Abstract
A well-formulated design space parametrisation is the key to the success of design optimisation. Most parametrisation methods require manual set-up which typically results in a restricted design space and impedes the generation of superior designs which may be found outside this restricted envelope. In this work, we adopt a NURBS-based automatic and adaptive parametrisation approach where the optimisation begins in a coarser design space and adapts to finer parametrisation during the optimisation. Our approach takes CAD descriptions as input and to alter the shape perturbs the control points of the NURBS patches that form the boundary representation. Driven by adjoint sensitivity information the control net is adaptively enriched using knot insertion. The sensitivity-driven parametrisation method is applied here to reduce the pressure loss of a U-bend passage of a turbine blade serpentine cooling channel.