Very slow convergence

AlbertoPi · June 4, 2017, 07:44

Hi everyone,
I have a convergence problem. I'm running a simulation of the naca 4412 at 13.87 deg of AOA and Mach 0.09 from the following link:https://turbmodels.larc.nasa.gov/naca4412sep_val.html

But my residuals, especially the Rho·Energy, converge very slowly. And I can not find the reason.
I've tried changing the CFL, linear solver, multigrid, CFL_REDUCTION_TURB, RELAXATION_FACTOR_TURB...
But I have not been able to accelerate the convergence. Can anyone help me?

I have attached the configuration file, evolution of the residues (restarted after more than 200,000 iterations), and a screenshot of Paraview with the mapping of energy residues on the mesh.

Best regards,
Alberto

omdemircan · June 4, 2017, 21:06

Looks like you are trying to solve in 2nd order discretisation. You could try convergence with 1st order, then change it back to 2nd order if you want.
You could also increase the initial CFL to higher values and switch off CFL adaptation parameter. This could provide faster convergence for a trade of possible instability of the solution.
By the way, how did you plot the residuals? Did you use a GNU plot in real-time?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
slow convergence in chtMultiRegionSimpleFoam	sven82	OpenFOAM Running, Solving & CFD	4	October 2, 2014 11:15
Force can not converge	colopolo	CFX	13	October 4, 2011 23:03
Convergence of CFX field in FSI analysis	nasdak	CFX	2	June 29, 2009 02:17
Slow convergence for Boundary Layer flow	Biga	Main CFD Forum	2	November 18, 2004 17:51
Slow convergence	Jesper	CFX	1	July 7, 2004 17:59

June 4, 2017, 21:06		#2
omdemircan New Member Osman Mirza Demircan Join Date: May 2017 Location: Ankara, Türkiye Posts: 29 Rep Power: 9	Looks like you are trying to solve in 2nd order discretisation. You could try convergence with 1st order, then change it back to 2nd order if you want. You could also increase the initial CFL to higher values and switch off CFL adaptation parameter. This could provide faster convergence for a trade of possible instability of the solution. By the way, how did you plot the residuals? Did you use a GNU plot in real-time?