Trailing edge separation pocket: physical or numerical?

MarcoCaboni · February 27, 2021, 17:28

Hi all,

Recently, I have been performing steady state, incompressible simulations using the k-omega SST RANS turbulence model on a wind turbine blade. For this blade, I have generated 2 grids which use the same number of chordwise points, but one of them uses 3-time larger number of spanwise points (see screenshot). The simulation done using the finer mesh exhibits a large trailing edge separation pocket (see screenshot) in a region where we wouldn’t normally expect separation to occur (very low AoA and no sudden variations of the blade’s planform). I have also noticed several much smaller pockets occurring in the finer mesh over the whole training edge of the blade. Using the coarser mesh most of the separation pockets (including the large one) disappear.

For the finer mesh, average residuals look fine, and field residuals do not indicate any grid-related issues around the pocket region. Also, the flow field around the pocket does not look abnormal. This is puzzling as I cannot conclude whether these pockets are physical (captured by a finer mesh) or just numerical artifacts.

Have you ever experienced anything similar? Or can you suggest possible tests or references to shed some light on the issue?

Many thanks!

LoUcAsss · March 1, 2021, 04:43

Hi,

Coarser mesh leads to higher numerical dissipation. So it is not surprising to observe separation as the mesh is finer.
However in your case, I would have expected to observe that while increasing the number of nodes chordwise. It is true that the separation is 3D and the span direction should have an impact on the separation.
I suggest you study the mesh convergence in all directions. And I think you can find many articles with similar issues on naca profiles.

Cheers,
Lucas

FMDenaro · March 1, 2021, 05:35

First of all, you should ensure that the residuals of all equations on the finer grid are below a good threshold. Then try to refine further the grid in all directions. Note that to assess the physical meaning of a flow separation you need to have a fine resolution in the BL.

February 27, 2021, 17:28	Trailing edge separation pocket: physical or numerical?	#1
MarcoCaboni New Member Marco Caboni Join Date: Jan 2019 Posts: 1 Rep Power: 0	Hi all, Recently, I have been performing steady state, incompressible simulations using the k-omega SST RANS turbulence model on a wind turbine blade. For this blade, I have generated 2 grids which use the same number of chordwise points, but one of them uses 3-time larger number of spanwise points (see screenshot). The simulation done using the finer mesh exhibits a large trailing edge separation pocket (see screenshot) in a region where we wouldn’t normally expect separation to occur (very low AoA and no sudden variations of the blade’s planform). I have also noticed several much smaller pockets occurring in the finer mesh over the whole training edge of the blade. Using the coarser mesh most of the separation pockets (including the large one) disappear. For the finer mesh, average residuals look fine, and field residuals do not indicate any grid-related issues around the pocket region. Also, the flow field around the pocket does not look abnormal. This is puzzling as I cannot conclude whether these pockets are physical (captured by a finer mesh) or just numerical artifacts. Have you ever experienced anything similar? Or can you suggest possible tests or references to shed some light on the issue? Many thanks! Last edited by MarcoCaboni; March 11, 2021 at 14:59.

March 1, 2021, 04:43	Refinement chordwise	#2
LoUcAsss New Member lucas s Join Date: Jul 2013 Location: Grenoble, France Posts: 12 Rep Power: 13	Hi, Coarser mesh leads to higher numerical dissipation. So it is not surprising to observe separation as the mesh is finer. However in your case, I would have expected to observe that while increasing the number of nodes chordwise. It is true that the separation is 3D and the span direction should have an impact on the separation. I suggest you study the mesh convergence in all directions. And I think you can find many articles with similar issues on naca profiles. Cheers, Lucas

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Discontinuity at the blunt trailing edge of an airfoil	klkl	OpenFOAM Running, Solving & CFD	0	October 28, 2020 05:42
Trailing edge sensitivity anlaysis	JanH	SU2	2	November 23, 2018 14:53
Difference between diffusion, dissipation and dispersion in physical and numerical	TurbJet	Main CFD Forum	9	October 6, 2018 19:15
Meshing near leading and trailing edge of propeller	surajp92	STAR-CCM+	1	June 1, 2017 02:40
[ICEM] Meshing Wing with low Y+ value and sharp trailing edge	sonic109	ANSYS Meshing & Geometry	0	July 27, 2015 10:29

March 1, 2021, 05:35		#3
FMDenaro Senior Member Filippo Maria Denaro Join Date: Jul 2010 Posts: 6,897 Rep Power: 73	First of all, you should ensure that the residuals of all equations on the finer grid are below a good threshold. Then try to refine further the grid in all directions. Note that to assess the physical meaning of a flow separation you need to have a fine resolution in the BL.