[Pyrokrates]

Hi,


I would like to calculate the gradient of some scalar value, stored in UDMI.


The way I found (User guide + forum posts) is some kind of transformation to UDSI and saving UDSI_G afterwards to UDMI.


I tried this but it didnt work for me... Do I require face values also?


First I use some DEFINE_ADJUST to initialize my UDSI:

DEFINE_ADJUST(DA_UDSI, d)
{
    Thread *t;
    cell_t c;
    face_t f;

    thread_loop_c (t,d)
    {
        begin_c_loop (c,t)
        {
            C_UDSI(c,t,0) = C_UDMI(c,t,0);
        }
        end_c_loop (c,t)
    }
}

In the next step, I would like to set the gradient values to UDMI:

DEFINE_ON_DEMAND(DOD_UDMI)
{
    Thread *t;
    cell_t c;
    Domain *d = Get_Domain(1);

    thread_loop_c (t,d)
    {
        begin_c_loop (c,t)
        {
            C_UDMI(c,t,1) = C_UDSI_G(c,t,0)[0];
            C_UDMI(c,t,2) = C_UDSI_G(c,t,0)[1];
            C_UDMI(c,t,3) = C_UDSI_G(c,t,0)[2];

        }
        end_c_loop (c,t)
    }
 }

For using UDS, I disable the calculation of UDS in solver menu.


In CFD-Post, I use some contour on plane for visualization. While C_UDMI(c,t,1) shows me some results, I get "holes" inside my plane for C_UDMI(c,t,2) and C_UDMI(c,t,3), as there are some cells where I get no value.




Thanks in advance for any help


Pyro

[Siba11]

Can you try this instead:

Switch on the UDS equation, and allocate 3 memory locations (since you'll be using C_UDMI thrice)

#include "udf.h"

DEFINE_ADJUST (name,d)
{
#if !RP_HOST  
Thread *t;
cell_t c;
face_t f;

thread_loop_c (t,d)
    {
        begin_c_loop (c,t)
        {
            C_UDMI(c,t,0) = C_UDSI_G(c,t,0)[0];
            C_UDMI(c,t,1) = C_UDSI_G(c,t,0)[1];
            C_UDMI(c,t,2) = C_UDSI_G(c,t,0)[2];

        }
        end_c_loop (c,t)
    }
#endif
}

This will save your x,y,z gradients in the 3 memory locations. Now you can plot the C_UDMI contours.

[Pyrokrates]

Hi and thank you for the answer.


I don´t understand the benefit or change, for doing this...


If I turn on the claculation of UDS, than my value changes over time as I have unsteady term and flux, ...


I already allocate enough memories and scalars.
My data are stored in UDMI 0, I write them to UDSI 0 and than I save them in UDMI 1, UDMI 2 and UDMI 3. So I use 4 UDMI and 1 UDSI.


Thanks in advance


Pyro

[Siba11]

C_UDSI is a solver macro that fetches you the value of your scalar variable. You cannot alter or change the value of C_UDSI by assigning it some value.

C_UDSI(c,t,0) = C_UDMI(c,t,0) 

//the above statement is wrong. It's no better than:

int a;
2 = a;

The correct assignment would be :

C_UDMI(c,t,0) = C_UDSI(c,t,0);