Different Core number produce different calculation results
 

Hi all,

I have met some situations recently. I set up an identical case and run in two different computers (one is 16 cores with 32 logic cores, the other one is 27 cores with 54 logic cores). The simulation showed an identical pressure trace before combustion and then a totall different combustion process afterwards. The one with less core predicted a faster combustion apparently. The combustion model is SAGE with RANS turbulence model, and following Converge recommd settings from example case.


My question is: does the core number affect calculation results? Is there any guidelines about setting up converge core number?

Thank you!

Hi Shanghua.

Yes, the core number can affect the simulation results due to numerical perturbation, especially when the combustion system is unstable. For example, a computational diesel engine will not exhibit these oscillations whereas the GDI type simulation will show cycle-to-cycle oscillations. You can also see this cycle-to-cycle oscillation in experimental engine data. More detailed information can be found in this paper:

https://asmedigitalcollection.asme.o...T06A010/245357

As for your second question, there is no guideline about setting up core number.

Best,
Chengqi