Hi all, I'am trying to apply gmres implicit scheme to my LES code. But I found it converged slowly at each step in the case of Decaying Isotropic Turbulence(DIT) simulation. I used 7 points central difference scheme for spatial discretization. And a LU-SGS preconditioner was used for the gmres method. Does anyone have suggestion for the time integration scheme in LES ? It is not suitable to use implicit scheme in LES ? Thanks!

I am not familiar with GMRES, but there is no fundamental reason that you can not use an implicit scheme. The issue is accuracy versus computational time. Your code has discretization errors due to your grid size and your time step size. For an implicit scheme, you can take very large time steps. This combined with the fact that implicit schemes are generally 1st or 2nd order means your error in the time discretization could be significant. So taking large time steps may not be a good thing. If this is the case you might as well use a faster explicit method and save on cpu time. The only way to check to see if this is the case is to do a time step study (similar to a grid study). Halve the time step and recompute. If the solution is unchanged, you are good.

[harish_0689@]

Hey EagleGao, even i am trying to use implicit schemes for the euler equation and using the GMRES matrix solver. Even i found its converging slower than the explicit in terms of CPU time. And i was searching for the LU-SGS solver. Can u suggest me how and where can i get the LU-SGS solver? i am in a serious need of that.