Research Fellows
Pavlos Alexias
Early Stage Researcher 4 at Engys IT
In the absence of parametrization, the surface nodes of the CFD mesh can be used
as design variables. The latter approach offers the richest possible design space (for
the given spatial discretization) allowing the generation of better optimum solutions
and the implementation of high-complexity constraints. However, any numerical
noise in the adjoint derivatives, combined with the fact that each surface node is
being perturbed independently from its neighbors, can create oscillations and irregularities.
To avoid reducing the smoothness of the surface and negatively impacting
the convergence of the optimization problem it is necessary to generate a smooth
representation of the gradient.
In the following figure, the difference between raw and smooth sensitivities is demonstrated.
