Adjoint Optimization combined with mesh morphing for CFD application
CHALLENGE - This study aims to investigate the applicability of Adjoint optimization combined with mesh morphing to the industrial practice, by the integration of commonly used industrial-grade simulation software. Adjoint techniques are efficient optimization methods in terms of accuracy of results and short computational cost, but normally are limited to in-house simulation codes, that allow the calculation of partial derivatives of the observable quantities within the model simulation.
SOLUTION - Quite recently, some CAE packages introduced this capability within their solvers. This is the case of ANSYS Fluent, which makes available the derivatives of a given observable (objective) as a function of mesh points coordinates. This, combined with a mesh morphing tool such as RBF Morph, allows to automatically compute the derivatives in function of design parameters, such as the amplifications of RBF solutions that control the shape of the mesh.Integrating these software in the optimization platform modeFRONTIER, it becomes possible to apply efficient gradient based optimization algorithms, with the big advantage of a full automatic process integration, the possibility to exploit the post-processing tools available in modeFRONTIER, and a very short number of design simulations to optimize the objective function.
BENEFITS - The advantage of the approach, which has proved to be highly competitive in terms of simulation time and efficiency of results compared to other classic approaches, relies on the fact that it can be applied linking together CAE software of large industrial diffusion, such as ANSYS, RBF Morph and modeFRONTIER, and it is not an intrusive method requiring to modify simulation codes, but is very simple to implement for any industrial application just using the available GUI interfaces.