Hello,

Did you check the flow-through time required and actual simulation time? If your time step is quite small then you may have to run many more iterations to reach the stationary state. Alternatively you can increase/decrease the time stem following the CFL criterion, which also has an influence on the convergence.

Hope it helps.

Vilas

Abdul: thank you, I will try to first get a solution for the decoupled equations and then use that as an initial condition. I still find it a bit weird as my solution converged very easily when I used a y+ value which was large enough so that FLUENT used wall functions.

Vilas: I am running FLUENT in steady mode so I don't see how CFL criterion can be applied here, can you please correct me if I'm wrong?

Thank you both

I am also using the pressure based solver as my flow is at low Mach, should I use the density based one?

try increasing the courant number

Every turbulence model has some requirement of y+ so check that. If you wish to use a wall function then no need of fine mesh. Just target y+ ~25.

Before going for transient simulation first run for steady state convergence with 1st order discretization. After convergence raise it to second order and finally do the transient simulation where the choice of time step will influence the convergence rate.

To save time and to ensure that thing are moving in right direction start from simple settings and then add complexity like curvature correction, low Reynolds correction, order if discretization and so forth.

Hi, simply forget Spalart-Allmaras closure. It is fundamentally wrong like k-eps and k-kL other closures. The only one I can recommend is k-omega by Wilcox which is at least qualitatively correct in a sense that the right physics is at its basis (the original stems from Landau and Kolmogorov). I attach a paper.

Article Universal equations and constants of turbulent motion

Don't forget that when you lower the grid space in y+, the streamwise grid space should decrease also.