Convergence time???
 

All,

I'm running a simple airfoil with a plenum inside for flow control at the trailing edge. When I run the case it takes about 1 and 1/2 days or so to get the solution to converge. I'm new to this flow solver so I was wondering if this sounded right to any of you who have more experience. I feel like its taking way to long for this case.

For completeness here are some of the solver settings and details of the grid: coupled implicit (have tried many cases with CFl # <= 5), node based, spalart-allmaras, 256405 - elements in grid

Any comments or suggestions are welcome...

Thanks, greg

Re: Convergence time???
 

There is really not a great deal of help that anyone can provide on this subject. There are so many variables that can affect the time a soultion takes to converge, apart from the ones you have mentioned. e.g. (in no real order) If the energy equation is on, if there is a reaction occuring, if you are solving for UDS's of UDF's, if you are using 1st/2nd order upwind, what sort of system you are using, what the processor/memory is, how many other users are occupying the CPU, what your convergence criterion is.

I have run simulations that have taken 5 mins and 30+ days. It takes as long as it takes with the resources you have avaliable.

Re: Convergence time???
 

HEHE!aaaaaaa:((

My simulation of unsteady cavityting flow in 3D is now running for 565 h and 17 min:(

Have fun:) MATEUS

Re: Convergence time???
 

I'm running Linux with 2 - 3.06Ghz Xeon processors, and 1 gig of ram. Currently I am the only one using this machine. I usually set convergence criteria to 10e-6, I've tried higher convergence criteria but the Cl, Cd, and Cm's don't converge.

Re: Convergence time???
 

Usually it is best to solve using first order upwind first. Then use that intial solution and to solve for second order upwind.

Try play with your under relaxation factors too. It may help your solution time.

Another method to get better convergence is to adapt your grid.

Hope this helps. Have fun!

Re: Convergence time???
 

Thanks for the suggestions...I've played around with the courant number and that certainly helps but not much. I have not been running first order so I'll give that a shot. I think that the grid adaptation is the key to success for my case. I'll have to try and figure out how that works.

Thanks again...

Re: Convergence time???
 

Normally I do a countour plot and find regions of high gradients. You can also look at the residuals and mass imbalance. I then find the minimum gradient and use that value to refine my grid.