CAD-free hydrodynamic optimisation using consistent kernel-based sensitivity filtering

摘要

This paper advocates the benefits of consistent kernel-based filter functions dedicated to the filtering of shape sensitivities obtained from CAD-free adjoint hydrodynamic shape optimisation procedures. Emphasis is given to explicit corrections of truncated Gaussian filters in discrete space subjected to first-order consistency constraints. The nature of the approach supports an irregular distribution of surface-mesh points without modifying the shape sensitivity influences. The methodology facilitates an apparently simple explicit Lagrangian surface parameterisation at a moderate computational surplus. Results indicate benefits of a consistent approach as regards the conservation of surface-mesh properties and the approach towards optima. Examples indicate the practical impact for aftship and wake-equalising duct examples. Accordingly, O(150) designs that feature large (local) deformations of the initial shape can be examined during a few hours of turn-around time using a single tool rather than a tool chain.