[snsd03090412]

Hi! Gus. I am working on the simulation of selective laser melting for time dependent heat source term in DEFINE_SOURCE Macros. I want to simulate one track situation to know the distribution of temperature. My volumetric heat source formula is as follows . The size of substrate is 0.5m*0.5m*0.3 m. The initial laser xyz position ( 0.01 m, 0.03 m, 0.028 m) move to the final position ( 0.04 m, 0.03 m, 0.028 m) as time goes by. But, after I interpret my udf, the result shows floating point exception. I don't know why it doesn't work, can any of you guys give me some suggestions?
#include "udf.h"
DEFINE_SOURCE(gaussian_heat_flux,c,t,dS,eqn)
{
float A = 0;
real x[3];
real se;
real s=0.000006;
real source; // heat flux
real d=0.000006;
real p = 450; // laser power
real n = 0.45; // process efficiency;
real a = 0.0015; // deposition half width
real b = 0.0009; // melt pool depth
real cc = 0.0015; // longitudinal ellipsoid axis c.
real FF = 1; // scaling factor
real vel = 0.01; // laser beam velocity
real time; // time
real x_coor; // x_coordinates
real y_coor; // y_coordinates
real z_coor; // z_coordinates
real x_ini = 0.01; // x_ini
real y_ini = 0.03; // y_ini
real z_ini = 0.028; // z_ini
real absorp = 0.4; // absorptivity of ss316
real circle = 3.14; // PI
real emis; // emissivity
real sigma = 0.0000000567; // stefan-boltzmann
real Tinf = 300.0; // environment temp
real coeff_h; // heat transfer coefficient
real dummy; // dummy variable
time = RP_Get_Real("flow-time");

C_CENTROID(x,c,t);
x_coor = x[0];
y_coor = x[1];
z_coor = x[2];
if (vel*time<=0.04)
{
dummy = 3 * pow((x_coor - x_ini - vel*time), 2) / pow(a, 2) + 3 * pow((y_coor - y_ini), 2) / pow(b, 2) + 3 * (z_coor - z_ini ) / pow(cc, 2);
se = (6 * pow(3, 0.5)*p*n*FF) / (a*b*cc*circle*pow(circle, 0.5));
source = se*exp(-dummy);
dS[eqn] = 0.0;
return source;

}
else
{
source = 0;
dS[eqn] = 0;
}}
DEFINE_PROPERTY(ss316_density, c, t)
{
real rho_lam;
real temp = C_T(c, t);
if (temp <= 1658)
{
rho_lam = -0.5212*temp + 8126.4;
}
else if (temp>1658 && temp <= 1723)
{
rho_lam = -5.9702*temp + 17161;
}
else if (temp > 1723)
{
rho_lam = -0.7653*temp + 8192.8;
}
return rho_lam;
}
DEFINE_PROPERTY(ss316_conductivity, c, t)
{
real rho_lam;
real temp = C_T(c, t);
if (temp <= 1658)
{
rho_lam = 0.0141*temp + 10.665;
}
else if (temp>1658 && temp <= 1723)
{
rho_lam = -0.0855*temp + 175.88;
}
else if (temp > 1723)
{
rho_lam = 0.0143*temp + 3.8433;
}
return rho_lam;
}
DEFINE_INIT(initial, d)
{
cell_t c;
Thread *t;
/* loop over all cell threads in the domain */
thread_loop_c(t, d)
{
/* loop over all cells */
begin_c_loop_all(c, t)
{
C_T(c, t) = 300.0;
}
end_c_loop_all(c, t)
}
}

[rishitosh]

hey snsd03090412!!
did u solve the problem??
I am trying solve similar kind of problem..
using your idea...
#include "udf.h"
#include "metric.h"
#include "math.h"

#define P 7000 /*peak power of laser*/
DEFINE_SOURCE(gaussian_heat_flux,c,t,dS,eqn)
{
real x[2];
real Q, a1, b, time, vel,dt;
real x1, x2, y[2];
a=b= 0.003;
time = RP_Get_Real("flow-time");
vel = 1e-3;
C_CENTROID(y,c,t);
x1 = y[0]; /*xmax=0.0005 ans xmin=-0.0005*/
x2 = y[1]; /*ymax=0 and ymin=-.0003*/
if (vel*time<=0.005)
{
Q=(3 * P / (M_PI*a1*b))*exp(- (3 * pow(x1-vel*time, 2) / pow(a1, 2)) - (3 * pow(x2, 2) / pow(b, 2)));
dS[eqn] = 0.0;
return Q;
}
else
{
return 0;
dS[eqn] = 0;
}
}

but its is not wrking well...
can you help in this regard..
thank you.

[pakk]

Originally Posted by rishitosh

a=b= 0.003;

This is not allowed. (Or it might technically be valid code, but will do something completely different then what you expect.)

I think you mean:

a=0.003;
b=0.003;

[obscureed]

I do not agree with you on this, Pakk -- that looks like perfectly valid C to me, and it should do the same as the two-line alternative that you propose. See https://en.wikipedia.org/wiki/Operat...ment_operators
There is a problem that variable "a" is not declared in either case -- I don't know why, but the variables are "a1" and "b". It surprises me that this file gets through the compiler without errors.

A serious error is that size-3 arrays need to be declared as "real y[3];", for example, so that y[0], y[1] and y[2] are available. (It seems not to matter for x[2], since x is not used.)

A (possibly benign, possibly dangerous) error is that no value is returned in the "else" branch.

I would advise you to check the equation for Q -- there is nothing clearly wrong, but... there seem to be some significantly different lengthscales in operation (0.3m to 0.0003m), which is possible; the factors of 3 are surprising (for a standard definition of widths "a1" and "b"); and the laser appears to be centred on 0.0 for the y-coordinate "x2". (It looks possible that exp(XXX) could evaluated for some large negative values of XXX -- I would hope that these would safely underflow to zero.)

Ed

[pakk]

I stand corrected...

[pakk]

By the way: the "Fluent-way" of defining a variable for position is

real y[ND_ND];

ND_ND here means the number of dimensions. In this way, if you go to a two-dimensional case and are able to use the same code, it uses a tiny little bit less memory. In most cases not enough to be noticeable, but it might be a good habit to use it always.

And although the "y[2]" part is definitely wrong in this code, I don't see how it could lead to a floating point exception... Are you compiling or interpreting?

[sumantam22]

Hi everyone,
I have written the UDF for moving gaussian heat source and after compiling it into fluent my temperature profile is not changing with position. Can anybody please guide me!!!!


#include "udf.h"
DEFINE_PROFILE(gauss2D_heat_flux, thread, position)
{
real x[ND_ND];
face_t f;
real current_time;
real Q = 250;
real PI = acos(-1);
real vel = 1e-3;
real r = 0.002;
real B = -3;
real A;
real d;
real dt;
real x_pos;
current_time = CURRENT_TIME;
dt = CURRENT_TIMESTEP;
A = (2 * Q) / (PI * pow(r, 2.));
begin_f_loop(f, thread)
{
F_CENTROID(x, f, thread);
x_pos = vel * current_time;
F_PROFILE(f, thread, position) = A * exp(B * (pow((x[0] - x_pos), 2.) + pow(x[1], 2.)) / pow(r, 2.));
}
end_f_loop(f, thread)
}

[AlexanderZ]

code seems to be correct,
compile, load udf, hook profile as a source, check units (for length)