% relative to base) is within this uncertainty and should be understood as one representative point in the optimal zone rather than an individually identified optimum. The optimization infrastructure supports multi-parameter Bayesian optimization, demonstrated on preliminary runs, but the Sobol phase alone was sufficient to characterize the 1D drag landscape. The results confirm that hull geometry at the chine has a strong and non-trivial influence on resistance in the semi-planing regime. The automated framework developed here provides a reusable foundation for future optimization campaigns with larger computational budgets.