Adaptive Mesh Refinement went wrong with OpenFOAM

preview_player
Показать описание
As we all know, it takes time and experiment to get good results from the CFD simulations. In this attempt, 923 mesh ended up 53k mesh with LOTS OF non-orthogonal elements. So, this is a reminder of we must be careful what we wish(inputs) from the OpenFOAM.
  1. Eren Çolak
  2. #CFD #OpenFOAM #dynamicMesh #adaptiveMeshRefinement
Рекомендации по теме
Adaptive Mesh Refinement 0:00:39

Adaptive Mesh Refinement went wrong with OpenFOAM

Kelvin–Helmholtz instability simulation 0:00:19

Kelvin–Helmholtz instability simulation with adaptive mesh refinement

Adaptive Mesh Refinement 0:00:49

Adaptive Mesh Refinement - Kármán Vortex Street ∇ OpenFOAM®

Adaptive Mesh Refinement: 0:46:02

Adaptive Mesh Refinement: Algorithms and Applications

A posteriori error 0:04:30

A posteriori error estimation and adaptive mesh-refinement techniques

Adaptive Mesh Refinement 0:19:09

Adaptive Mesh Refinement (AMR) | OpenFOAM

Adaptive Mesh Refinement 0:00:26

Adaptive Mesh Refinement & Coarsening

Flow around a 0:00:42

Flow around a cylinder - Adaptive Mesh Refinement (AMR) - Fluid Mechanics - OpenFOAM 2112

Simulation of a 0:00:09

Simulation of a falling droplet using adaptive mesh refinement (a more accurate interface capturing)

Direct numerical simulation 0:13:34

Direct numerical simulation of turbulent flows through adaptive mesh refinement

Adaptive Mesh Refinement 0:00:10

Adaptive Mesh Refinement in CAT3410 simulation

Adaptive mesh refinement 0:01:22

Adaptive mesh refinement of a shock tube | Skill-Lync

Adaptive Mesh Refinement 0:10:01

Adaptive Mesh Refinement in SolidWorks Simulation

adaptive mesh refinement 0:00:36

adaptive mesh refinement

Demo-flood- Adaptive mesh 0:00:21

Demo-flood- Adaptive mesh refinement method-2D shallow water equations

Adaptive Mesh Refinement 0:00:04

Adaptive Mesh Refinement simulation of adiabatic hypersonic jet

Adaptive Mesh Refinement 0:00:11

Adaptive Mesh Refinement on a blunt body (STAR-CCM+)

OpenFOAM simulation - 0:00:31

OpenFOAM simulation - Adaptive mesh refinement (AMR)

Adaptive mesh refinement 0:00:16

Adaptive mesh refinement of Rayleigh-Plateau instability

TAC: Optimizing Error-Bounded 0:20:01

TAC: Optimizing Error-Bounded Lossy Compression for 3D Adaptive Mesh Refinement Simulations

Numerical Diffusion with 0:00:06

Numerical Diffusion with Adaptive Mesh Refinement (AMR) Enabled in CONVERGE

Flow past a 0:00:18

Flow past a square cylinder with adaptive mesh refinement

Adaptive mesh refinement 0:00:53

Adaptive mesh refinement

blastFoam CFD Code 0:00:09

blastFoam CFD Code Development Update Including Adaptive Mesh Refinement (AMR)

Комментарии
Автор

I don't understand.
What went wrong?

hopkinsjr
Автор

Does it actually make sense to use AMR for a steady state solver? For velocity gradients for example i want to get only these regions of my mesh where the velocity gradients is greater than some certain level?

volkanatar
Автор

You are always going to get non-orthogonal elements using an oct-tree (or quad-tree) method for refining the mesh. This is because (2d example ahead) the cell next to the refined cells becomes a skewed pentagon (meaning that the line connecting the cell centroids does not pass through a face centroid) and the faces connecting the unrefined and refined elements are not orthogonal (meaning that the line connecting the cell centroids is not orthogonal to the face). Hence, you need to use corrected interpolation schemes and non-orthogonal correctors. Notice that even your original grid had very skewed non-orthogonal faces (in the o-grids). This is not an issue of OpenFOAM, it is simply a characteristic of the finite volume mesh.

However, I would say OpenFOAM is handling the situation quite well here. I do not see anything obviously wrong in the solution from your video and I would not say that the adaptive mesh refinement went wrong. Even if OpenFOAM had a more advanced algorithm for AMR which was somehow capable of preserving orthogonality at the refined-unrefined interface, you would still have bad elements in your o-grid.

Also, notice that ANSYS FLUENT's AMR employs, at its core, the same principle for splitting the elements and therefore suffers of similar issues.

This said, I like your simulations!

fmuni
Автор

What metric were you using for your adaptation rule? Maybe basing the adaptation on pressure differentials might solve this problem.

atatoraman