[Sanoja]

I have included SRF model to the chtMultiRegionFoam solver, to include the effect of rotating reference frame, as I want to simulate a heat exchanger that rotates.

I have included the SRF model to the solver (named the solver as myHeatExFoamomega) and then changed all the files in the Fluid region to use relative velocity (replaced U with Urel).

The solver compiled without errors, and I am trying to simulate a counter current cylindrical heat exchanger which rotates around its centre axis (Z axis).

I have 3 regions in the heat exchanger, 2 fluid and 1 solid. I have attached the changeDictionaryDict files used and the fvSolution files.

I cannot get this simulation to run, as I get extremely small time steps. If I use fixed time step, for example, 0.001, it runs fast for upto about 0.29 and then crashes, with unphysicaly low temperature in one of the fluid regions.

I really appreciate if some one can help me with this issue. Some tips to get the simulations running.

I have attached the fluid folder used in the new solver and the .c file of the solver as well.

[Sanoja]

I have changed the boundary conditions and started running in serial, and then the simulation works for low rpm, like, 1000 rpm. But I need the rpm to be in the range or 10000.
So, I believe I will have to initialize a simulation with the results from a simulation with low rpm.

[blais.bruno]

Originally Posted by Sanoja

I have changed the boundary conditions and started running in serial, and then the simulation works for low rpm, like, 1000 rpm. But I need the rpm to be in the range or 10000.
So, I believe I will have to initialize a simulation with the results from a simulation with low rpm.

Are you running a steady or an unsteady simulation?

What are your initial conditions? Be careful that Urel=0 can be highly unstable as an initial condition. The best idea is generally to start with Urel=omega*R...

In the context of mixing this has hepled me greatly.

Otherwise you might want to validate with a simpler example (like a couette flow) first before testing something bigger.

[Sanoja]

Thank you for your advise Bruno.

I'm trying to run an unsteady case. And have used the absolute velocity as the inlet value (axial velocity) and then used the same value in the internal fields and given as Urel. This worked for 100 rpm.

Then I tried to use the out put values from the last time step of that simulation into an other case with 10000 rpm.

Simulations are really slow as I am running in serial mode (faced difficulties with the parallel running). And I am waiting to see whether this is going to converge or not. If not, then I will try one with the way you have suggested.

[Sanoja]

My simulations with 10000 rpm has failed (where I used the results from a simulation with 100 rpm), was getting way too high temperatures as a results of too big velocity components (in some places. colse to the outer walls, due to high swirels).

So, I was thinking of using Urel as suggested by blais.bruno !!

But now, I am wondering how to split the calculated Urel value (using Urel = r_mean*omega) into the x, y components. Any ideas ???

[blais.bruno]

Originally Posted by Sanoja

My simulations with 10000 rpm has failed (where I used the results from a simulation with 100 rpm), was getting way too high temperatures as a results of too big velocity components (in some places. colse to the outer walls, due to high swirels).

So, I was thinking of using Urel as suggested by blais.bruno !!

But now, I am wondering how to split the calculated Urel value (using Urel = r_mean*omega) into the x, y components. Any ideas ???

You really need to impose the right initial condition. I believe that's the issue you have right now.

in SRF type of models, you have two velocities. Uabs is in the laboratory frame of reference and Urel is in the rotating frame of reference. Therefore, Uabs=0 as initial condition is not equal to Urel=0 as initial condition. If you start with Urel=0 as initial condition, your simulations will most likely diverge (or least be very unstable) for high RPM values.

I hope that is sufficiently clear

[Sanoja]

Thank you blais.bruno

I don't use U_rel = 0 as initial values.

Now I have decided to increase rpm step by step. Like using results from 100 rpm case in 200 rpm simulation, an so on. Seems to work But time consuming