CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)

 naveen_010 February 2, 2011 12:40

Dear users

I understand both steady and unsteady solvers will involve calculation of time steps. I usually calculate time step for unsteady case by CFL criteria. For steady case there will be psedo time step, how does the solver calculates this psueotime step .

Thanks & Regards
vnnnn.

 julien.decharentenay February 3, 2011 02:31

Hi,

It depends on the software you are talking about. Implementation of steady-state solver usually uses either of the following:

- An under-relaxation term (in place of a transient term).
Note 1: some solver do solve for the steady-state equation (without transient or under-relaxation terms), but usually needs to get close to the solution first (using an under-relaxation or transient term).
Note 2: the formulation of the under-relaxation term is so close to the transient one, that it is my belief that you may even be able to work it backward to an actual time. However, it is standard practice to employ different under-relaxation factors for different variables.

- a local time-step based on a local CFL. In this approach an individual time-step is computed for each individual control volume/cell and used during the integration. Consequently you loose time consistency, but convergence is faster.

Regardless of the approach the software employs, the "converged" solution should have its transient term/under-relaxation term going to zero.

Sincerely, Julien

 naveen_010 February 3, 2011 14:18

timesteps

Dear julien

Thanks for the information.
I use star-cd for my runs, do you have any idea how does star-cd calculates the time step in steady state. I run my cases by specifying the number of iterations.

My belief is steady or unsteady solver, the outer iterations are time step and inner one are iterations.

Thanks & Regards
Vnnnn

 julien.decharentenay February 3, 2011 20:49

For steady-state, STAR-CD uses under-relaxation factors. There is no time-step in the integration of the solver equations. Note that there may be time-steps associated with specific models, lagrangian particle tracking for example.

The correct terminology is as follows: