IISc Bangalore

Adjoint-based unstructured anisotropic mesh adaptation for compressible flow simulations

Department Seminar

Title: Adjoint-based unstructured anisotropic mesh adaptation for compressible flow simulations

Speaker: Dr Aravind Balan
Postdoctoral Fellow, NASA Langley Research Center, USA

Date & Time: 4th September 2019 at 1600 hrs

Venue: Aerospace Engineering Auditorium

Abstract :

Despite the significant advancement in computational power in the recent decades, it is still too computationally expensive to perform high-fidelity fluid flow simulations around complex geometries.
One of the effective ways to decrease the cost, without compromising the accuracy, is to perform simulations on adapted meshes. Mesh adaptation, however, is yet to be proven as a robust technology to be used in industrial settings. The CFD Vision 2030 Study, conducted by NASA, has the goal of formulating a research strategy for developing a visionary CFD capability in the year 2030. The study identifies that mesh generation and adaptivity will continue to be the significant bottlenecks in the CFD work flow. To this end, the present work aims at incorporating adjoint-based anisotropic mesh adaptation capabilities into the stabilized finite-element branch of the legacy FUN3D flow solver, developed at NASA Langley Research Center. Mesh adaptation using adjoints can reduce the error in scalar output functionals, such as lift or drag coefficient in external aerodynamics, in a more efficient manner compared to that of feature-based or of uniform mesh refinement. For the mesh adaptation, a metric-based approach is used, where metric tensors encode information about simplicial mesh elements that can be passed to a metric-conforming mesh generator to produce the required anisotropic mesh. We follow a method where both the adjoint and the primal solutions are used to construct the metric field. Results for inviscid and turbulent flow simulations on adaptive meshes will be compared with results obtained using a solution-based approach that controls the Lp norm of Mach number interpolation error to evaluate the effectiveness and efficiency of the adaptive mesh technology being developed.

Date(s) - 04/09/2019
4:00 pm

Map Unavailable