search

gdtk-uq / gdtk

The Gas Dynamics Toolkit (GDTk) is a set of software tools for simulating high speed fluid flow, maintained at The University of Queensland and the University of Southern Queensland, Australia.
https://gdtk.uqcloud.net/
Other
62 stars 17 forks source link

Issue: Finest grid turbulent flow simulations take more computational time with e4-nk-dist solver #56

Closed blackysak closed 6 months ago

blackysak commented 6 months ago

Hi, I am trying to run Shock-wave boundary layer interaction simulation using e4-nk-dist solver. My computational mesh is hybrid (layer for quad cell near the wall and triangular cells rest of the domain) and consists of 1.3 lakhs cells in total. I am running the simulations in parallel with 24 processors. I am running simulations for Mach number of 5 with k-omega turbulence model. The smallest cell size is 1e-6. Due this smallest cell size, the pseudo time step is coming out around 1e-8. Because of this smallest pseudo time step it is takes large amount of computational time for one simulation. Is this the case for any simulations with finest grids? Is there any way to speed up the simulations apart from the parallelization? Kindly let me know. PS: I am attaching the pictures of computational grid and terminal output.

1 2 3

Thanks. Sakthi

uqngibbo commented 6 months ago

Hi Sakthi,

From what I can see, it looks like this simulation probably converged a long time ago, but didn't stop because the stopping criterion was too tight.

The first entry in the "relative" column is the global relative residual. Yours is down to 7.631005e-11, a drop of 10 orders of magnitude from where it started. This is a very converged flowfield.

For next time, I would set the stopping tolerance a little higher, anything less than 1e-8 is usually fine. You can do this by changing the right field in the SteadyStateSolver options:

stop_on_relative_global_residual = 1.0e-8,