'box with wall' simulation project by hermanowicz


#1

I created a new simulation project called 'box with wall':

various CAD inputs


More of my public projects can be found here.


#2

Hello @hermanowicz, do you need any help with fixing the simulation and make it run successfully? Cheers!


#3

Hi,
In my project “box with wall” I want to simulate passive scalar transport. I need a transient simulation of a tracer to get the residence time distribution. For this, I need records of at least 20-30 timesteps with the simulation time from 0 to 3x residence time. For this model, the residence time Vol/Flow is about 6 sec so I set the end time value of 20 sec with 2 sec time step. I chose adjustable time step with writing results every 2 sec of “runtime”. I presume “runtime” is the time in the model so I will get (20 sec/2 sec +1) = 11 records. Is this correct?

I am confused by your terminology of “clocktime”, “runtime”, “adjustable runtime”. In the previous runs I tried smaller timestep to get more records but the run terminated due to excessive memory.

This last run “Tran 20 s 2s step adj write 2s” seems to require small memory but quite long CPU time. I set the limit to 30000 sec and only got the first step simulated (i.e., 2 sec) even with a very coarse grid. By this token, reaching 20 sec simulation time will require 8 times more CPU.

Can I do something to run this better?

Also, I need to integrate the value of the scalar over a specific cross-section of the box at different times (assuming I get more than one timestep simulated). This represents an average concentration of the tracer. Is there a procedure to do this easily?

Thanks a lot for your help. Your whole team has been really helpful!!


#4

@hermanowicz - nice setup! Yes, you’re correct with the runtime definition (agreed btw that the wording could be clearer - what would be good names for you?). I just briefly checked on your sim setup - with such low velocities and below 0.5 million cells, that should indeed be faster. I’ll check it tomorrow in more detail but I’m sure there is something that can significantly speed up the the sim. Also @gholami or @Ali_Arafat might be able to help.


#5

Hi @hermanowicz @dheiny,

In a Transient analysis the “Courant Number or CFL” determines the actual ‘Time step’ being used in the simulation and defines a stability criterion. In brief it is based on the flow velocity and the smallest cell size in the mesh (see link for details).

So, for a given flow condition and mesh, a certain fixed value of ‘CFL’ will give a certain fixed value of ‘Time step’ .

Now, in your case as you are using “Automatic Time Stepping” , (which is highly recommended) with a CFL of 0.5,

The actual Time Step used is Not ‘2 sec’ as you have specified. If you see the logs the value used is about ;

Courant Number mean: 0.00311384809156 max: 0.499998505841
deltaT = 6.27974789015e-05

1- Now, if you do Not want to change the flow conditions or the mesh, then the only way to make it faster ( i.e use larger Time steps) would be to increase the CFL to 0.9. You may try a value close to or higher than 1.0 but it would increase then the chance of divergence.

2- If you use a coarser mesh, than for the same CFL , the Time Step can be larger also.

I hope this would help.

Best,
Ali