CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Density- vs. Pressure-Based Solvers (http://www.cfd-online.com/Forums/main/120387-density-vs-pressure-based-solvers.html)

Dwayne July 5, 2013 13:05

Density- vs. Pressure-Based Solvers
 
Hey Guys,

I know the difference between density-based and pressure-based flow solvers and I know their extensions to regimes of arbitrary Machnumbers. So preconditioning and density-velocity-pressure-coupling.

My question is how I decide wich solver I should use in what situation. What are the advantages and disadvantages of both? I only found out that density-based solvers are better in solving shocks.

Thanks for help!

immortality July 7, 2013 04:07

there may be not a clear boundary between them.
Density based solvers are originated for high velocity flows,although pressure based have improved a lot for high velocities too but still density based's are better in mach around ond and higher,i think it depends on the problem if others add something is nice.
I'd like to know preconditioning if you tell me a bit.

Dwayne July 7, 2013 06:51

Quote:

Originally Posted by immortality (Post 438206)
there may be not a clear boundary between them.
Density based solvers are originated for high velocity flows,although pressure based have improved a lot for high velocities too but still density based's are better in mach around ond and higher,i think it depends on the problem if others add something is nice.
I'd like to know preconditioning if you tell me a bit.

Thanks, in compressible flows information is carried with flow velocity and also with speed of sound. So you have to consider both velocities is your stability consideration. For explicit methods the CFL-Condition must be fulfilled wich causes very small timesteps because of the high speed of sound.

A system is than called bad conditioned. In other words: A system is bad conditioned because of the large disparity of the eigenvalues of the jacobian. By multiplying the timederivation with an preconditioning-matrix you can force the eigenvalues to be of similiar order. This also solves the accuracy problem of density based solvers at low mach numbers.

immortality July 7, 2013 08:13

thanks,may multiplying a matrix to time derivation cause an inaccuracy in each time step in real unsteady problems?
and what nSweep is better in your opinion?

Dwayne July 7, 2013 08:53

That's right the preconditioned equations have only the stationary solution in common with the orginal equations. But the time accuracy can be restored by dual time stepping. I just found out that this is really time consuming so this is an disadvantage of preconditioned density-based solvers. Thanks for the hint.

What do you mean with nsweeps? Density-based solvers don't need iterative methods I thought or am I wrong?


All times are GMT -4. The time now is 16:24.