LESDES deltas

cfdmarkus · December 8, 2008, 05:11

Hi all

I had a look at the DES model currently implemented in OF1.5. For computing the DES length scale one needs a delta defined as delta()=C*max(dx,dy,dz). This definition is a crucial point in DES and should be used.

I could only find definitions of delta() which are based on the cuberoot delta()=C*(dx*dy*dz)^1/3.

Is the definition of delta based on the maximum grid spacing readily available in OF, or do I have to implement it?

Best regrads,
Markus

lakeat · January 11, 2009, 07:10

Hi, Markus, how did you do that now? I am interested!

Daniel

braennstroem · January 12, 2009, 01:56

Hi,

did you take a look at: src/turbulenceModels/LES/LESdeltas/PrandtlDelta/PrandtlDelta.C
or
src/turbulenceModels/LES/LESdeltas/cubeRootVolDelta/cubeRootVolDelta.C

I think this could be a good start to implement this!?

Fabian

lakeat · January 12, 2009, 03:06

Well, I am not poor in using C++, so I need help.

PrandtlDelta.C and cubeRootVolDelta.C does not give me much infomation, since it seems the length of each cell is needed, But in fvMesh.H, only mesh().V() etc. is provided, so I'm not sure how to get mesh().maxLength() in a easiest way, could someone give me a hint? Many thanks.

Daniel

lakeat · January 12, 2009, 04:11

Sorry, I mean I'm good at C++ programing.

lakeat · January 12, 2009, 04:13

Sorry, I mean I'm NOT good at C++ programing.
I hate my keyBoard

cfdmarkus · January 12, 2009, 04:41

Hi Daniel

Sorry I haven't had enough time to work on this issue.
But I still believe that a delta based on the maximum grid spacing is crucial for the old DES versions. Therefore, sooner or later I will have to implement it as well.

I am also wondering if there is a command to return the maximum grid spacing of a cell !?

Markus

eugene · January 12, 2009, 09:03

Considering that FOAM cells are represented as arbitrary polyhedra, what would the definition of "grid spacing" be?

lakeat · January 12, 2009, 10:09

Yes, but isn't that most of the DES Simulations now are carried on structured hexahedra?

eugene · January 12, 2009, 10:20

The cells might look like hexahedra, but the internal representation is 100% face based. FOAM meshes are completely unstructured. The best you could do would be to ask for a maximum edge length, which could be misleading in some situations, but would do what you need provided all cells were hex shaped.

I do not know of any unambiguous definitions that will comply to the Spalart method for getting max(dx,dy,dx) on arbitrary meshes.

cfdmarkus · January 12, 2009, 10:30

Ok I get your point.

I just don't believe that the cubeRootVolDelta is an appropriate length scale for high aspect ratio hexahedral cells.
Also, the only way to ensure the correct behaviour of DES for attached flows is the design of "DES-grids" (with high aspect ratio cells inside the boundary layer) combined with the "MaxDelta" definition.

Markus

lakeat · January 12, 2009, 10:30

Fully Agree!

And have you tried DDES, can you give some comments? Thanks

lakeat · January 12, 2009, 10:42

To Markus,

For me, maxDelta is just a method, and an easy way to control the appropriate length scale, to control the proper RANS region.
1. "Easy" but not so easy in preprocessing, that is "DES-grids" design process.
2. "Easy" but not best, why do you think cubeRootVolDelta will violate the "DES-law"?

eugene · January 12, 2009, 11:19

We will be trying DDES this year, so I have no personal experience of it yet. Initial indications are that it is a definite improvement.

Assuming you have a hex dominant mesh with surface layers, then the only place that cuberoot vol will be different from max xyz, will be in the transition zone between wall distance and cell size length scales. Near the wall the wall distance remains unchanged and far from the wall, the cuberoot is equal to the max length. In any case, it doesn't seem to make a massive difference to the results. I did a study in my thesis (which is available for download from Hrv's foamcfd website), so you can follow my logic and make up your own mind.

cfdmarkus · January 12, 2009, 11:20

I am not sure what you mean by "DES-law".

Using the cubeRootVolDelta makes it much more difficult to maintain RANS in attached flows and consequently, the switch from RANS to LES can occur well inside the boundary layer. A switch inside the boundary layer usually has negative consequences (drop in skin friction). A length scale defined by delta=max(dx,dy,dx) combined with anisotropic cells will help to prevent a switch from RANS to LES inside the boundary layer. Thats why I want to use delta=max(dx,dy,dx).

Markus

cfdmarkus · January 12, 2009, 11:37

I came across an upgraded version of DDES called IDDES. If you are interested:

A hybrid RANS-LES approach with delayed-DES and wall-modelled LES capabilities
International Journal of Heat and Fluid Flow, Volume 29, Issue 6, December 2008, Pages 1638-1649
Mikhail L. Shur, Philippe R. Spalart, Mikhail Kh. Strelets, Andrey K. Travin

Markus

lakeat · January 12, 2009, 12:44

I know this article. But according to Wilcox，i won't use it for its GREAT complexity, for i think its improvement is limited both in engineering-requirement results and basic theory.
It's too complex to have a model like that. :-) just my thought.

Good night!

lakeat · January 12, 2009, 23:42

Hi, Markus!

Do you have the original paper of DES by Spalart. May I have a copy of it?

Thank you.
My Email is lakeat@gmail.com

Daniel

---------------------------
Spalart, P.R., Jou, W.-H., Strelets, M., Allmaras, S.R.: Comments on the feasibility of LES for wings, and on a hybrid
RANS/LES approach. In: Proceedings of first AFOSR international conference on DNS/LES, Ruston, Louisiana. Greyden
Press, 4–8 Aug (1997)

braennstroem · January 13, 2009, 04:36

Hi,

could it possible to use the gradient of a virtual field, which then could be used to determine the max value!?
For attached flows it might be better to use a different approach with a hard, kind of 'predefined', interface position. As the name suggests, DES is more suitable for detached flows. There is a review about hybrid LES/RANS methods of Froehlich and von Terzi, which is quite interesting.

By the way, it would be quite nice, if you guys support the Turbulence Special Interest Group!?

Fabian

cfdmarkus · January 13, 2009, 08:04

Daniel:
yes I should have a hardcopy of the original DES97 paper. If I find a scanner I can send it to you (might take some time).

Fabian:
I am really not sure about your first point. However, constructing a MaxDelta for hexahedral cells shouldn't be too difficult.
I agree that the definition of a proper interface location is clearly superior to any global-hybrid method. However, these zonal-hybrid methods bring up a lot of other problems and therefore, global-hybrid methods are at least a good starting point.

Do you have any experience in running hybrid simulations in OF wich use a predefined interface location? I am very interested to hear about your experiences regarding this issue!

Markus

December 8, 2008, 05:11	Hi all I had a look at the	#1
cfdmarkus Member Markus Weinmann Join Date: Mar 2009 Location: Stuttgart, Germany Posts: 77 Rep Power: 17	Hi all I had a look at the DES model currently implemented in OF1.5. For computing the DES length scale one needs a delta defined as delta()=Cmax(dx,dy,dz). This definition is a crucial point in DES and should be used. I could only find definitions of delta() which are based on the cuberoot delta()=C(dxdydz)^1/3. Is the definition of delta based on the maximum grid spacing readily available in OF, or do I have to implement it? Best regrads, Markus

January 11, 2009, 07:10	Hi, Markus, how did you do tha	#2
lakeat Senior Member Daniel WEI (老魏) Join Date: Mar 2009 Location: Beijing, China Posts: 689 Blog Entries: 9 Rep Power: 21	Hi, Markus, how did you do that now? I am interested! Daniel __________________ ~ Daniel WEI ------------- Boeing Research & Technology - China Beijing, China Email

January 12, 2009, 01:56	Hi, did you take a look at:	#3
braennstroem Senior Member Fabian Braennstroem Join Date: Mar 2009 Posts: 407 Rep Power: 19	Hi, did you take a look at: src/turbulenceModels/LES/LESdeltas/PrandtlDelta/PrandtlDelta.C or src/turbulenceModels/LES/LESdeltas/cubeRootVolDelta/cubeRootVolDelta.C I think this could be a good start to implement this!? Fabian

January 12, 2009, 03:06	Well, I am not poor in using C	#4
lakeat Senior Member Daniel WEI (老魏) Join Date: Mar 2009 Location: Beijing, China Posts: 689 Blog Entries: 9 Rep Power: 21	Well, I am not poor in using C++, so I need help. PrandtlDelta.C and cubeRootVolDelta.C does not give me much infomation, since it seems the length of each cell is needed, But in fvMesh.H, only mesh().V() etc. is provided, so I'm not sure how to get mesh().maxLength() in a easiest way, could someone give me a hint? Many thanks. Daniel __________________ ~ Daniel WEI ------------- Boeing Research & Technology - China Beijing, China Email

January 12, 2009, 04:11	Sorry, I mean I'm good at C++	#5
lakeat Senior Member Daniel WEI (老魏) Join Date: Mar 2009 Location: Beijing, China Posts: 689 Blog Entries: 9 Rep Power: 21	Sorry, I mean I'm good at C++ programing. __________________ ~ Daniel WEI ------------- Boeing Research & Technology - China Beijing, China Email

January 12, 2009, 04:13	Sorry, I mean I'm NOT good at	#6
lakeat Senior Member Daniel WEI (老魏) Join Date: Mar 2009 Location: Beijing, China Posts: 689 Blog Entries: 9 Rep Power: 21	Sorry, I mean I'm NOT good at C++ programing. I hate my keyBoard __________________ ~ Daniel WEI ------------- Boeing Research & Technology - China Beijing, China Email

January 12, 2009, 04:41	Hi Daniel Sorry I haven't h	#7
cfdmarkus Member Markus Weinmann Join Date: Mar 2009 Location: Stuttgart, Germany Posts: 77 Rep Power: 17	Hi Daniel Sorry I haven't had enough time to work on this issue. But I still believe that a delta based on the maximum grid spacing is crucial for the old DES versions. Therefore, sooner or later I will have to implement it as well. I am also wondering if there is a command to return the maximum grid spacing of a cell !? Markus

January 12, 2009, 09:03	Considering that FOAM cells ar	#8
eugene Senior Member Eugene de Villiers Join Date: Mar 2009 Posts: 725 Rep Power: 21	Considering that FOAM cells are represented as arbitrary polyhedra, what would the definition of "grid spacing" be?

January 12, 2009, 10:09	Yes, but isn't that most of th	#9
lakeat Senior Member Daniel WEI (老魏) Join Date: Mar 2009 Location: Beijing, China Posts: 689 Blog Entries: 9 Rep Power: 21	Yes, but isn't that most of the DES Simulations now are carried on structured hexahedra? __________________ ~ Daniel WEI ------------- Boeing Research & Technology - China Beijing, China Email

January 12, 2009, 10:20	The cells might look like hexa	#10
eugene Senior Member Eugene de Villiers Join Date: Mar 2009 Posts: 725 Rep Power: 21	The cells might look like hexahedra, but the internal representation is 100% face based. FOAM meshes are completely unstructured. The best you could do would be to ask for a maximum edge length, which could be misleading in some situations, but would do what you need provided all cells were hex shaped. I do not know of any unambiguous definitions that will comply to the Spalart method for getting max(dx,dy,dx) on arbitrary meshes.

January 12, 2009, 10:30	Ok I get your point. I jus	#11
cfdmarkus Member Markus Weinmann Join Date: Mar 2009 Location: Stuttgart, Germany Posts: 77 Rep Power: 17	Ok I get your point. I just don't believe that the cubeRootVolDelta is an appropriate length scale for high aspect ratio hexahedral cells. Also, the only way to ensure the correct behaviour of DES for attached flows is the design of "DES-grids" (with high aspect ratio cells inside the boundary layer) combined with the "MaxDelta" definition. Markus

January 12, 2009, 10:30	Fully Agree! And have you t	#12
lakeat Senior Member Daniel WEI (老魏) Join Date: Mar 2009 Location: Beijing, China Posts: 689 Blog Entries: 9 Rep Power: 21	Fully Agree! And have you tried DDES, can you give some comments? Thanks __________________ ~ Daniel WEI ------------- Boeing Research & Technology - China Beijing, China Email

January 12, 2009, 10:42	To Markus, For me, maxDelt	#13
lakeat Senior Member Daniel WEI (老魏) Join Date: Mar 2009 Location: Beijing, China Posts: 689 Blog Entries: 9 Rep Power: 21	To Markus, For me, maxDelta is just a method, and an easy way to control the appropriate length scale, to control the proper RANS region. 1. "Easy" but not so easy in preprocessing, that is "DES-grids" design process. 2. "Easy" but not best, why do you think cubeRootVolDelta will violate the "DES-law"? __________________ ~ Daniel WEI ------------- Boeing Research & Technology - China Beijing, China Email

January 12, 2009, 11:19	We will be trying DDES this ye	#14
eugene Senior Member Eugene de Villiers Join Date: Mar 2009 Posts: 725 Rep Power: 21	We will be trying DDES this year, so I have no personal experience of it yet. Initial indications are that it is a definite improvement. Assuming you have a hex dominant mesh with surface layers, then the only place that cuberoot vol will be different from max xyz, will be in the transition zone between wall distance and cell size length scales. Near the wall the wall distance remains unchanged and far from the wall, the cuberoot is equal to the max length. In any case, it doesn't seem to make a massive difference to the results. I did a study in my thesis (which is available for download from Hrv's foamcfd website), so you can follow my logic and make up your own mind.

January 12, 2009, 11:20	I am not sure what you mean by	#15
cfdmarkus Member Markus Weinmann Join Date: Mar 2009 Location: Stuttgart, Germany Posts: 77 Rep Power: 17	I am not sure what you mean by "DES-law". Using the cubeRootVolDelta makes it much more difficult to maintain RANS in attached flows and consequently, the switch from RANS to LES can occur well inside the boundary layer. A switch inside the boundary layer usually has negative consequences (drop in skin friction). A length scale defined by delta=max(dx,dy,dx) combined with anisotropic cells will help to prevent a switch from RANS to LES inside the boundary layer. Thats why I want to use delta=max(dx,dy,dx). Markus

January 12, 2009, 11:37	I came across an upgraded vers	#16
cfdmarkus Member Markus Weinmann Join Date: Mar 2009 Location: Stuttgart, Germany Posts: 77 Rep Power: 17	I came across an upgraded version of DDES called IDDES. If you are interested: A hybrid RANS-LES approach with delayed-DES and wall-modelled LES capabilities International Journal of Heat and Fluid Flow, Volume 29, Issue 6, December 2008, Pages 1638-1649 Mikhail L. Shur, Philippe R. Spalart, Mikhail Kh. Strelets, Andrey K. Travin Markus

January 12, 2009, 12:44	I know this article. But accor	#17
lakeat Senior Member Daniel WEI (老魏) Join Date: Mar 2009 Location: Beijing, China Posts: 689 Blog Entries: 9 Rep Power: 21	I know this article. But according to Wilcox，i won't use it for its GREAT complexity, for i think its improvement is limited both in engineering-requirement results and basic theory. It's too complex to have a model like that. :-) just my thought. Good night! __________________ ~ Daniel WEI ------------- Boeing Research & Technology - China Beijing, China Email

January 12, 2009, 23:42	Hi, Markus! Do you have the	#18
lakeat Senior Member Daniel WEI (老魏) Join Date: Mar 2009 Location: Beijing, China Posts: 689 Blog Entries: 9 Rep Power: 21	Hi, Markus! Do you have the original paper of DES by Spalart. May I have a copy of it? Thank you. My Email is lakeat@gmail.com Daniel --------------------------- Spalart, P.R., Jou, W.-H., Strelets, M., Allmaras, S.R.: Comments on the feasibility of LES for wings, and on a hybrid RANS/LES approach. In: Proceedings of first AFOSR international conference on DNS/LES, Ruston, Louisiana. Greyden Press, 4–8 Aug (1997) __________________ ~ Daniel WEI ------------- Boeing Research & Technology - China Beijing, China Email

January 13, 2009, 04:36	Hi, could it possible to us	#19
braennstroem Senior Member Fabian Braennstroem Join Date: Mar 2009 Posts: 407 Rep Power: 19	Hi, could it possible to use the gradient of a virtual field, which then could be used to determine the max value!? For attached flows it might be better to use a different approach with a hard, kind of 'predefined', interface position. As the name suggests, DES is more suitable for detached flows. There is a review about hybrid LES/RANS methods of Froehlich and von Terzi, which is quite interesting. By the way, it would be quite nice, if you guys support the Turbulence Special Interest Group!? Fabian

January 13, 2009, 08:04	Daniel: yes I should have a h	#20
cfdmarkus Member Markus Weinmann Join Date: Mar 2009 Location: Stuttgart, Germany Posts: 77 Rep Power: 17	Daniel: yes I should have a hardcopy of the original DES97 paper. If I find a scanner I can send it to you (might take some time). Fabian: I am really not sure about your first point. However, constructing a MaxDelta for hexahedral cells shouldn't be too difficult. I agree that the definition of a proper interface location is clearly superior to any global-hybrid method. However, these zonal-hybrid methods bring up a lot of other problems and therefore, global-hybrid methods are at least a good starting point. Do you have any experience in running hybrid simulations in OF wich use a predefined interface location? I am very interested to hear about your experiences regarding this issue! Markus