Mesh effects in high curvature flows

JamesPMCFC · October 19, 2022, 06:48

Dear all,

I'm running a conjugate case on a co-rotating cavity using URANS to capture the unsteady structures that form in the fluid.
However, using OpenFOAM I've observed an unexpected mesh effect that occurs in the velocity that has affected the development of these structures.
These do not appear to form when using other CFD softwares (CFX/Fluent) and I have attempted to use different meshes in OpenFOAM but this effect still occurs.

I've investigated using different gradient schemes and while I get the best structure formation and matching to experimental results using Gauss linear, the mesh effects are noticeably worse.
LeastSquares reduces the mesh effects (not completely removing them) however the structure formation and temperature distributions are further from experimental results.

I've recently implemented a version of the SRFFoam model into chtMultiRegionFoam to help with resolving the flow relative to the high rotation of the cavity (8000rpm), but this mesh effect occurs using both a relative (solving Urel) approach or using a MRFzone (solving U).

I've attached a screenshot of my mesh alongside the mesh effect, as well as my fvSchemes and fvSolution files, happy to provide any other information if I've missed anything!

Any help would be greatly appreciated!
James

8000rpmLinearMeshEffect.jpg
fvSchemes_CHT.txt
fvSolution_CHT.txt

joegi.geo · November 3, 2022, 17:04

Well, my guest is due to the sensitivity of the turbulence model to system rotation.

Try using a non-linear eddy viscosity model or a Reynolds stress model, I am quite sure that the results will improve.

It is likely that in Fluent or CFX the curvature correction option is enabled.

Maybe you an share the case?

October 19, 2022, 06:48	Mesh effects in high curvature flows	#1
JamesPMCFC New Member James Join Date: Jan 2021 Posts: 1 Rep Power: 0	Dear all, I'm running a conjugate case on a co-rotating cavity using URANS to capture the unsteady structures that form in the fluid. However, using OpenFOAM I've observed an unexpected mesh effect that occurs in the velocity that has affected the development of these structures. These do not appear to form when using other CFD softwares (CFX/Fluent) and I have attempted to use different meshes in OpenFOAM but this effect still occurs. I've investigated using different gradient schemes and while I get the best structure formation and matching to experimental results using Gauss linear, the mesh effects are noticeably worse. LeastSquares reduces the mesh effects (not completely removing them) however the structure formation and temperature distributions are further from experimental results. I've recently implemented a version of the SRFFoam model into chtMultiRegionFoam to help with resolving the flow relative to the high rotation of the cavity (8000rpm), but this mesh effect occurs using both a relative (solving Urel) approach or using a MRFzone (solving U). I've attached a screenshot of my mesh alongside the mesh effect, as well as my fvSchemes and fvSolution files, happy to provide any other information if I've missed anything! Any help would be greatly appreciated! James 8000rpmLinearMeshEffect.jpg fvSchemes_CHT.txt fvSolution_CHT.txt

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
sliding mesh problem in CFX	Saima	CFX	46	September 11, 2021 07:38
A high order curvilinear mesh generation(converting) tool is strongly recommended!	hyperman	Mesh Generation & Pre-Processing	0	June 22, 2018 08:04
[snappyHexMesh] SnappyHexMesh for internal Flow	vishwa	OpenFOAM Meshing & Mesh Conversion	24	June 27, 2016 08:54
Icemcfd 11: Loss of mesh from surface mesh option?	Joe	CFX	2	March 26, 2007 18:10
separation region in corner flows submitted to curvature effects	Stephane	Main CFD Forum	2	July 13, 1998 19:06

November 3, 2022, 17:04		#2
joegi.geo Senior Member joegi Join Date: Nov 2009 Location: genoa Posts: 103 Rep Power: 16	Well, my guest is due to the sensitivity of the turbulence model to system rotation. Try using a non-linear eddy viscosity model or a Reynolds stress model, I am quite sure that the results will improve. It is likely that in Fluent or CFX the curvature correction option is enabled. Maybe you an share the case?