Unable to use Crank NicholsonTime differentiation

Using SimScale Fluid Flow / CFD
1

Im using pimplefoam and unable to find Crank NicholsonTime differentiation.

2

Hi Sebastian here,
you can find all information about our numeric schemes here:

please be aware that not all schemes are implemented.

Best regards Sebastian

3

Dear @SBlock , I was checking this out, and as @agoney mentioned, the Crank-Nicholson scheme is not implemented for the PIMPLE algorithm under incompressible flow.

Dear @agoney, I advise you to either use the backward scheme (if you need to attain 2nd order accuracy) or use an Euler scheme, download the case and run it in OpenFOAM.

If you need to understand how the backward scheme works, I have written a description in my paper: Paper, section 2.3. You can also check the reference posted there: “Moukalled, F.; Mangani, L.; Darwish, M. Temporal Discretization: The Transient Term. In The Finite Volume Method in Computational Fluid Dynamics; Springer: Berlin/Heidelberg, Germany, 2016; pp. 489–533.”.

Happy CFDing,

Jairo

4

Hi @jairogut .

I am using Openfoam for my thesis work, i have already carried out my validation studies using crank Nicholson scheme. Due to my system limitation i have opted for Simscale.

@jairogut and @SBlock any idea regarding the scalability of Simscale.

I have a i7 machine and use 6 cores to run my cases on my machine. But to my surprise when i was running the same case in Simscale it took a lot longer for the solution to finish.

I have uploaded the same exact mesh numerical settings except for crank Nicholson scheme.

The case took 53 mins to run on my laptop and it took 56 mins to run on Simscale with 16 cores .

Any idea why this is happening?

5

I think I may have a very good idea of what is happening (although I am not a SimScale employee).

  1. SimScale cores have lower specifications.
  2. SimScale may use some “in-house” tweaks or by-default configurations to make the cases more stable (as the average user does not know all he is doing). These numerical configurations may be more robust but more expensive. For steady-state simulations (SIMPLEC), I am positive they have automatic overrelaxation control, which is not the case for OpenFOAM.

By the way, there should not be a noticeable difference in CPU time between the Crank Nicholson and backward schemes. But again, solution-wise, for the immense majority of cases and configurations, there should not be a significant difference between C-N and backward schemes. Try to use a more efficient, element distribution-wise grid to lower your cost.

Happy CFDing.

Jairo