[saosao]

Hi
I am new in udf and I want to exponentiate the "C_U_RG(c,t)[1]" to 2.5. But the fluent gives this error:
Error: Divergence detected in AMG solver: uds-0
Error Object: #f

when I want to exponentiate the "C_U_RG(c,t)[1]" to 2.0, it works very well.
what is wrong about my udf? my udf is as follows:

#include "udf.h"
enum
{
T22,
N_REQUIRED_UDS
};
real y;
real z;


DEFINE_SOURCE(T22_source, c, t, dS, eqn)
{
y=C_U_RG(c,t)[1];
z= pow(y,2.5);
dS[eqn] = -1;
return z - C_UDSI(c, t, T22);
/*Source*/

}

DEFINE_SOURCE(ymom_source, c, t, dS, eqn)
{

dS[eqn] = 0.;
return C_UDSI(c, t, T22);

}

[AlexanderZ]

You UDF seems correct to me, most likely this is model problem, change convergence criteria, under-relaxation factors, time step, mesh.
your code

DEFINE_SOURCE(T22_source, c, t, dS, eqn) 
{ 
y=C_U_RG(c,t)[1];
z= pow(y,2.5);
dS[eqn] = -1; 
return z - C_UDSI(c, t, T22); 
/*Source*/ 

}

how did you get dS[eqn] = -1; ???
try to change to dS[eqn] = 0;

best regards

[saosao]

the problem seems to be related to float number 2.5 in "pow(y,2.5)". when the power is changed to 2.0 (pow(y,2.0)) the code is working very good.

[obscureed]

Hi Saosao,

I think to use pow, you need to put

#include "math.h"

at the start of the UDF.

It surprises me that pow(x,2.0) is OK, but it is just about possible that the compiler automatically simplifies pow(x,2.0) into (x*x).

The manual explains that C_U_RG(c,t) is not necessarily available -- but presumably you have followed the instructions about "solve/set/expert...Keep temporary solver memory from being freed".

In any case like this, you can try to debug the problem just by printing to the terminal window, using Message. If the case is large, so that printing a message for every cell would be overwhelming, I tend to select one in every 1000 cells (or even 100000 etc), by an arbitrary test on the integer c -- please see below.

Good luck!
Ed

#include "udf.h"
enum
{
   T22,
   N_REQUIRED_UDS
};

DEFINE_SOURCE(T22_source, c, t, dS, eqn)
{
  real y; /* Declare variables inside function where possible. */
   real z;

   y=C_U_RG(c,t)[1];
   z= pow(y,2.5);
   dS[eqn] = -1;
   if(c%1000==37) /* Select ~1 in 1000 cells; 37 is arbitrary */
     Message("In T22_source: y=%12g z=%12g T22=%12g\n",
       y,z,C_UDSI(c,t,T22));

   return z - C_UDSI(c, t, T22);
}

[saosao]

It does not work.
I think the error related to the power 2.5, which causes divergence in the begining of the solution. Is there any way to substitude pow(y,2.5)?

[obscureed]

Hi Saosao,


Try the Message command (and the include) that I suggested, and check that the pow function is working as intended.


Note that pow(x,2.5) will fail if x is negative (which pow(x,2.0) might not). So, if x could ever be negative, you need to provide some safeguards.


Once you have pow working, then any divergence etc is related to the wider problem, as AlexanderZ has already pointed out.


Good luck!
Ed

[saosao]

Thank you very much
The problem was solved. As you mentioned, the error is related to negative form of x.