CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   time step??????? (https://www.cfd-online.com/Forums/main/11368-time-step.html)

vijesh joshi May 6, 2006 03:48

time step???????
 
hi friends. What is this time step? and why it is required? what is its role in a steady state problem? these are my questions in this message. Please find the answers and help me out. thank you.

ganesh May 6, 2006 06:18

Re: time step???????
 
Dear Vijesh,

The governing equations of fluid flow are cast as an unsteady state problem, so that they are hyperbolic for all flow regimes. Note that there is a distiction between steady state subsonic and supersonic problems as being elliptic and hyperbolic, which would necessiate use of different solution procedures for these regimes, and would create complications in the transonic regime. Therefore, for solution to steady--state problems, these unsteady problems are marched in time, with a suitable timestep, based on stability considerations, until some residue (RMS value of error in density/energy) falls below a specified value epsilon, which is of the order of 1e-6.

Hope this helps.

Regards,

Ganesh


vijesh joshi May 6, 2006 06:49

Re: time step???????
 
Thanks a lot...Mr.Ganesh. If I get some doubt on the same topic, I will post the message again.

Ben May 6, 2006 06:54

Re: time step???????
 
I would suggest that if you dont know what a time step is then you are probably in the wrong business!

Bak_Flow May 6, 2006 15:49

Re: time step???????
 
Hi Vijesh,

it sounds like you need a good intro text to numerical methods and CFD.

Check out the books page of CFD-online. A couple of useful suggestions are:

Versteeg and Malalasekera: An introduction to Computational Fluid Dynamics. In particuar chapter 8 covers the solution of transient probmes and using a transient to under-relax a solution to steady state.

Ferziger and Peric: Computational Methods for Fluid Dynamics. This is a bit broader and deeper than Versteeg and Malalasekera

Regards,

Bak_Flow

taw May 7, 2006 03:46

Re: time step???????
 
This is for Ganesh,

From your response I think u have a good insight in choosing time step. But my problem is as follows.

I am studying flow instability, using CFD tools. The main idea is to generate the TS waves and study how they grow downstream. For certain time steps I choose I can track this waves, like delT= 1e-5, but when I want check grid or time independency and run it on delT=1e-6, 1e-7, these waves disappears. I expect I would get more resolved and sharper waves. Is there anything like this happen or what could the explanation be? ta thanks in a dvance

diaw May 7, 2006 04:08

Re: time step???????
 
Taw,

Can you be more specific about your flow regime & how you are currently determining a suitable time-step?

diaw... (Des Aubery)


taw May 7, 2006 21:01

Re: time step???????
 
Diaw,

Consider a simple plate flow, I used Re = 400, the typical problem I solved is that explained by Fasel, the paper I am referring could give of what I am doing. He used a full Navier-Stokes solver to study plate flow instability. Hope this can give you a more idea what I am doing.

Rist, U. and Fasel, H., Direct Navier-Stokes Simulations of Transition and Turbulence in a Flat Plate Boundary Layer, J. Fluid Mech.298:211-248, 1995

This much, Thanks


diaw May 7, 2006 21:22

Re: time step???????
 
Hi Taw,

Thanks for your reply. I'll take a look at your paper reference to get more idea of Fasel's approach.

In my own research, I have found it most useful to concentrate on locking the Courant Number into a tight band, which then sets the required time-step, relative to mesh & velocity. You are looking to capture waves, & courant basically locks together the first two terms of the wave equation. It has a lot to do with model information flow-rate versus physical flow-rates.

Try this simple rule & tell me what you see... I'm sure that it will get you on the right path. You can then vary your 'timing' to suit the fact that viscous stresses will be very active within the boundary layer itself.

diaw... (Des Aubery)

taw May 8, 2006 00:14

Re: time step???????
 
Hi Diaw,

Thankyou very very much, you really indeed directed me to the right path, but I have to workout how I could capture the courant, convective and diffusive scales from my simulation, and I will get back to you. great

with many thanks, taw


All times are GMT -4. The time now is 11:56.