Dimensionless grid spacing in wall units, deltax-plus and deltaz-plus

balrog_f · May 22, 2017, 10:00

Is there a possibility to plot or calculate the x-plus and z-plus values like you can plot the y-plus value in CFX?

ghorrocks · May 22, 2017, 20:02

What is x+ and z+? I hope you are not confused in thinking that y+ is in the y direction. y+ gives you the normalised wall distance perpendicular to the wall, so is applicable for any orientation of wall.

balrog_f · May 23, 2017, 00:35

The Ansys CFX guide for LES investigates a turbulent channel flow and mentions x+ and z+:
https://www.sharcnet.ca/Software/Ans...l#bp_turbu_s_3

It is defined as:
- x+ is the non-dimensional grid spacing in the streamwise direction
- z+is the non-dimensional grid spacing in the spanwise direction

Also Chapman mentions these values in his work "Grid point requirements for LES" which is a good reference to get a proper mesh for my wall resolving LES, in my opinion. x+ ~ 100 & z+ ~ 20

"I hope you are not confused in thinking that y+ is in the y direction": I really thought that because it has the y distance of the first node from the wall in its definition? But I think I know what you mean: You can just use it for every wall and when the wall is for example perpendicular to the x-axis you get x+ with the integrated y+ function in CFX right?

But does this also works if I am using translational periodic boundaries in spanwise an streamwise directions because there is no real wall defined?

ghorrocks · May 23, 2017, 00:56

y+ is used to refer to the normalised distance from the wall in RANS simulations. The Dx+ and Dz+ you link to are the normalised grid spacing in the x and z directions for a LES simulation. So keep in mind that they might look similar but they are totally different variables. They are not the 3 components of a vector, for instance.

The Dx+ and Dz+ variables you mention are defined in CFX. If you want these variables then you will need to write CEL expressions to get them. But if you read the link you sent you will find that they are not intended to be used for post-processing - they are guides to assist you in defining an appropriate grid for an LES simulation based on the geometry and flow.

balrog_f · May 23, 2017, 01:34

But the definitions of the three values look quite similar. So when should I calculate it or where are they defined in CFX? With which expression do I have access to them. What is the CEL expression for them? Can't find anything in the documentation.
I thought I have to calculate them manually with the respective formulas after I did my RANS Simulation and then adapting my mesh like I do with the y+ value. But you meant it is not a post processing thing. What is it then?
What do you recommend on how to do/calculate it and especially WHEN? It must be somewhere between finishing RANS solution and starting LES

ghorrocks · May 23, 2017, 09:14

y+ is a widely used parameter to give a normalised distance from the wall. It is used in all turbulence models (RANS, LES etc).

Dx+ and Dz+ are parameters only used in LES modelling, and are just used to estimate the mesh size required to model the flow. I guess you can also use it to check your mesh is OK on a LES model, but then you have to calculate the turbulent viscosity which is not straight forward.

LES is not a topic for those inexperienced at CFD. Have you read the rest of the link you provided? It gives a very thorough explanation of why LES is very challenging and not for the feint hearted. Are you aware of the issues it discusses? You don't simply click the LES option and a useful LES takes place.

balrog_f · May 23, 2017, 13:01

Yes I have read the rest of the link as well as some papers and books about LES but I agree with you that LES is a huge topic and I still don't get everything what can go wrong or what is bad for your solution because that needs a lot of experience. The good thing about CFX and LES is that CFX really takes you by the hand when it comes to LES, that's my opinion. Back to topic: Why do you think I have to calculate the turbulent viscosity? Am I overseeing something. The formulas are for channel flow:

x+ = ((u*) *delta_x)/viscosity
z+ = ((u*) *delta_z)/viscosity

with u* = sqrt(wallshearstress/density)
delta_x = boundarylayerthickness/N_x
delta_z =boundarylayerthickness/N_z
N is number of nodes in respective direction
boundarylayerthickness is estimated with half channel height

Please correct me if I am misunderstanding something essentially here

ghorrocks · May 23, 2017, 19:54

The link you posted does not define x+ or z+. If defines Dx+ and Dz+ and they have a different definition to what you just posted.

Are you talking about Dx+ and Dz+ as defined in the link you sent or x+ and z+ as you define in post #7?

balrog_f · May 24, 2017, 08:21

That's where my mistake is maybe. I really thought it is the same. Somebody wrote it in the forum, will post the link later. Can you explain the difference to me please?

Sent from my iPhone using CFD Online Forum mobile app

ghorrocks · May 24, 2017, 19:28

I think I have explained what they appear to be in each post I have put on this thread.

rob3rt 0ng · August 13, 2017, 23:52

Quote:

Originally Posted by balrog_f

Yes I have read the rest of the link as well as some papers and books about LES but I agree with you that LES is a huge topic and I still don't get everything what can go wrong or what is bad for your solution because that needs a lot of experience. The good thing about CFX and LES is that CFX really takes you by the hand when it comes to LES, that's my opinion. Back to topic: Why do you think I have to calculate the turbulent viscosity? Am I overseeing something. The formulas are for channel flow:

x+ = ((u*) *delta_x)/viscosity
z+ = ((u*) *delta_z)/viscosity

with u* = sqrt(wallshearstress/density)
delta_x = boundarylayerthickness/N_x
delta_z =boundarylayerthickness/N_z
N is number of nodes in respective direction
boundarylayerthickness is estimated with half channel height

Please correct me if I am misunderstanding something essentially here

Hi, it looks correct to me. But, it should be called Dx+ and Dz+ instead of x+ and z+.

Regards
Robert

Gang Wang · March 10, 2020, 11:05

Quote:

Originally Posted by rob3rt 0ng

Hi, it looks correct to me. But, it should be called Dx+ and Dz+ instead of x+ and z+.

Regards
Robert

Hi!

I agree with ur opinion , it's the definition of Dx+ and Dz, but I still have one question about u*(ustar), the wallshearstress is a vector, which I could get from the functionObject from OpenFOAM utilities, right?

Best,
Gang Wang

May 22, 2017, 10:00	Dimensionless grid spacing in wall units, deltax-plus and deltaz-plus	#1
balrog_f Member Join Date: Apr 2017 Posts: 53 Rep Power: 9	Is there a possibility to plot or calculate the x-plus and z-plus values like you can plot the y-plus value in CFX?

May 23, 2017, 00:35		#3
balrog_f Member Join Date: Apr 2017 Posts: 53 Rep Power: 9	The Ansys CFX guide for LES investigates a turbulent channel flow and mentions x+ and z+: https://www.sharcnet.ca/Software/Ans...l#bp_turbu_s_3 It is defined as: - x+ is the non-dimensional grid spacing in the streamwise direction - z+is the non-dimensional grid spacing in the spanwise direction Also Chapman mentions these values in his work "Grid point requirements for LES" which is a good reference to get a proper mesh for my wall resolving LES, in my opinion. x+ ~ 100 & z+ ~ 20 "I hope you are not confused in thinking that y+ is in the y direction": I really thought that because it has the y distance of the first node from the wall in its definition? But I think I know what you mean: You can just use it for every wall and when the wall is for example perpendicular to the x-axis you get x+ with the integrated y+ function in CFX right? But does this also works if I am using translational periodic boundaries in spanwise an streamwise directions because there is no real wall defined? Last edited by balrog_f; May 23, 2017 at 00:37. Reason: wrong value fo z+

May 23, 2017, 09:14		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,869 Rep Power: 144	y+ is a widely used parameter to give a normalised distance from the wall. It is used in all turbulence models (RANS, LES etc). Dx+ and Dz+ are parameters only used in LES modelling, and are just used to estimate the mesh size required to model the flow. I guess you can also use it to check your mesh is OK on a LES model, but then you have to calculate the turbulent viscosity which is not straight forward. LES is not a topic for those inexperienced at CFD. Have you read the rest of the link you provided? It gives a very thorough explanation of why LES is very challenging and not for the feint hearted. Are you aware of the issues it discusses? You don't simply click the LES option and a useful LES takes place. EddySouth likes this.

May 22, 2017, 20:02		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,869 Rep Power: 144	What is x+ and z+? I hope you are not confused in thinking that y+ is in the y direction. y+ gives you the normalised wall distance perpendicular to the wall, so is applicable for any orientation of wall.

May 23, 2017, 00:56		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,869 Rep Power: 144	y+ is used to refer to the normalised distance from the wall in RANS simulations. The Dx+ and Dz+ you link to are the normalised grid spacing in the x and z directions for a LES simulation. So keep in mind that they might look similar but they are totally different variables. They are not the 3 components of a vector, for instance. The Dx+ and Dz+ variables you mention are defined in CFX. If you want these variables then you will need to write CEL expressions to get them. But if you read the link you sent you will find that they are not intended to be used for post-processing - they are guides to assist you in defining an appropriate grid for an LES simulation based on the geometry and flow.

May 23, 2017, 01:34		#5
balrog_f Member Join Date: Apr 2017 Posts: 53 Rep Power: 9	But the definitions of the three values look quite similar. So when should I calculate it or where are they defined in CFX? With which expression do I have access to them. What is the CEL expression for them? Can't find anything in the documentation. I thought I have to calculate them manually with the respective formulas after I did my RANS Simulation and then adapting my mesh like I do with the y+ value. But you meant it is not a post processing thing. What is it then? What do you recommend on how to do/calculate it and especially WHEN? It must be somewhere between finishing RANS solution and starting LES

May 23, 2017, 13:01		#7
balrog_f Member Join Date: Apr 2017 Posts: 53 Rep Power: 9	Yes I have read the rest of the link as well as some papers and books about LES but I agree with you that LES is a huge topic and I still don't get everything what can go wrong or what is bad for your solution because that needs a lot of experience. The good thing about CFX and LES is that CFX really takes you by the hand when it comes to LES, that's my opinion. Back to topic: Why do you think I have to calculate the turbulent viscosity? Am I overseeing something. The formulas are for channel flow: x+ = ((u) delta_x)/viscosity z+ = ((u) delta_z)/viscosity with u* = sqrt(wallshearstress/density) delta_x = boundarylayerthickness/N_x delta_z =boundarylayerthickness/N_z N is number of nodes in respective direction boundarylayerthickness is estimated with half channel height Please correct me if I am misunderstanding something essentially here

May 23, 2017, 19:54		#8
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,869 Rep Power: 144	The link you posted does not define x+ or z+. If defines Dx+ and Dz+ and they have a different definition to what you just posted. Are you talking about Dx+ and Dz+ as defined in the link you sent or x+ and z+ as you define in post #7?

May 24, 2017, 08:21		#9
balrog_f Member Join Date: Apr 2017 Posts: 53 Rep Power: 9	That's where my mistake is maybe. I really thought it is the same. Somebody wrote it in the forum, will post the link later. Can you explain the difference to me please? Sent from my iPhone using CFD Online Forum mobile app

May 24, 2017, 19:28		#10
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,869 Rep Power: 144	I think I have explained what they appear to be in each post I have put on this thread.