SimScale CAE Forum

Simulation cancelled - abnormally long simulation time?

Hello everyone,

I am new to SimScale, and currently attempting to simulate multiphase flow (water and air) in a static mixer, which I will next build with my own hands! My hope is that SimScale will provide useful insights into the designing phase, (e.g. shape and number of mixing elements required to achieve good mixing without creating too much pressure drop).

So far I’ve managed to resolve my simulation issues by reading through the forum, which has been very helpful, but now I face the following question:

how can I make my simulation less computationally painful?

My previous run ran for 172 mins before being canceled, as the maximal allowed time had elapsed.
I already decreased the Delta t (to 0.001 sec), as advised.
I can always re-run with a smaller time step, but I feel like it shouldn’t take that long, and that there is something making it abnormally long.

I’d appreciate any comment or idea on how to make this less painful.

Cheers,
Aymar

https://www.simscale.com/workbench/?pid=1263498155602346615&mi=run%3A90%2Csimulation%3A56&mt=SIMULATION_RUN

@core_support - can you jump in here please?

Best,

Jousef

Hi!

Multiphase simulations are actually quite different from other analysis types. In a very short explanation, multiphase analysis are inherently transient, whereas for most other analysis types, you will have steady-state simulations.

Since the CFL condition is a timestep limitation for multiphase analysis, here are some good practices for multiphase:

  • A coarse and uniform hex-dominant mesh generally works best. Note that boundary layers are not really desired, as they will impose limits due to the CFL condition. If you really need boundary layers, go for 1 at most
  • Use a very small time step at the beginning of the simulation, and make sure that the timestep size is automatic
  • I noticed that you are using pressure inlets and outlets. If you are interested in a specific velocity, I’d use a velocity inlet instead. With pressure inlet, you might be inducing high velocities within the domain if you are not cautious

With these changes (especially the first pointer), you should achieve a more manageable runtime.

Cheers

1 Like