advection/transient scheme-Turbulence numerics

PegahMhr · July 21, 2023, 11:38

Hi,

I need some thoughts regarding the simulation of flow over a cylinder. The Reynolds number range is between (3000 to 18000). The Results which I get for Re number over 8000 with an Upwind advection scheme, First order backward Euler transient scheme, and First Order Turbulence numerics (Case 1) show a lot of difference with the high-resolution advection scheme, Second order backward Euler transient scheme, and high-resolution Turbulence numerics (Case 2). Simulations in the first case are closer to the experiment.
I am wondering why? and Is it ok if I implement the first case options for the turbulence problem like this?

Opaque · July 21, 2023, 15:43

Are you comparing different numerical schemes on the same mesh? If so, there is no surprise you notice differences.

Which is better? Difficult to say since I cannot tell which one is closer to mesh-independent convergence.

How do you find out? You refine the mesh and compare solutions with the same numerics as you keep refining your mesh. At some point, those solutions will converge. I would expect the solution using First order numerics (Upwind) will converge slower than the Specified Blend Factor/High-Resolution schemes.

Making decisions on a single mesh solution w/o verification of mesh-independent behavior, i.e. due diligence, is not a correct use of numerical methods.

ghorrocks · July 21, 2023, 23:10

Opaque's comments are essential for the numerical side of things.

And there is a physical side as well. In the range of Reynolds number from 3000 to 18000 flow over a cylinder has a laminar boundary layer which goes turbulent in the wake. This means a standard turbulence model is not suitable as the critical part of the flow (the boundary layer) is laminar.

This model is in a very tricky part of the Reynolds number regime where the flow is both laminar and turbulent. If you are trying to get lift or drag on the cylinder I would recommend you try a laminar simulation as that will model the boundary layer more accurately. If you are trying to model the wake then a turbulence model is appropriate as that part of the flow is turbulent.

If you are trying to model the whole thing accurately you will find this quite challenging. It will require either an LES approach, or maybe a turbulence transition model.

PegahMhr · July 24, 2023, 11:02

Thank you for your help

July 21, 2023, 11:38	advection/transient scheme-Turbulence numerics	#1
PegahMhr New Member Pegah Join Date: Jun 2022 Posts: 3 Rep Power: 3	Hi, I need some thoughts regarding the simulation of flow over a cylinder. The Reynolds number range is between (3000 to 18000). The Results which I get for Re number over 8000 with an Upwind advection scheme, First order backward Euler transient scheme, and First Order Turbulence numerics (Case 1) show a lot of difference with the high-resolution advection scheme, Second order backward Euler transient scheme, and high-resolution Turbulence numerics (Case 2). Simulations in the first case are closer to the experiment. I am wondering why? and Is it ok if I implement the first case options for the turbulence problem like this?

July 21, 2023, 15:43		#2
Opaque Senior Member Join Date: Jun 2009 Posts: 1,804 Rep Power: 32	Are you comparing different numerical schemes on the same mesh? If so, there is no surprise you notice differences. Which is better? Difficult to say since I cannot tell which one is closer to mesh-independent convergence. How do you find out? You refine the mesh and compare solutions with the same numerics as you keep refining your mesh. At some point, those solutions will converge. I would expect the solution using First order numerics (Upwind) will converge slower than the Specified Blend Factor/High-Resolution schemes. Making decisions on a single mesh solution w/o verification of mesh-independent behavior, i.e. due diligence, is not a correct use of numerical methods. __________________ Note: I do not answer CFD questions by PM. CFD questions should be posted on the forum.

July 21, 2023, 23:10		#3
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	Opaque's comments are essential for the numerical side of things. And there is a physical side as well. In the range of Reynolds number from 3000 to 18000 flow over a cylinder has a laminar boundary layer which goes turbulent in the wake. This means a standard turbulence model is not suitable as the critical part of the flow (the boundary layer) is laminar. This model is in a very tricky part of the Reynolds number regime where the flow is both laminar and turbulent. If you are trying to get lift or drag on the cylinder I would recommend you try a laminar simulation as that will model the boundary layer more accurately. If you are trying to model the wake then a turbulence model is appropriate as that part of the flow is turbulent. If you are trying to model the whole thing accurately you will find this quite challenging. It will require either an LES approach, or maybe a turbulence transition model. Opaque likes this. __________________ Note: I do not answer CFD questions by PM. CFD questions should be posted on the forum.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Turbulence postprocessing	Mohsin	FLUENT	2	October 3, 2016 14:18
Troubles Developing Turbulence	fatirishman53	OpenFOAM Running, Solving & CFD	1	September 20, 2015 11:44
specify the turbulence intensity and the turbulence length	bennoman75	Main CFD Forum	0	May 15, 2013 05:36
K - epsilon VS SST turbulence model	Maicol	Main CFD Forum	0	November 30, 2012 16:25
Vortex shedding, FSI-analysis, turbulence numerics	tallknuseren	CFX	3	May 10, 2010 04:31

July 24, 2023, 11:02		#4
PegahMhr New Member Pegah Join Date: Jun 2022 Posts: 3 Rep Power: 3	Thank you for your help