ESR4 | Kaloyan Kirilov

Adjoint-based mesh refinement for high-order flow simulations

ESR4 | Kaloyan Kirilov

Imperial College London

Host Institution: Imperial College London

Phd awarding institution: Imperial College London

Master Title: Advanced Computational Methods for Aeronautics

Research Interest: Mesh Generation, Adaptive Mesh Refinement, Spectral/hp Element Method, Topology Optimisation

To achieve one of the main goals of the SSECOID project, a framework supporting high-fidelity fluid simulations for industrial application will be created. These are usually very complex geometries generating complex fluid flow structures. Thereafter, they have stringent requirements on both the discretization (meshing), its versatility and interfaces with 3rd party CAD software and last but not least on the computational efficiency of the fluid solver.

Therefore, the initial research plan includes three main topics: The first one is validating and further developing an existing end-to-end workflow supporting high-order curvilinear meshing for such complex geometries in the Nektar++ framework. The main research goal, however, is creating an automatic mesh refinement process (Adaptive Mesh Refinement / AMR), which minimizes the spatial error of a specific global target parameter of the simulation.

These error metrics are usually the reason for simulating in the first place and some examples in the aerospace context could be lift, drag, etc. These sensitivity metrics are estimated after an initial simulation and they control the resolution through classical mesh adaptation techniques. The project aims at combining all three of them – local change of the solver expansion order (p-type), movement of the nodes in an element around discontinuities (r-type) and whenever this not being sufficient, splitting/merging elements for resolution control (h-type). Once developed, the natural validation of such algorithm is the simulation of compressible flow, where shockwaves are generated, move inside the domain and are reflected from surfaces.

The supersonic flow around an airfoil (part of Testcase 2), such as the NACA-0012 will likely be used for this study. Last but not least, the accuracy and computational efficiency will be tested together with the ability for capturing flow features in scale resolving turbulent simulations. A good validation study could be an incompressible simulation (Large Eddy Simulation) for the industrial geometry part of Testcase 1.

[wpdm_login_form]