Learning Forum

[jesseahlquist]

Setup: Following tutorial at

Geometry: Symmetric 10cm radius cylinder w/ hieght = 20cm; inlet = 4cm; outlet = 2cm;

Mesh Quality:

Minimum Orthogonal Quality: 6.2682e-01; Maximum Aspect Ration = 1.07262e+01

Solution Setup:

- transient // gravity (z = -9. // Vol. of Fluid -- Water and Air // Energy ON

- Inlet: Vol Frac. Water = 1 / Velocity Mag. = 0.01 m/s; Outlet Backflo Vol. Frac. Water = 0

- Fluid Vol. = mixture // Solution Method -- See Picture // Solution Controls = Default

- Initialization: Standard from Inlet // Adapted Region = See Picture Included

- Patched Fluid Zone volume fraction of water = 0 // autosave every 5 time steps

- time step = 0.01s; number of time steps = 5000; max iterations/time step = 10;

Questions:

(1): After running the simulation, I viewed the animation of the volume fraction of water and found that the inlet stopped supplying water almost immediately (<<1sec). Is there a common reason, or reason based on the setup for why this would have happened?

(2): Based on the residual plot below, I was having difficulty judging the convergence of the solution. For a transient solution, is it expected to have oscillating residuals? Although most of the residual values "stabilized" around 10^-3 i.e. roughly stable I wasn't convinced. The simulation had maximum iterations/time step = 10; however, at the end of the simulation convergence was occurring with just two iterations, would this be justification for convergence?

(3): In terms of simulation speed, this took > 8hrs the other day which will not work going forward. I understand you can use the parallel computing option to speed up the process, but I'm wondering what the recommended processes assigned are to optimize speed, but also ensure limited wear and tear on my everyday computer.

[jesseahlquist]

[jesseahlquist]

[jesseahlquist]

[Rob]

Just a comment, the images you've shown are for a steady calculation. 

In transient you will expect to see a saw tooth residual profile as the solver converges each time step, updates time (residual spikes) and then converges for as long as you give it.

If the water didn't come in through the inlet, check you've got the phases defined correctly, and have the correct phase assigned to the inlet as it's easy to get that wrong (everyone does it at least once). 

Parallel will speed up the simulation, but transient CFD calculations do take time, and 8 hours isn't actually that bad. Only tutorials typically run in tens of minutes, everything else will take a while longer. 

 

[jesseahlquist]

 Thanks for the response rwoolhou!

Which setting would be turning it into a steady calculation? The SIMPLE scheme, higher term relaxation, or First Order Upwind?

I thought that could possibly be the case. A further question: Is there important significance to the rogue spike of the energy residual?

 

 

 

 

 

 

This boundary phase setup seems correct to me...do you see what the issue could be?

Okay good too know. Patience shall become my virtue.

[Karthik Remella]

Hello,

Have you reduces the values of your residuals? Please lower your residual criteria and check. You should start to see the saw-tooth like curves that rwoolhou is talking about. This might improve your results.

You inlet velocity condition looks about right if your secondary phase is water. 

Please let us know your findings.

Thank you.

Best Regards,

Karthik

[Rob]

Transient is in the General section of the tree (at the top), I don't think any other settings can turn it off, so you may have missed it during the set up.  Unless you need to calculate temperature there's no need to solve energy, the spikes are probably a result of something else going on in the solution.