how to solve a steady problem explicitly
Dear all, I am a little bit confused how I can solve a steady-state problem EXPLICITLY. do I have to write code for an unsteady situation and then give a long time?
thanks. Meda |
Re: how to solve a steady problem explicitly
Yes, until the time derivatives will be 0
|
Re: how to solve a steady problem explicitly
yes, you need to specify a time step, and it must satisfy the stability condition, then let it run until the difference between two consecutive time step is very small.
|
Re: how to solve a steady problem explicitly
In a steady state solution you can use local time stepping. Each cell progresses at it's own CFD criterium. As the solution is steady state the final converged outcome does not depend on the (local) timestep. This can significantly decrease your CPU time. In normal unsteady problems (explicit) your timestep is determined by the smallest cell with the largest velocity (via CFL criterium giving you a small overall timestep).
Bart |
Re: how to solve a steady problem explicitly
hi Bart.how can I define the local time stepping for each cell? doesnt it take more time to compute the criteria for each cell than to define a genaral criteria?
|
Re: how to solve a steady problem explicitly
I general criteria also requires you calculate a CFL timestep for every cell in between timesteps. I mean, how do you determine your timestep now and would it really make a big difference? However if your velocity scales and cell sizes in your domain vary only slightly then it may not make sense to use local timestepping.
|
Re: how to solve a steady problem explicitly
You can also employ local-preconditioning, which alters PDE such that all waves travel at an equal speed, thereby maximizing the minimum local time step (local time step is limited by the slowest wave speed of the PDE system). See this van Leer's paper --> Download at http://citeseerx.ist.psu.edu/viewdoc...10.1.1.43.8861
Good luck, bkjohn |
All times are GMT -4. The time now is 02:13. |