Adjoint system-based optimisation of ydrofoils using algorithmic differentiation

Abstract

Hydrofoils are the biggest performance differentiators on modern racing yachts. They significantly affect the speed, the state and the trim of the yacht system. This requires detailed optimisation and the modelling of the whole system during the optimisation. However, due to the high computational cost, hydrofoils are still optimised in isolation of the yacht system with most studies using only a small number of design variables. This introduces inaccuracies and leaves considerable potential inaccessible.
Therefore, a stationary physics model of an entire yacht is developed. It includes an advanced lifting line method for hydrofoil force prediction and a detailed parametric model of the hydrofoil with 68 design variables. As this is too computationally expensive to optimise with conventional strategies, the physics model is integrated into a gradient-based optimisation routine, where the gradient is computed with the adjoint method. The adjoint method can compute the gradient at small cost, independent of the number of design variables, and is the key to detailed design. The adjoint method is only applied to the bottleneck of the physics model using algorithmic differentiation. The remainder of the model is differentiated with the easier to implement finite difference method. The combined gradients are provided to an optimisation algorithm.
The framework is used to optimise the hydrofoils of an AC75 America’s Cup yacht. It is validated with a parametric study and literature. A single-condition optimisation of the hydrofoil is performed for upwind

More information

Identifiers

ORCID for Rafael Andreas Maximilian Tannenberg: orcid.org/0009-0004-4250-1361

ORCID for Stephen Turnock: orcid.org/0000-0001-6288-0400

Catalogue record

Export record