Skip to main content
Menu

Research

Adjoint-based multi-level and multi-physics methods for shape optimisation and error estimation.

Principal investigator: Jens-Dominik MUELLER

Adjoint-based sensitivity analysis is an essential ingredient for use of CFD in shape optimisation, error estimation, mesh adaptation, uncertainty quantification and may aspects of flow control. The group at QMUL is at the forefront of developing efficient adjoint and tangent sensitivity tools using automatic differentiation (AD) software tools for a variety of applications. The work of the group is funded by Rolls Royce, as well as through the EC research projects FlowHead, About Flow and IODA. The current research focus is on the development of

  • general and robust multi-purpose adjoint solvers with fully automated adjoint code derivation (Christakopoulos 2011, Jones 2011)

  • automatic multi-purpose geometric parametrisation toolboxes suitable of generalist industrial applications with geometric constraints, (Yu 2011, Xu 2014, Xu 2015)

  • multi-physics adjoint coupling for multi-disciplinary optimisation of strongly coupled problems, (Forsythe 2008)

  • stable and efficient fixed-point iterative schemes for compressible and incompressible flow discretisations, (Cusdin 2006, Christakopoulos 2011, Xu 2015, Akbarzadeh 2015)

  • multi-level methods for adaptive mesh refinement and spatio-temporal filtering of flow fields for unsteady (LES/DES) adjoint sensitivity analysis. (Xu 2015, Hueckelheim 2015)

More detailed information on some of the projects of the group members can be found on the webpages for About Flow and IODA. The activities of the group are embedded in a strong network of European and international collaborations with academic and industrial collaborators, as present in these projects.

 

The group seeks a PostDoc to join for two years to contribute to its core research activities. You should have a good knowledge of CFD and adjoint methods and an experience in one or several of shape optimisation, mesh adaptation, LES/DES, higher-order methods, multi-disciplinary coupling. Please don't hesitate to contact Dr. J.-D. M?ller (j.mueller@qmul.ac.uk).