[avd28]

Hi all,

I'm trying to simulate diffusion of a species through a boundary wall in my system. I'm trying to set a user defined species mass fraction at the walls with a UDF using the DEFINE_PROFILE function.

The specified mass fraction is actually a function of the concentration gradient at the wall. Hence I need to use the cell variable macro C_YI_G, which stores the gradient at the cell wall.

The usage of this variable with pressure based solver involves using 2 seperate commands

set/solve/expert

and

(rpsetvar 'species/save-gradients? #t)

which I have already done.

I'm using a parallel system and a moving mesh and have included provisions for that in the UDF as well. The UDF is as follows

#include "unsteady.h"
DEFINE_ON_DEMAND(save2)
{
#if !RP_HOST 
    save2();
    Message0("\n\n   Done! \n   You can check this through plotting the contours of user defined node memory (e.g. print on walls)\n");
#endif /*!RP_HOST */
}

void save2()
{
    Domain *domain;
    cell_t c;
    Thread *t;
    Node *v;
    int n;
    domain=Get_Domain(1);
    /*Store the mesh node coordinates in user-defined node memory, this data is used in the dynamic mesh*/
    thread_loop_c (t,domain)
    {
        begin_c_loop (c,t)
        {
            c_node_loop (c,t,n)
            {
                v = C_NODE(c,t,n);
                N_UDMI(v,0) = NODE_X(v);
                N_UDMI(v,1) = NODE_Y(v);
                N_UDMI(v,2) = NODE_Z(v);
            }
        }
        end_c_loop (c,t)
    } 
}

DEFINE_PROFILE(conc_sink,t,i)
{
  cell_t c;
  face_t f;
  real cx;
  begin_c_loop(c,t)
  {
  cx = C_YI_G(c,t,i)[0];
  begin_f_loop(f,t)
  {
    F_PROFILE(f,t,i) = cx*0.003;
  }
  end_f_loop(f,t)
  }
  end_c_loop(c,t)
}

This code compiles and loads correctly, but when I try to initialize the problem, it crashes. I realize of course that an initizlializing a gradient variable at the beginning of a solution may pose some problems. Does anyone know how to deal with this, and also how to use such variables in a solution?

[pakk]

There is the clean but complicated solution, and the easy dirty solution.

Easy dirty: make sure that conc_sink is not called when you initialize. So unload the library before you initialize, or change the relevant boundary condition into something else, and change it back after you initialize.

Clean but complicated: Add a check in your code such that the gradient is only read if you have already initialized your data. Based on the Fluent manual, I would guess one of these three formats would work:

if (NNULLP(THREAD_STORAGE(t,C_YI_G))) {... }

if (DATA_VALID_P(C_YI_G)) { ... }

if (DATA_VALID_P()) { .... }

But I never tested this myself, so if you do, please tell me if which one (if any) worked.

[mahdi-united]

Originally Posted by pakk

There is the clean but complicated solution, and the easy dirty solution.

Easy dirty: make sure that conc_sink is not called when you initialize. So unload the library before you initialize, or change the relevant boundary condition into something else, and change it back after you initialize.

Clean but complicated: Add a check in your code such that the gradient is only read if you have already initialized your data. Based on the Fluent manual, I would guess one of these three formats would work:

if (NNULLP(THREAD_STORAGE(t,C_YI_G))) {... }

if (DATA_VALID_P(C_YI_G)) { ... }

if (DATA_VALID_P()) { .... }

But I never tested this myself, so if you do, please tell me if which one (if any) worked.

hi i wrote a udf for getting the species mass fraction gradients by define_adjust and using the storing commands as follow:


DEFINE_ADJUST(gradient, domain)
{
Thread *t;
Thread **pt;
cell_t c;
Domain *pDomain = DOMAIN_SUB_DOMAIN(domain,P_PHASE);
real voidx, voidy, voidz=0.0;


mp_thread_loop_c (t,domain,pt)
{
if (FLUID_THREAD_P(t) && n_udm > 0)
{
Thread *tp = pt[P_PHASE];
begin_c_loop (c,t)
{
C_UDMI(c,t,0) = C_YI_G(c,tp,0)[0];
}
end_c_loop (c,t)
}
}

}


this code compile and initialize correctly without any error
but in calculating in 1st iteration appear the message "the f1 process could not be started" and fluent crashes.
Where do you think the problem is?

[AlexanderZ]

what is it?

Domain *pDomain = DOMAIN_SUB_DOMAIN(domain,P_PHASE);

you dont use pDomain in your code

try DEFINE_EXECUTE_AT_END instead of DEFINE_ADJUST

[vinerm]

The code works fine. I wrote it and tested it at my end before sharing it. The issue is with the case not with the code.

[mahdi-united]

Originally Posted by AlexanderZ

what is it?

Domain *pDomain = DOMAIN_SUB_DOMAIN(domain,P_PHASE);

you dont use pDomain in your code

try DEFINE_EXECUTE_AT_END instead of DEFINE_ADJUST

hi
why do i use the DEFINE_EXECUTE_AT_END??
i want to store species mass fraction gradients.

[mahdi-united]

Originally Posted by vinerm

The code works fine. I wrote it and tested it at my end before sharing it. The issue is with the case not with the code.

how to send my case to you?

[AlexanderZ]

Originally Posted by mahdi-united

hi
why do i use the DEFINE_EXECUTE_AT_END??
i want to store species mass fraction gradients.

read manual first. stop wasting our time
ansys fluent customization manual

[mahdi-united]

Originally Posted by vinerm

The code works fine. I wrote it and tested it at my end before sharing it. The issue is with the case not with the code.

hi


i type this command but not work
solve>set>expert
reply yes to question keep memory from being freed?
and

(rpsetvar 'species/save-gradients? #t)


how to store the species mass fraction gradients in fluent?
what is the its another command?