Learning Forum

[cibi]

Hello everyone,

I am running a steady state simulation of a fluid flowing inside a small tube within a bigger tube and I am using a multiphase model with the following settings:

As boundary conditions I have got for the secondary phase volume fraction = 1 at the inlet of the small tube and volume fraction = 0 at the inlet of the bigger tube.

After running my simulation, even if the residuals seem to have reached convergence it looks like my fluid never reached the end of the small tube:

If I run my simulation even just for 10 additional iterations, the contour is already looking better:

I was wondering why I don't achieve this result after running more than 1000 iterations but I do with just a few iterations?

These are the solution settings that I am using which I think are the recommended ones for steady state multiphase simulations:

I am not sure what I am doing wrong and what I am missing

Thank you!

[Rob]

What changed between the first and second run?

[cibi]

Hi Rob, I didn't change anything in the settings. The only difference is that the second run starts from the solution of the first run (I set 10 iterations withouth initializing the solution)

[Rob]

Odd, so, you wrote out case & data, read back in and did 10 iterations? Have you got the residual plot from the restart?

[cibi]

I didn't read it back but continued on the same Fluent window. This is the residual plot, it was just a test to see if something changed so maybe I should run it for more iterations than 10?

 

[Rob]

Run for more iterations. The massive spike in VOF isn't ideal, and suggests something isn't working right. What are the UDS for? 

[cibi]

Hi Rob, I've run again the simulation these are the residuals after 1250 iterations:

And these are the residuals after adding 600 iteations on top:

I ran the first 1250 itrations in batch mode through Workbench in Optislang and the other 600 manually by opening the workbench file and changing the number of iterations. 

The UDS are for the mean age of air and for the PLI calculation, I've got a spike in them as well (these are reported in the same cell zone of my domain and on the mixture phase). In the first two iteration I get the message:         Stabilizing uds-0 to enhance linear solver robustness.
        Stabilizing uds-1 to enhance linear solver robustness. and after this the values seem to be okay.

 

[Rob]

That spike really isn't good - look at the scale on the plots. What is PCI? 

[cibi]

Yes the spike is not good but I'm not sure what is causing it. I've computed the same UDS in other simulations with different geometry without problems

PLI is the platelet lysis index which I am computing like this: 

shear_stress = C_MU_L(c, t)*C_STRAIN_RATE_MAG(c, t);
source1 = C_R(c, t)*(pow(3.66*pow(10, -6), 1 / 0.77)*pow(shear_stress, 3.075 / 0.77));

 

 

 

 

[Rob]

Where is the source applied? Can you plot shear stress on the odd results and check you have the amount of source you think you have?

[cibi]

It's applied to the mixture. I could compute the PLI using the formula but I am not sure on the value of time to use because the simulation is steady state.

However, I have run the same simulation without the multiphase model and those huge spikes are gone:

 

The final values of the UDS do not change between the two simulations though, for example for the avg scalar 0 I get this with the multiphase model:

And this is from the simulation without multiphase model:

 

 

[Rob]

Won't the terms be phase (material) based? What is the source doing? Ie what's it attached to?

[cibi]

I am using the multiphase model just as a tracking tool for the blood entering the vein through the small tube, I have seen it in this paper by de Oliveira (2020):

So since my phases are identical I applied the source to the mixture.

This is how I defined the sources (one for residence time using the mean age of air theory and one for PLI)

 

 

[Rob]

Yes, maa goes to scalar mass, but what is the sigma attached to? 

[cibi]

Sigma is also attached to scalar mass

[Rob]

OK thanks, so the sources are linked solely to the UDS which isn't then attached to any fluid properties: just needed to clarify. 

There's no reason that the solver should stick like that, equally, I'd expect it to need more than 10 iterations to correct like it has. Not sure, I'll do more digging.