Have been trying to set up a thermosyphon or NCL. Couple of issues

In an open volume I can get the NCL working no problem. If I take the center out of the volume, basically where v = 0 at the center, I get some type of inverse flow which is very odd.

Setting up closed loop with the settings from the transient heat tutorial, but with the flow solved for velocity vector. To do that I used the setting for the volume from the pipe tutorial. I keep getting an error message in the log:

RANS Flow failed
Solver singularity: smoothSolver: Solving for epsilon, Initial residual = nan, Final residual = nan, No Iterations 100

Any idea what is happening?

Also is there any way to reset the simulation once you have run a time on it? E.g. to clear the sim back to the set-up point.

And for a last question I tried to set up the NCL under compressible and the iteration did not converge. NCL's ostensibly should have a density component to the sim based on the thermal input. Is there any way to do an NCL as compressible or density variable in this software.


Transient Incompressible Heat Transfer

I have no experience with Thermosiphon applications, but I think I understand what you are trying to achieve.

The incompressible heat transfer solver in Caedium assumes the Boussinesq approximation holds. So variations in fluid density are assumed to be negligible and buoyancy is driven by gravity.

Are you trying to solve steady-state or unsteady? Natural convection can often require unsteady (transient) simulations. Matching the time step to your transient simulation is crucial, so try making it successively smaller to avoid singularities.

Another thing to check is that you have gravity in the correct direction for your geometry.

It's also worth checking your mesh according to "Mesh Diagnostics."

To reset a simulation click the Beginning button. Also make sure you have turned archiving off.

The compressible solvers in Caedium assume relative large variations in density for Mach numbers greater than 0.3, so for your application I don't think they are applicable.



Using transient and incompressible for simulation. NCL's take some time to develop. Finally got rid of most of the singularity problems by simplifying the wall geometry.

That said am getting some very weird results from the sim. Started with the target conditions for the NCL which is about 100C temp variation between the hot and cold columnar sections. First the flow starts off in the opposite direction flowing toward the heated column then it reverses instantaneously and then drops to some type of sinusoidal state where it accelerates, decelerates, reverses direction and repeats. Also velocities are very small. According to an incompressible PGF calc they should be pretty high, big pressure differential short distance.

Thought it might be that the columns did not have enough delta so I increased them to 1000C delta which should produce near sonic conditions (based on actual NCL's) in an NCL or thermosiphon but shows the same behavior as previous sim and only comes up to a max 1 ms.

Essentially this thing should function like a closed boundary land-sea breeze but is not. Big difference with NCL's is most of the calcs are based on density. That is the only thing I can figure for why the sim results are the way they are. Ran it as compressible and it does not even develop a flow.

Would it be better to assign a heat flux to the walls or possibly an embedded outlet or possibly try to assign different density regions? Is that possbile?


Natural Convection Loop (NCL) Suggestions

I'd stick with the incompressible heat transfer solver for your simulations.

If you haven't already I'd suggest trying to increase the mesh density (Accuracy:Custom->Resolution = higher) in the regions where you expect the highest temperature gradients - note this will also require that you decrease your time step if you are running transient.

Another option is to try a higher order difference scheme, i.e., Substance:Air->Solver->Schemes->Divergence = Linear Upwind

If shouldn't matter whether you assign a temperature or heat flux. There's currently no way to assign different densities for the same flow volume. You could try initializing with a small non-zero velocity to nudge the initial startup in the right direction.