CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Fluent UDF and Scheme Programming (http://www.cfd-online.com/Forums/fluent-udf/)
-   -   calculate distance from wall to neighbourig cell centroid (http://www.cfd-online.com/Forums/fluent-udf/63211-calculate-distance-wall-neighbourig-cell-centroid.html)

 rr123 April 1, 2009 04:24

calculate distance from wall to neighbourig cell centroid

dear all,

i need to calculate the distance between the wall and the centroid of the cell touching this wall. my system is 2D.

this is the main code snippet:

begin_c_loop (c,t)
{
c_face_loop(c,t,n)
{
{
f = C_FACE(c,t,n);
F_CENTROID(x,f,t);
wf = x[1]; /* for 2D */
C_CENTROID(cell,c,t);
nc = cell[1];
dist = FABS(wf-nc);
}
}
}
end_c_loop (c,t)

am i right ? is it correct ordering of the loop structure ?

best regards.

 HenrikS April 1, 2009 06:37

I think your code will do what you want, yes. If you have any problems with
you might consider using
although I guess they would work equally well.

Anyway, you could use C_WALL_DIST(c,t) directly instead of looping, I guess?

/Henrik

 rr123 April 1, 2009 06:41

dear henrik,

i didnt get you clearly with the using of the C_WALL_DIST(c,t).

can you remake the code snippet with your suggestion so that i can try from my end.

sincerely.

 HenrikS April 1, 2009 07:07

In your code, you calculate "dist" for every cell next to a wall boundary. My guess is that you want to use this value for some purpose elsewhere in your code. Now, there is a macro called C_WALL_DIST(c,t) which I believe (although I have not verified that) will give you this distance directly.

begin_c_loop (c,t)
{
dist = C_WALL_DISTANCE(c,t);
}
end_c_loop (c,t)

This is probably not only shorter but also quicker. Anyway, I suggest you validate this approach before relying on it. You can store "dist" in a UDM and compare to the cell wall distance in a contour plot in Fluent for example.

/Henrik

 rr123 April 1, 2009 07:47

dear henrik,

thanks for the reply. so you mean that the c_face_loop which i used, is no more required with your code snippet using C_WALL_DIST(c,t) ?

one more query:
i want to add a source term, S_epsilon in the epsilon eqn in the k-epsilon model. which DEFINE macro must i choose ?

many thanks in advance for the help.
best regards.

 HenrikS April 1, 2009 07:56

It was quite some time ago since I worked with wall distances in UDFs, but I seem to remember that the distance between a cell centre and the nearest wall should be available in C_WALL_DISTANCE(c,t). (As you can see, it is available for plotting under Grid...):

"Cell Wall Distance is the distribution of the normal distance of each cell centroid from the wall boundaries. Its unit quantity is length. "

Anyway, if this is what C_WALL_DISTANCE(c,t) returns, then all you need is the cell thread and index (c and t) to obtain it, i.e. no need to loop over faces and compare coordinates yourself.

To add a source term for the epsilon equation (to use inside your domain), write a DEFINE_SOURCE macro and hook it up under the source terms for the boundary conditions of the fluid zone.

Hope this helps!

/Henrik

 HenrikS April 1, 2009 07:58

You would also have to add some kind of check to see whether the current cell is next to a boundary, if you are only interested in boundary cells (which seems the case in your first post).

 rr123 April 1, 2009 08:05

yes, Henrik you are righ indeed. my only intention is to calculate the distance from the centroid of the cell which touches the wall.
so thats why i intended to use the THREAD_TYPE macro.

so if i use C_WALL_DIST(c,t) macro, what should i write to implement that check ?

sincerely.

 HenrikS April 1, 2009 08:11

begin_c_loop(c, t)
{
dist = 0.0;
c_face_loop(c, t, n)
{
{
dist = C_WALL_DIST(c,t);
}
}
if (dist > 0.0)
{
/* do whatever it is you wish to do with dist here... */
/* if dist = 0.0 then this cell does not touch the boundary */
}
}
end_c_loop(c, t)

or something like that... :-)

/Henrik

 rr123 April 1, 2009 08:13

henrik, thanks again.

but how do we know that BOUNDARY_FACE_THREAD is the wall ??

sincerely.

 HenrikS April 1, 2009 08:23

Hmm... BOUNDARY_FACE_THREAD_P(t) expands to a function that returns TRUE if Thread *t is a boundary face thread. I don't know for sure, maybe this would also include inlets and outlets. Then maybe stick with your initial code or supply an array containing the wall thread IDs, whatever you prefer.

 coglione April 2, 2009 03:16

Hello,

there is a macro THREAD_TYPE(tf) which returns an integer for each boundary face thread tf. If the boundary face is a wall this number will be "3"
--> if (THREAD_TYPE(tf)) == 3) can be used to check wether the boundary is a wall.

cheers

 rr123 April 2, 2009 03:32

hi,

is

the same as

sincerely.

 HenrikS April 2, 2009 03:42

so those should be the same.

 rr123 April 2, 2009 05:27

thanks Henrik and coglione !!

i have two small queries:

i am trying to add a secondary source term, S to the epsilon eqn of the k-epsilon model.
S = max[0.83(l/l_e)(epsilon/k),0]
where, l tubulence length scale and l_e is the is the near-wall equilibrium length scale.
i am using the DEFINE SOURCE(name,c,t,dS,eqn) macro.

so the query is do i need to define my dS to be ds[eqn] = 0.0 or shall i need to assign an expresison to it ?
i read the UDF manual, but is not clear for me.

and if i need to assign an expresison to dS, then with what should i
differetiate S. With l or with l_e ? i mean, what is the dependent variable in the above for S ?

 HenrikS April 2, 2009 06:18

You do not need to specify an expression for dS, you can set it to 0.0. If you have convergence issues, consider adding an expression for dS. As I understand it, dS should then be the derivative with respect to epsilon?

 coglione April 2, 2009 07:48

Yes, the dependent variable here is epsilon as the source is defined for this quantity. in this specific case a linearization of the source term will not add stability because this is only true if dS/depsilon is negativ which will never be possible since the definition of source as >0.0 ! Thus dS = 0.0 is the only possibility and FLUENT automatically does so if you do not specify dS at all.

cheers

 rr123 April 2, 2009 10:25

thanks for helping me henrik and max !
i keep you informed of this.

 rr123 April 3, 2009 10:39

dear henrik and max,

why should the S (in my case) be differentiated w.r.t epsilon ?
i mean, what is the reason behind it ??

because, in some examples i find that for modifying a source term in the omega equation (of k-omega model), the S is differentiated w.r.t. x-velocity.

pls help me catch what i am failing to understand.

kind regards.

 Mohamed Refaat January 7, 2015 17:56

Quote:
 Originally Posted by HenrikS (Post 211557) I think your code will do what you want, yes. If you have any problems with if(THREAD_TYPE(tf) == THREAD_F_WALL) you might consider using if (BOUNDARY_FACE_THREAD(tf)) although I guess they would work equally well. Anyway, you could use C_WALL_DIST(c,t) directly instead of looping, I guess? /Henrik
dear Henrik
i have the same problem of rr123
when i use this macro C_WALL_DIST(c,t) this message appeared

D:\PROGRA~1\ansys\ANSYSI~1\v145\fluent\fluent14.5. 0\win64\3d\fl1450s.exe received fatal signal (ACCESS_VIOLATION)
1. Note exact events leading to error.
2. Save case/data under new name.
3. Exit program and restart to continue.