CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   ILU for Navier stokes problems (https://www.cfd-online.com/Forums/main/11886-ilu-navier-stokes-problems.html)

 Raju July 27, 2006 10:08

ILU for Navier stokes problems

Hello, I am trying to use ILU preconditioner for Navier stokes problems. None of the standard ILU preconditioners in the SPARSKIT package is working due to the presence of zero's in the pressure (continuity) equation. Does any one has any suggestion about how to implement ILU for Navier stokes problems.

 Tom July 27, 2006 10:23

Re: ILU for Navier stokes problems

You could try modifying the continuity equation to include a pressure term; e.g. replace div(u) = 0 by something like

k.[p(n) - p(n-1)] + div(u) = 0

where p(n) is the value of p(n) is p at the n'th iteration and k is a constant.

 saudagar July 27, 2006 16:01

Re: ILU for Navier stokes problems...useless reply

This reminds me of the famous song...

Ye ILU ILU kya hai ? Ye ILU ILU (what is this ILU ILU ?)

ILU ka matlab I love you ! (ILU means I love you...)

I hope having a little fun on this forum is not that bad...

 KK.Khan July 28, 2006 14:10

Re: ILU for Navier stokes problems...useless reply

Try to Use Artificial Compressibility Method of

Chorin which include pressure in the continuity equation.

cheer

KK.Khan

 ztdep July 28, 2006 22:01

Re: ILU for Navier stokes problems

Hi: I also want to apply the SPARSKIT to the N-S equations, I think we can use a block implicit method. what is you idea

 rt July 29, 2006 14:15

Re: ILU for Navier stokes problems

precence of zeros in diagonal entries is not desirable and is related to poor physical or numerical modeling, but there are some cure (althouth i recommend modifiung modeling, projection method instead probably penalty method), you can use pivoting to change position of zeros to off-diagonals or you can use BILUM package, it is opensource due to Y. Saad and designed for solution of such ill-pose systems, it is not only rubost but also enjoys properties of algebraic multigrids (scalability and fast convergence) free of AMG limitations such as SPD condition.

 All times are GMT -4. The time now is 12:14.