[sfn]

Here is example udf from ANSYS Fluent 17 help (2.5.1.3. Example 1):

/* reflect boundary condition for inert particles */
#include "udf.h"
DEFINE_DPM_BC(bc_reflect,p,t,f,f_normal,dim)
{
real alpha; /* angle of particle path with face normal */
real vn=0.;
real nor_coeff = 1.;
real tan_coeff = 0.3;
real normal[3];
int i, idim = dim;
real NV_VEC(x);

#if RP_2D
/* dim is always 2 in 2D compilation. Need special treatment for 2d
axisymmetric and swirl flows */
if (rp_axi_swirl)
{
real R = sqrt(P_POS(p)[1]*P_POS(p)[1] +
P_POS(p)[2]*P_POS(p)[2]);
if (R > 1.e-20)
{
idim = 3;
normal[0] = f_normal[0];
normal[1] = (f_normal[1]*P_POS(p)[1])/R;
normal[2] = (f_normal[1]*P_POS(p)[2])/R;
}
else
{
for (i=0; i<idim; i++)
normal[i] = f_normal[i];
}
}
else
#endif
for (i=0; i<idim; i++)
normal[i] = f_normal[i];

if(p->type==DPM_TYPE_INERT)
{
alpha = M_PI/2. - acos(MAX(-1.,MIN(1.,NV_DOT(normal,P_VEL(p))/
MAX(NV_MAG(P_VEL(p)),DPM_SMALL))));
if ((NNULLP(t)) && (THREAD_TYPE(t) == THREAD_F_WALL))
F_CENTROID(x,f,t);

/* calculate the normal component, rescale its magnitude by
the coefficient of restitution and subtract the change */

/* Compute normal velocity. */
for(i=0; i<idim; i++)
vn += P_VEL(p)[i]*normal[i];

/* Subtract off normal velocity. */
for(i=0; i<idim; i++)
P_VEL(p)[i] -= vn*normal[i];

/* Apply tangential coefficient of restitution. */
for(i=0; i<idim; i++)
P_VEL(p)[i] *= tan_coeff;

/* Add reflected normal velocity. */
for(i=0; i<idim; i++)
P_VEL(p)[i] -= nor_coeff*vn*normal[i];

/* Store new velocity in P_VEL0 of particle */
for(i=0; i<idim; i++)
P_VEL0(p)[i] = P_VEL(p)[i];

return PATH_ACTIVE;
}
return PATH_ABORT;
}

This udf is work, but i have a question about "alpha" variable. Why do they use it? I mean they just calculate the value and don't use it in further calculations. Without this line

alpha = M_PI/2. - acos(MAX(-1.,MIN(1.,NV_DOT(normal,P_VEL(p))/ MAX(NV_MAG(P_VEL(p)),DPM_SMALL))));

i will get the same results. Am I wrong or is it little error in the code?

[pakk]

You are correct in saying that without this line, you will get the same results. The line has no effect on the simulation.

(I even suspect that the compiler recognizes this, and that the line is completely ignored.)

But I would not call it an error. An error is something that (potentially) causes problems. This line will not cause problems. And you have to see it in the context of a manual: this specific UDF does not use the angle alpha, but if somebody wants to change the UDf to make it angle-dependent, the angle is already there.

[sthinmind]

As pakk said Alpha is not used in this code. The definition of Alpha however is wrong. The way that Alpha is defined, it is the angle of particle path with the surface itself and not with the surface normal.
or "angle of particle path with face normal"=pi/2-Alpha