Learning Forum

[Max]

Hi Fluent experts,

I am trying to make the UDF for the Realizable k-epsilon turbulent viscosity based on the paper of Shih et al. and that expressed in the Fluent theory guide.

The UDF code is for 2D simulation as follows:

#include "udf.h"
#include "math.h"

DEFINE_TURBULENT_VISCOSITY(tur_vis,c,t)
{
  real mu_tur;
  real S11, S12, S22, W12, W12r, S;
  real SMU, WAS, PHI, USTAR;
  real AS;
  real CMU, a1, a2, a3, a4, a5, a6;
  real PHI2, F2;

/***/
  real betas=0.09;
  real A0=4.04;
  real rho=C_R(c,t);
  real k=C_K(c,t);
  real w=C_D(c,t);
  real CMUL=C_MU_L(c,t);
  real y=C_WALL_DIST(c,t);

/***/
  real wr=-1.485446130;       /***rotation speed unit: 1/s***/

/***/
  S11=0.5*(C_DUDX(c,t)+C_DUDX(c,t));
  S12=0.5*(C_DUDY(c,t)+C_DVDX(c,t));
  S22=0.5*(C_DVDY(c,t)+C_DVDY(c,t));
 
  W12=0.5*(C_DVDX(c,t)-C_DUDY(c,t));

/*** Cal vorticity with angular velocity wr ***/
 
  W12r=W12-wr;

/*** Cal U*  ***/

  USTAR=sqrt(S11*S11+2.0*S12*S12+S22*S22+2.0*W12r*W12r);

/*** Cal Phi ***/

  SMU=sqrt(S11*S11+2.0*S12*S12+S22*S22);
  WAS=(pow(S11,3)+pow(S22,3)+3.0*pow(S12,2)*(S11+S22))/pow(SMU,3);
  PHI=1.0/3.0*acos(max(-1.0,min(sqrt(6.0)*WAS,1.0)));

/*** Cal As ***/
 
  AS = sqrt(6.0)*cos(PHI);

/*** Cal CMU ***/
 
  CMU=1.0/(A0+AS*USTAR*k/w);

/*** Cal Viscosity ***/

  mu_tur=CMU*rho*k*k/w;
  
  return mu_tur;
}

The simulations were performed for both the UDF and the default viscous model. However, there is a large deviation between these simulations' averaged results. I do not know why and why this difference is because the UDF code was based on the expression in the Fluent theory guide. I also compared the simulation between the UDF code and the default turbulent viscosity of the SST k-omega model and got a good agreement between the results.

Could anyone please help?

Thank you very much.

[Rob]

I'm assuming PHI is in radians? Otherwise the code doesn't look to have any obvious problems. I'd favour more intenting and comments for future reference and I'd tend to avoid 2-3 character variables where possible in case anything overlaps with a Fluent value. 

I've not had a detailed look, and won't be doing any debugging. If you plot contours, lines  etc are the results off by an offset, factor or just random differences?

[Max]

 

 

Thank Rob,

I think PHI is in radian.

I compare the torque on the object versus the simulation time which shows the difference in the case of Realizable k-e (see the figure – series 1 is UDF and series 2 is the defaul model) but not for SST k-w.

 

 

[Rob]

That looks like an offset, how constant is that value?