Shape Optimization of Sailing Yacht Keel Bulb Using RANS Simulation

Sammanfattning

Shape optimisation has become a key tool for improving hydrodynamic performance and reducing drag in modern marine architecture. By systematically exploring the influence of geometric parameters on flow behaviour, optimisation loops enable designers to identify efficient configurations while accounting for physical and operational constraints. However, when high-fidelity Computational Fluid Dynamics (CFD) simulations are involved, the associated computational cost can become a major limitation. In this context, surrogate-based optimisation (SBO) offers an efficient framework by replacing a large number of expensive simulations with a predictive model trained on a limited dataset, thereby making CFD-driven optimisation more suitable for industrial applications. An illustrative and relevant application of this approach is the shape optimisation of a sailing yacht keel bulb. With the increasing use of hydrofoils on modern racing yachts, the relative contribution of appendages to the total drag has become more significant. Optimising the keel bulb geometry is therefore of particular interest for performance improvement. Moreover, this problem can be formulated with a limited number of design variables while still capturing meaningful hydrodynamic effects, making it a suitable case study for the development and validation of an optimisation framework. The objective of this thesis is to develop a complete and automated optimisation loop for the CFD-based shape optimisation of an IMOCA yacht keel bulb. The work includes the creation of a parametric geometric modeller, the automation of Reynolds-Averaged Navier--Stokes (RANS) simulations, and the implementation of a Gaussian Process Regressor (GPR) surrogate model. Special attention is given to the selection of initial and adaptive sampling strategies in order to minimise the number of CFD evaluations while preserving surrogate accuracy. The optimisation is performed under a volume constraint, with the aim of reducing hydrodynamic drag. The resulting methodology is intended to provide a robust and reusable framework that can be extended to more complex marine optimisation problems.