Skip to main content

Adaptive mesh refinement for turbulent flows: methodology and applications

Time: Thu 2022-03-03 10.30 - 11.30

Location: Faxén, Teknikringen 8

Video link: Hybrid e-Seminar (Zoom)

Participating: Daniele Massaro (KTH)

Export to calendar

Abstract. The main goal of adaptive mesh refinement (AMR) is to increase the accuracy
of the solution at a lower computational cost than what was achievable
with a conformal mesh. The two main ingredients of AMR are the mesh refinement
strategy and the error estimation. This latter is discussed here by
comparing the spectral error indicator (SEI), a local measure of the quadrature
and truncation errors, and the adjoint error estimator (AEE), a goal-oriented
and dual-weighted residuals estimator. We consider the fully turbulent flow over
a three-dimensional periodic hill to see the different mesh resolutions and flow
solutions.


When it comes to high-order methods such as the spectral element method
adapted in the CFD code Nek5000, we also need to take care of the discontinuities
arising at non-conformal interfaces (wiggles). Considering the turbulent
flow in a straight pipe, we use the turbulent kinetic energy terms as wiggles
indicator to figure out the hanging node effect on the solution. We can claim
that the jumps in derivatives are uniquely related to an inadequately resolved
mesh, guaranteeing the robustness of our code and allowing significant savings
for internal wall-bounded flows as well.


Eventually, the latest AMR application in a direct numerical simulation
(DNS) is presented: the flow around a stepped cylinder. This moderately complex
geometry consists of two cylinders with different diameters joint at one
extremity and it represents a good model for real-world applications, e.g. the
foundations of offshore wind turbines. We exploit AMR capabilities to investigate
the wake vortex dynamics and junction region behaviour as a function of
Reynolds number (ReD = 150, 1000, 5000). Moreover, we study a reduced-order
model, as proper orthogonal decomposition (POD) or dynamic mode decomposition
(DMD), which provides a better understanding of the flow dynamics.

Page responsible:Ardeshir Hanifi
Belongs to: FLOW
Last changed: Mar 02, 2022