economical convergence

  • Last Post 07 June 2019
akin3rd posted this 07 June 2019

I am currently running a 2-d simulation on an airfoil at re=6m. 

The simulation uses a mesh of about 100k cells and has y+<1 

i have set my residuals to e-6 but it takes so long to converge (usually over 20k iterations). is there a residual value low enough to guarantee accurate results but high enough to reduce time to final convergence ? 


Order By: Standard | Newest | Votes
rwoolhou posted this 07 June 2019

The defaults (1e-3 for flow, 1e-6 for energy) and monitors are usually good enough for most problems. Why did you choose 1e-6 ? 

akin3rd posted this 07 June 2019

I read in some tutorial that e-3 is too high for proper convergence and e-6 is better

 I am using the k-omega SST turbulent model and the Mach number is about 0.28 so the flow is assumed incompressible and subsonic.

I had to reduce the under-relaxation factors of pressure momentum, kinetic energy and dissipation from their default values to 0.2, 0.5,0.5  and 0.5 respectively.

rwoolhou posted this 07 June 2019

That sounds like it ought to work.  Check the mesh quality & resolution. Are you getting flow separation?