Adjoint RANS with filtered shape derivatives for hydrodynamic optimisation

Abstract

The adjoint RANS method has been implemented in the framework of an unstructured general-purpose finite volume code following the derive-then-discretise strategy (continuous adjoint approach). An explicit filtering technique is applied to the shape derivatives in order to extract noise from the mesh-based representation of high-resolution. In a CAD-free optimisation strategy the method is applied to a semi-circular profile in incompressible, external flow. It was observed that the filtering is particularly relevant at high Reynolds-numbers. In the complex scenario of ship hull design, the filtered shape derivative can directly be applied to the base-line configuration in order to support both manual and automatic shape optimisation.