[mkozic]

Applying ANSYS FLUENT discrete phase model (DPM) I obtained water droplet collection efficiency without difficulties for 2D body (circular cylinder with diameter of 10.16 cm, free stream velocity of 80 m/s and median volumetric diameter of 16 μm). Then I tried Eulerian approach using User-Defined-Scalars (UDS). The C source code was checked and initialization performed via DEFINE_INIT. Values on the boundaries and in the domain were according to those defined in DEFINE_INIT. But during iterations values of uds on the wall were not changed, whereas all cell variables were put to zero. I suppose that DEFINE_PROFILE didn't work properly, but I don't know why. I send a part of the source code consisting of the UDFs for initialization, wall boundary condition for volume fraction and source term for x-momentum equation only for Reynolds number between 0 and 0.1 for brevity. Has anyone had a similar problem ?

#include "udf.h"
#define W_DENS 998.2 /* water density */
#define ALFA_INF 1.3E-06 /* water volume fraction at inlet */
#define VX_INF 80.0 /* water x-velocity at inlet */
#define VY_INF 0. /* water y-velocity at inlet */
#define DROP_DIAM 16.0E-06 /* water droplet diameter */
#define AIR_VISC 17.8E-06 /* air viscosity */

DEFINE_INIT(init,mixture_domain)
{
Domain *subdomain,*dom_water;
int phase_domain_index;
int body_ID=3; /* body ID=3 */
c_t c;
face_t f;
Thread *c_thread;
Thread *t_body;
sub_domain_loop(subdomain,mixture_domain,phase_dom ain_index)
{
if(DOMAIN_ID(subdomain)==3) /* water DOMAIN_ID=3 */
thread_loop_c(c_thread,subdomain)
{
begin_c_loop_all(c,c_thread)
{
C_UDSI(c,c_thread,0)=ALFA_INF; /* water volume fraction in cell */
C_UDSI(c,c_thread,1)=VX_INF; /* water x-velocity in cell */
C_UDSI(c,c_thread,2)=VY_INF; /* water y-velocity in cell */
}
end_c_loop_all(c,c_thread)
}
}
dom_water = Get_Domain(3);
t_body=Lookup_Thread(dom_water,body_ID); /* body faces thread */
begin_f_loop(f,t_body)
{
F_UDSI(f,t_body,0)=ALFA_INF; /* water volume fraction at body face*/

F_UDSI(f,t_body,1)=VX_INF; /* water x-velocity at body face */
F_UDSI(f,t_body,2)=VY_INF; /* water y-velocity at body face */
}
end_f_loop(f,t_body)
}
DEFINE_PROFILE(alfa_wall,t_wall,i) /*water volume fraction at body */
{
face_t f;
real A[ND_ND];
real ax,ay;
real vx_abs_f,vy_abs_f,skpr;
begin_f_loop(f,t_wall)
{
F_AREA(A,f,t_wall);
ax=A[0];
ay=A[1];
vx_abs_f=F_UDSI(f,t_wall,1); /* water x-velocity at body surface */
vy_abs_f=F_UDSI(f,t_wall,2); /* water y-velocity at body surface */
skpr=vx_abs_f*ax+vy_abs_f*ay; /* scalar product of velocity vector
and surface normal vector */
if(skpr>0.)
F_PROFILE(f,t_wall,i)=F_UDSI(f,t_wall,0);
else
F_PROFILE(f,t_wall,i)=0.;
}
end_f_loop(f,t_wall)
}
DEFINE_SOURCE(x_mom_water,c,t_water,dS,eqn)
{
Thread *t_air,*mixture_thread;
int phase_domain_index=0;
real source_ux,air_dens,v_slip,rey_nu,cd,ef,ka;
real alfa_c,ux_water_c,uy_water_c,ux_slip,uy_slip;
mixture_thread=THREAD_SUPER_THREAD(t_water); /* mixture thread */
t_air=THREAD_SUB_THREAD(mixture_thread,phase_domai n_index); /* primary
phase thread*/
alfa_c=C_UDSI(c,t_water,0); /* water volume fraction in cell */
ux_water_c=C_UDSI(c,t_water,1); /* water x-velocity in cell */
uy_water_c=C_UDSI(c,t_water,2); /* water y-velocity in cell */
ux_slip=C_U(c,t_air)-ux_water_c; /* cell x-slip velocity */
uy_slip=C_V(c,t_air)-uy_water_c; /* cell y-slip velocity */
air_dens=C_R(c,t_air); /* cell air density */
v_slip=sqrt(ux_slip*ux_slip+uy_slip*uy_slip); /* cell slip velocity */
rey_nu=air_dens*v_slip*DROP_DIAM/AIR_VISC; /* Reynolds number */
if(rey_nu==0)
ka=0.;
else if(0.1>rey_nu>0.)
{
cd=24./rey_nu;
ef=cd*rey_nu/24.;
ka=18.*AIR_VISC*ef/(W_DENS*DROP_DIAM*DROP_DIAM);
}
source_ux=ka*alfa_c*W_DENS*ux_slip;
return source_ux;
}

[aminrbspace]

Hi,

1) my problem is computing collection efficiency with fluent.
I know, I must use UDS, also I have all of equations, but I dont know how could I define continunity for phase of water in UDS.

2) when eulerian_multiphase has actived in fluent, it will solve both equations of two phase(air and droplet disperse)couple , here why we must use UDS??(it has confused me)

I become happy to help me.

Regards,

Originally Posted by mkozic

Applying ANSYS FLUENT discrete phase model (DPM) I obtained water droplet collection efficiency without difficulties for 2D body (circular cylinder with diameter of 10.16 cm, free stream velocity of 80 m/s and median volumetric diameter of 16 μm). Then I tried Eulerian approach using User-Defined-Scalars (UDS). The C source code was checked and initialization performed via DEFINE_INIT. Values on the boundaries and in the domain were according to those defined in DEFINE_INIT. But during iterations values of uds on the wall were not changed, whereas all cell variables were put to zero. I suppose that DEFINE_PROFILE didn't work properly, but I don't know why. I send a part of the source code consisting of the UDFs for initialization, wall boundary condition for volume fraction and source term for x-momentum equation only for Reynolds number between 0 and 0.1 for brevity. Has anyone had a similar problem ?

#include "udf.h"
#define W_DENS 998.2 /* water density */
#define ALFA_INF 1.3E-06 /* water volume fraction at inlet */
#define VX_INF 80.0 /* water x-velocity at inlet */
#define VY_INF 0. /* water y-velocity at inlet */
#define DROP_DIAM 16.0E-06 /* water droplet diameter */
#define AIR_VISC 17.8E-06 /* air viscosity */

DEFINE_INIT(init,mixture_domain)
{
Domain *subdomain,*dom_water;
int phase_domain_index;
int body_ID=3; /* body ID=3 */
c_t c;
face_t f;
Thread *c_thread;
Thread *t_body;
sub_domain_loop(subdomain,mixture_domain,phase_dom ain_index)
{
if(DOMAIN_ID(subdomain)==3) /* water DOMAIN_ID=3 */
thread_loop_c(c_thread,subdomain)
{
begin_c_loop_all(c,c_thread)
{
C_UDSI(c,c_thread,0)=ALFA_INF; /* water volume fraction in cell */
C_UDSI(c,c_thread,1)=VX_INF; /* water x-velocity in cell */
C_UDSI(c,c_thread,2)=VY_INF; /* water y-velocity in cell */
}
end_c_loop_all(c,c_thread)
}
}
dom_water = Get_Domain(3);
t_body=Lookup_Thread(dom_water,body_ID); /* body faces thread */
begin_f_loop(f,t_body)
{
F_UDSI(f,t_body,0)=ALFA_INF; /* water volume fraction at body face*/

F_UDSI(f,t_body,1)=VX_INF; /* water x-velocity at body face */
F_UDSI(f,t_body,2)=VY_INF; /* water y-velocity at body face */
}
end_f_loop(f,t_body)
}
DEFINE_PROFILE(alfa_wall,t_wall,i) /*water volume fraction at body */
{
face_t f;
real A[ND_ND];
real ax,ay;
real vx_abs_f,vy_abs_f,skpr;
begin_f_loop(f,t_wall)
{
F_AREA(A,f,t_wall);
ax=A[0];
ay=A[1];
vx_abs_f=F_UDSI(f,t_wall,1); /* water x-velocity at body surface */
vy_abs_f=F_UDSI(f,t_wall,2); /* water y-velocity at body surface */
skpr=vx_abs_f*ax+vy_abs_f*ay; /* scalar product of velocity vector
and surface normal vector */
if(skpr>0.)
F_PROFILE(f,t_wall,i)=F_UDSI(f,t_wall,0);
else
F_PROFILE(f,t_wall,i)=0.;
}
end_f_loop(f,t_wall)
}
DEFINE_SOURCE(x_mom_water,c,t_water,dS,eqn)
{
Thread *t_air,*mixture_thread;
int phase_domain_index=0;
real source_ux,air_dens,v_slip,rey_nu,cd,ef,ka;
real alfa_c,ux_water_c,uy_water_c,ux_slip,uy_slip;
mixture_thread=THREAD_SUPER_THREAD(t_water); /* mixture thread */
t_air=THREAD_SUB_THREAD(mixture_thread,phase_domai n_index); /* primary
phase thread*/
alfa_c=C_UDSI(c,t_water,0); /* water volume fraction in cell */
ux_water_c=C_UDSI(c,t_water,1); /* water x-velocity in cell */
uy_water_c=C_UDSI(c,t_water,2); /* water y-velocity in cell */
ux_slip=C_U(c,t_air)-ux_water_c; /* cell x-slip velocity */
uy_slip=C_V(c,t_air)-uy_water_c; /* cell y-slip velocity */
air_dens=C_R(c,t_air); /* cell air density */
v_slip=sqrt(ux_slip*ux_slip+uy_slip*uy_slip); /* cell slip velocity */
rey_nu=air_dens*v_slip*DROP_DIAM/AIR_VISC; /* Reynolds number */
if(rey_nu==0)
ka=0.;
else if(0.1>rey_nu>0.)
{
cd=24./rey_nu;
ef=cd*rey_nu/24.;
ka=18.*AIR_VISC*ef/(W_DENS*DROP_DIAM*DROP_DIAM);
}
source_ux=ka*alfa_c*W_DENS*ux_slip;
return source_ux;
}

[mkozic]

Full eulerian multiphase model can not solve the collection efficiency problem properly, but UDS must be used because the wall boundary conditions are not standard. The wall boundary conditions must be prepared using UDF functions in order to allow the droplets that impinge the wall to escape out of the numerical domain.

[aminrbspace]

Hi,

Tnx a lot for your answer, but I have some eesay that they do it, with full eulerian mutiphase that define permeable wall for phase of water.could you help me for using UDS? I know UDS in fluent.

bests,

Originally Posted by mkozic

Full eulerian multiphase model can not solve the collection efficiency problem properly, but UDS must be used because the wall boundary conditions are not standard. The wall boundary conditions must be prepared using UDF functions in order to allow the droplets that impinge the wall to escape out of the numerical domain.

[rebeccakath]

Originally Posted by mkozic

Applying ANSYS FLUENT discrete phase model (DPM) I obtained water droplet collection efficiency without difficulties for 2D body (circular cylinder with diameter of 10.16 cm, free stream velocity of 80 m/s and median volumetric diameter of 16 μm). Then I tried Eulerian approach using User-Defined-Scalars (UDS). The C source code was checked and initialization performed via DEFINE_INIT. Values on the boundaries and in the domain were according to those defined in DEFINE_INIT. But during iterations values of uds on the wall were not changed, whereas all cell variables were put to zero. I suppose that DEFINE_PROFILE didn't work properly, but I don't know why. I send a part of the source code consisting of the UDFs for initialization, wall boundary condition for volume fraction and source term for x-momentum equation only for Reynolds number between 0 and 0.1 for brevity. Has anyone had a similar problem ?

I want to obtain collection efficiency by DPM method. But I don't know how to get the parameter.　Could you tell me about your method? Do you use UDF obtain the velocity and position of the particles?