[Jim87]

Hallo,

Modelling the properties of high viscosity materials (polymer melt), leads to bad residuals and divergence in my test-models.

I use a UDF for the compressibility of the polymer fluid. This UDF (dependents of temperature and pressure) works really fine at low viscosity (air , water etc). When I am going up with may viscosity (over 1 Pas) every solver that I tested leads to a divergence (usually pressure correction).

- I tested pressure and density based solvers
- I screwed the solution contolls to a absurd low level
- I changed timesteps up and down.
- I used soft uprising setting to suspend wrong initial conditions


Does anyone have experience in modelling high viscosity models and have an advice for me. Even my Ansys support is insecure what the problem might be.

[Firas]

May you give brief description of your model to see how can I help you ..I dealt with hight viscosity before .

[Jim87]

The Model:


Geometrie:The testroom is 5*20*40 mm. 4000 identical hexahedrons.

Boundary Condition: In my easiest case there is only a velocity inlet (rather slow 0,01 m/s). I disabled the pressure outlet in oder to raise the compression effect. This works very very fine at low viscosity with good results and nearly perfect residuals.

- The walls are at a lower temperature than the fluid (450K)
- realising a velocity outlet leads not to an solution

--------------------------------------------------------------------------
Material: In my opinion the compressive UDF may be the origin of all problems.

Density is turned to an UDF function:

#include "udf.h"

DEFINE_PROPERTY(UDF_4Koeff_bar_cm_,c,t)
{
/************************************************** **************************
Standard UDF Order
************************************************** ***************************/
double pressure = 0.0;
double temp = 0.0;
double s1 = 19598.341796875;
double s2 = 2.7665386199951;
double s3 = 5604.6762695312;
double s4 = 22733.578125;

double spezVolu = 0.0;
double Rho = 0.0;

/************************************************** **************************
point and select temperature and pressure
************************************************** ***************************/

temp = C_T(c,t);
pressure = C_P(c,t);

/************************************************** *************************
equation
************************************************** ***************************/

spezVolu = s1 / (s4 + pressure) + (s2 * temp) / (s3 + pressure);
Rho = 1/spezVolu;
Rho = Rho * 1000;

return Rho;
}

/************************************************** **************************
Gleichung 4 Koeffizienten nach Schmidt
************************************************** ***************************/

The viscosity will be an UDF of a carreau fluid (testet in many projects, this one works)


Solver settings are pressure ore density based ansys standard cases. All Problems need to run transient and with time step sizes of ca. 0.001 sec.

If there are further questions I'll try to give detailed reports.

[Jim87]

I tried my best on finding a solution and reworked the udf and the material datas. The problem of dealing high viscosity could be handeld by very very low under relaxation factors.

But from viscosities of 1 Pas and more my hole setup and simulation becomes weird. Values that I fixed in the material data udf by maximum and minimum loops change their values and the residuals rise to cray heights (see figur [viscosity to iterations]).

Can anyone imagine a reason why high viscosities seem not capable in Fluent? My UDF's work brilliant for less viscosity (figure). Maybe there is a solver that I should use or setting fpr dealing these materials?

[flotus1]

Fluent can handle high viscosities.

I dont know what your UDF does, but for finding the cause of your problem it might be better to deactivate it for now.

I can imagine that the way you increase the viscosity in your simulation causes problems. Starting with a very low viscosity, the flow will be turbulent. Thus the simulation is an under-resolved DNS and eddies start to build up.
Now increasing the viscosity on this flow field is very likely to cause trouble or even lead to divergence.
Have you tried starting the simulation with a high viscosity from the beginning?

[Jim87]

Hallo,
The probelm seems to be the combination of compression and high viscosities.

I tried high and constant viscosities at the beginning but the residuals are as high as you can see in the attached figgure. I need eveb lower under relaxation factors even to get those bad residuals, otherwise the simulation diverges after a few iterations.

There needen't be a UDF for my viscosity. High and constant usually leed to this high residuals and very bad solutions.

-> Do you choose special solvers and settings for high viscosities in unsteady (transient) simulations?

[flotus1]

Wait a minute... What boundary conditions did you use? According to one of the previous posts, you use velocity inlet and velocity outlet? This is under-constrained if the fluid is compressible.

[Jim87]

The discussed cases are testcases, I used different boundaries to check the UDF's. Generall the problem is detected in this cases

1. Velocity inlet, pressure outlet. -> checking the flow properties.

2. Velocity or Pressure inlet and no outlet -> checking the compressibility.

3. Pressure outlet and no inlet (patching the pressure). -> ckecking time dependent decompression value.

Velocity In and Outlet at the same time shouldn't work,I agree.

The final viscosity modell is a carreau model. The viscoisty should rang from 300 Pas to 1 Pas.

[blackmask]

Maybe you should reduce your time step by an order of square root of (Re) when you switch from the large Re number to small number? Also you might need to adjust the preconditioning parameters, if there is any.