Monday, February 17, 2020       [ login ]
International Centre for
Numerical Methods in Engineering

events

Coffee Talk - "Sensitivity analysis of embedded bodies using the adjoint method and applied to airfoil optimization for the full-potential equation", by Marc Núñez

Published: 13/01/2020

Wednesday, January 22th, 2020. Time: 15h.

Place: O.C. Zienkiewicz Conference Room, C1 Building, UPC Campus Nord, Barcelona

ABSTRACT

In optimization, different methods or techniques can be considered depending on the nature and characteristics of the studied problem. In the context of large-scale cases where the computational cost is high, gradient-based optimization algorithms stand as an effective approach to advance in the design space.

Nonetheless, gradient-based optimization problems require the computation of the sensitivity of the objective functions with respect to the design parameters,  which is also costly if the number of parameters is high. For this reason, the adjoint method is used to accurately compute the gradient of cases where the number of design parameters is much higher than the number of objective functions, with a total cost that is roughly equivalent to the original problem in steady cases.

Additionally, when considering the analysis of large and complex structures that change its shape, standard body-fitted meshes may present difficulties in the pre-processing and post-processing phases, as well as when dealing with the shape movement of the studied body. In this sense, embedded geometries are utilized to overcome these problems at the expense of some analysis accuracy.

In this talk, the adjoint method will be introduced using embedded geometries and applied to airfoil optimization using the full-potential equation.

SPEAKER CV

Marc Núñez is an aerospace engineer from the UPC (ESEIAAT) and a PhD student at CIMNE. He is part of the research group KratosMultiphysics and his research is focused on shape optimization problems using embedded bodies. His work is carried out in the framework of the ExaQUte project.