Transient Simulation Taking Very Long

I am currently trying to simulate airflow around an airfoil with high angle of attack. Since the residuals and coefficients did not converge for a steady-state simulation, I’m currently trying to do a transient simulation.

However, its taking extremely long, even when I increased the timestep. I’m also worried about it consuming too much core hours. Is this normal for a transient simulation, or is there a way to reduce the computing time by adjusting the settings?

Below is the link to the running simulation.

Thank you!

Hi @Semix, thanks for posting on the forum.

Could you share your project with support?

It’s common that transient studies take longer to run and therefore consume a lot of core hours. They are not trivial and you need to achieve a high mesh quality/uniformity + a good balance between timestep size and simulation efficiency. These 2 resources should be very helpful:

One quick tip to consume less core hours however is to reduce the number of processors your simulation is running on. A good approach is to use 1 core per 500k cells so that the machine does not run out of memory (e.g. if you have 2M cells, 2M/500k = 4 processing cores).

Cheers
Igor

Hi Igor,

I’ve shared the project with you. My simulation takes around 200 cores for 8%, but is that typical?

Also, thank you for the links!

Hi @Semix,

Yes, it’s fairly typical. If you reduce the number of cores manually as I’ve mentioned above, you should see a reduction in the amount of core hours you’d consume.

Keep in mind a velocity of 84m/s is on the verge of the Incompressible solver’s capability and you’d need a very small timestep to compensate it so that the Courant number is controlled (see resource I’ve sent above). In your case, since the Maximal courant number is set to 0.7, assuming a maximum speed of 84m/s and I’ve measured the smallest element to have an edge lenght of ~0.002m:

\displaystyle C = u\frac{\Delta t}{\Delta x} \tag{1}

\displaystyle 0.7 =84 \frac{\Delta t}{0.002} \tag{2}

\displaystyle \therefore \Delta t \approx 1.7\times 10^{-5}s\tag{3}

Which would take very long to run given your end time of 2 seconds.

Cheers