Learning Forum

[ll00023]

I use this code to calculate stokes number.

settings:

reporting User Value 0. (I suppose this is the stokes number I get. but I don't know why its name is this one)

it works in the case EBU example from FLUENT.

but for my case with 2 million mesh and 40 million particle parcels, it crashed.

I can't post all the code here. the code is correct. so what setting should I change?

#include "udf.h"
#include "dpm.h"

DEFINE_DPM_SCALAR_UPDATE(StokesNumber,c,t,init,p)
{
    cphase_state_t *cp;
    cp = p->cphase;
    /* cphase_state_t *c = &(p->cphase); */

   ......

 

    /* Stokes Number */
    P_USER_REAL(p,0) = tau_p / tau_f;

}

[Rob]

Please confirm, 40 million parcels? I suspect you just ran out of storage array or RAM. 

[ll00023]

This is the memory for 88 cores. the memory is enough for simulation.

If I use Define_dpm_scalar_update, does the tp stores all the particle parcel info?

I have tested the code in URANS case of EBU example. but it didn't work in my URANS and LES case.

 

[Rob]

OK, so the code works on a small model. Assuming the set up in the large model is identical then the only difference is size/count. With 40M parcels I can easily see the memory allocation being exceeded. Retry with significantly fewer parcels and see what happens. 

[ll00023]

I tested in RANS case, which has 15120 particle parcel from dpm summary report.

this could exclude particle parcel number's effect. but I got similar error. I might need to test a simpler code for this scalar later. 

Thank you Rob, appreciate your reply.

[Rob]

OK, thanks. So, if the code is also failing on a smaller model the problem is with the UDF in some way. 

One quick check, add a few more UDM and UDS memory locations. As I don't know what else you're calling there's always a risk you missed something: this is a test and when running working code I assign the correct number of locations! 

[Rob]

Continues here /forum/forums/topic/udf-crash/