CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Time-accurate preconditioning for LES/DNS of turbulent flows (http://www.cfd-online.com/Forums/main/118652-time-accurate-preconditioning-les-dns-turbulent-flows.html)

francesco_capuano May 31, 2013 16:20

Time-accurate preconditioning for LES/DNS of turbulent flows
 
Dear all,

I have some some questions and doubts regarding the use of compressible codes for solving low-Mach (order 0.01-0.1) turbulent flows (using LES or DNS).

The main question is the following: is it really necessary to use time-accurate preconditioning techniques (e.g. dual-time stepping) for turbulent computations?

Some LES codes do that, but some others simply use an explicit time-integration with small time-steps. I don't quite understand whether the use of preconditioning is a matter of time-step restriction, accuracy, or both. Any experience on that?

Thanks in advance.

FMDenaro May 31, 2013 17:13

Quote:

Originally Posted by francesco_capuano (Post 431270)
Dear all,

I have some some questions and doubts regarding the use of compressible codes for solving low-Mach (order 0.01-0.1) turbulent flows (using LES or DNS).

The main question is the following: is it really necessary to use time-accurate preconditioning techniques (e.g. dual-time stepping) for turbulent computations?

Some LES codes do that, but some others simply use an explicit time-integration with small time-steps. I don't quite understand whether the use of preconditioning is a matter of time-step restriction, accuracy, or both. Any experience on that?

Thanks in advance.

Hi,
my opinion is that a pure compressible code could work at low Mach without preconditioning provided that the scheme is very accurate and the time step is constrained by the CFL condition based on the sound velocity.
Accuracy is important as also small errors in computing the density can result in high pressure perturbation.

cfdnewbie May 31, 2013 18:13

We use explicit time integration (with proper CFL bounds) all the time for Ma=0.1 computations, both LES and DNS and haven't encountered any problems.


All times are GMT -4. The time now is 16:38.