# should y+ be upper than 30 in high-Re in all cells?

 Register Blogs Members List Search Today's Posts Mark Forums Read

June 8, 2013, 09:38
should y+ be upper than 30 in high-Re in all cells?
#1
Senior Member

Ehsan
Join Date: Oct 2012
Location: Iran
Posts: 2,208
Rep Power: 23
in my case I see in some areas yPlus is lower than 30 as attached.
but in most cells bigger than 30.is it a problem and I should raise yPlus by coarsening the mesh further?(its a narrow tube that furthur coarsening may lead to a not very good mesh)
Attached Images
 yPlus.jpg (14.1 KB, 31 views)
__________________
Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King.
To Be or Not To Be,Thats the Question!
The Only Stupid Question Is the One that Goes Unasked.

 June 8, 2013, 10:53 #2 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 you love posting... 8 topics in only 24 hours. In traditional k-epsilon, the y+ should not go below 11. The limit for the viscous sub layer is often set at y+ = 11 (if you go to the code, you can see that the laminar limit is y+=11). It is not good to have a y+ smaller than 11 because the flow is laminar in this region and since k-epsilon is only turbulent, your assumption become wrong. To avoid this, you can use K-epsilon RNG or Realizable because they are "extended" to the laminar region. However, it doesn't meet they are accurate in region with y+ < 11 ! They are just "better" and they avoid some crazy behavior (especially where 11 < y+ < 30. And for your other topic where you cannot see the boundary layer, don't use the contour ! You need to use a curve plot perpendicular to the wall (use the filter of intersection in paraview and re-scale the data where you want).

 June 8, 2013, 11:01 #3 Senior Member     Ehsan Join Date: Oct 2012 Location: Iran Posts: 2,208 Rep Power: 23 Hi Heckmann thanks. no,I love more to solve my problems sooner! thanks.but it seems I have forgotten that to say I'm using kOmegaSST. does all cells have to be greater than 30 exactly or its sufficient that most be greater? I didn't grasp what you said about BL. whould you please answer in that thread more clearly? __________________ Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King. To Be or Not To Be,Thats the Question! The Only Stupid Question Is the One that Goes Unasked.

 June 8, 2013, 11:11 #4 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 If you use kOmegaSST, ALL your cells MUST have a y+ smaller than 5 ! kOmegaSST is used to solve the viscous sub layer. On this purpose, you must have at least one point in this region. To get good results with kOmegaSST, you should have a maximum y+ of 1 or smaller (I usually aim 0.5).

 June 8, 2013, 11:33 #5 Senior Member     Ehsan Join Date: Oct 2012 Location: Iran Posts: 2,208 Rep Power: 23 so can't possible to use kOmegaSST in high-Re form? I think its revered! see it:http://www.openfoam.org/mantisbt/view.php?id=835#c2167 any comments? __________________ Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King. To Be or Not To Be,Thats the Question! The Only Stupid Question Is the One that Goes Unasked.

June 8, 2013, 11:36
#6
Senior Member

Ehsan
Join Date: Oct 2012
Location: Iran
Posts: 2,208
Rep Power: 23
Quote:
 And for your other topic where you cannot see the boundary layer, don't use the contour ! You need to use a curve plot perpendicular to the wall (use the filter of intersection in paraview and re-scale the data where you want).
it's similar to plot on line,yes?
__________________
Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King.
To Be or Not To Be,Thats the Question!
The Only Stupid Question Is the One that Goes Unasked.

 June 8, 2013, 11:51 #7 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 1) What is the point of using kOmegaSST in high-Re form (y+ > 30) ? Why not kOmega or even kEpsilon ? 2) yes, it's like plot over a line where you line start at the wall and ends just oustide of the boundary layer. You can get the profile.

 June 8, 2013, 12:18 #8 Senior Member     Ehsan Join Date: Oct 2012 Location: Iran Posts: 2,208 Rep Power: 23 1)i have read that kOmegaSST is better in separation phenomena,do you think its better to use pure kOmega? 2) then you mean I have to realize the limits of BL from the velocity profile,yes?there isn't maybe a more automatic way based on velocity or more dissipation or viscosity or something else? __________________ Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King. To Be or Not To Be,Thats the Question! The Only Stupid Question Is the One that Goes Unasked.

 June 8, 2013, 12:59 #9 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 1) in the case of a y+ < 5 only ! kOmega SST has been designed to work within the viscous layer, if you don't have a y+ < 5, you are outstide of the "domain of validity". 2) just draw a line that start outside your domain (1mm) and goes into your domain (1cm for example) and that is perpendicular to the wall. Then plot the velocity magnitude over this line. It will be equal to 0 at the wall and increase as it goes into the domain (and it will be null outside of the domain).

 June 8, 2013, 13:04 #10 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 And if I remember well (need to check), kOmega SST works also for y+ > 30 because it has a condition that switch between two models. If you y+ is too big, kOmega SST should behave as kEpsilon does (and therefore, you get all the disadvantages of kEpsilon).

 June 8, 2013, 13:23 #11 Senior Member     Ehsan Join Date: Oct 2012 Location: Iran Posts: 2,208 Rep Power: 23 OK.thanks.I'm going to do some runs by y+<1.I'm doing now grid independence for y+>30 method. how can I set y+<1?only one cell should be under y+<1 ?and if I fine the mesh for grid independence should y+'s be exactly the same(when bothare lower than 1)? and if I set k as 1e-10 or 1e-11 then what value should I set for omega? I read in two threads 1)one saying it should be omegaWallFunction with value 1e-10 2) in another it had came that it should be fixedValue with a high value by a formula. which should I use? __________________ Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King. To Be or Not To Be,Thats the Question! The Only Stupid Question Is the One that Goes Unasked.

 June 8, 2013, 14:33 #12 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 All your wall cells must be under 1 to get a good accuracy no matter what happen. When you run a grid dependency, you should change the number of point in the domain. You can refine the lateral number of point, but it's hard to change the wall distance. To make simple, your y+ must be lower than 5 but it also should not be too small (let say 0.1 for example). You can make a grid with an average y+ = 1 and another one with y+ = 0.2. The use of "fixedValue" for k comes from it's physical meaning. Since k tends to be infinitely small at the wall. It's not wrong to use the wallFunctions since the wallFunctions actually behave like fixedValue (or/and zeroGradient) that some users described, when the mesh is small enough. It's just that you do it manually. I always keep the omega wall function and sometimes when my mesh is fine enough, I use fixed k at 1e-12. No matter what, the wall functions are here to make your life easier so use them.

 June 8, 2013, 15:10 #13 Senior Member     Ehsan Join Date: Oct 2012 Location: Iran Posts: 2,208 Rep Power: 23 I want more to know about omega values if I set: Code: ```walls { type compressible::omegaWallFunction; Cmu 0.09; kappa 0.41; E 9.8; value x; }``` what should I replace in x position?a very low value like 1e-12?or very high value like 1e10? thanks. __________________ Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King. To Be or Not To Be,Thats the Question! The Only Stupid Question Is the One that Goes Unasked.

 June 9, 2013, 00:10 #14 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 Well, you should really study more because this kind of answer can be found quite easily... The value entry is here for two reasons. One it is for paraview that needs a value member to plot the data, and second, it is here because the wall function derive from a fixed value condition and therefore it must have the entry of the wall function but also from the boundary it is derived from. I'm this case the value has no influence on the computation. It is just here to give a value to paraview. The wall function will give the good value by himself. JR22 and bill_liu like this.

 June 9, 2013, 02:47 #15 Senior Member     Ehsan Join Date: Oct 2012 Location: Iran Posts: 2,208 Rep Power: 23 i have set it 1e-11.is it ok? The run is too slow compared with high-Re. How can find out flow really need a low-Re or not?my case is similar to shock tube case although very different BC's. __________________ Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King. To Be or Not To Be,Thats the Question! The Only Stupid Question Is the One that Goes Unasked.

 June 9, 2013, 03:57 #16 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 Open a boundary layer theory book and it will tell you how to "compute" the Reynolds number in the region close to the wall.

 June 9, 2013, 05:17 #17 Senior Member     Ehsan Join Date: Oct 2012 Location: Iran Posts: 2,208 Rep Power: 23 oh! I certainly know how to compute Re number. maybe you haven't heard about low-Re and high-Re grid? I mean that type of mesh and related turbulent method.notice? __________________ Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King. To Be or Not To Be,Thats the Question! The Only Stupid Question Is the One that Goes Unasked.

 June 9, 2013, 05:31 #18 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 I'm not talking about the Reynolds number of the flow but about the Reynolds number of the boundary layer. It is slightly different...

 June 9, 2013, 05:44 #19 Senior Member     Ehsan Join Date: Oct 2012 Location: Iran Posts: 2,208 Rep Power: 23 I asked a clear question... __________________ Injustice Anywhere is a Threat for Justice Everywhere.Martin Luther King. To Be or Not To Be,Thats the Question! The Only Stupid Question Is the One that Goes Unasked.

 June 9, 2013, 06:12 #20 Senior Member   HECKMANN Frédéric Join Date: Jul 2010 Posts: 238 Rep Power: 14 I certainly doubt that you can have a low Reynolds grid in a shock tube simulation since your cells are surely too large in the flow direction. The only way you can get "locally" a low Reynolds grid is in the case of an adverse pressure gradient that leads to a detachment (see the boundary layer Reynolds number). In this case, the boundary layer makes a reverse that locally slow down the fluid. Some cells can then be running at a "low Reynolds" regime as you say. Ps. if you really have a "low reynolds grid" in a shock tube, just run it in LES not in RANS.