CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Timestep (http://www.cfd-online.com/Forums/cfx/20213-timestep.html)

mefpz March 4, 2004 05:19

Timestep
 
I'm a new comer on cfd. I am always confused how to choose Timestep. What is the criteria?

test March 4, 2004 10:41

Re: Timestep
 
Which version of CFx are you talking about??

for cfx5.6...

To start with go through the chapter on Advice on flow modelling in the help manual. This will clear lot of your other doubts..

There are three ways you can define the tiemstep: Physical time step: Even though you are solving for a steady state solution, you are marching ahead in time. A physical timescale would typically be 1/3* residence time. The residence time can be rougly be calulated as maximum length of the domain/average velocity. This is true for simple flows. For more complicated physics the timestep has be be determined by case to case basis.

If the flow is density driven you will have to restrict the timestep used to the bouyancy timescale(given in the chapter mentioned above)

Local timescale factor: Suitable for high speed flows, where you have largely varying mesh sizes. COnvergence is good but final result is a bit off. 5 is the most typical value.

An idea of the average residence time can be had from the average scale information printed in the out file just before the start of the iteration. Look out for Avection timescale

Hope this helps

mefpz March 4, 2004 20:47

Re: Timestep,responce to Test
 
Thanks for your advice. It's what I have been sought. However I have another question: Timestep in transcient problem and Timescale in steady problem. Does they have the same phsical attributes, that is, a timescale is one that a timestep used in the steady problem?

mefpz March 4, 2004 20:52

Discuss with test or anybody else!
 
Please test read the preceding message! Thanks

mefpz March 4, 2004 22:37

Re: responce to Test
 
For example: I will solve a transcient problem. First, In order to obtain an initial value file, I will finish a steady simulation using a timescale say 3s, Then continue to do the transcient problem using the timestep, how to choose the timestep? Shoud the timestep less than timescale 3s?

test March 5, 2004 00:39

Re: Timestep,responce to Test
 
Hi,

As mentioned earlier, even for a transient problem we progress ahead in time. The advantage in a steady state case is that you can use relatively large timesteps as compared to those in the transient runs. For eg. for a simple flows, transient timescale which can be started is 1/10 of the residence time.

For a transient case you get to a steayd state solution for each timestep. Start with a small timestep, once the initial transience is over (say 40-50 iterations) the solution would slowly reduce its sensitivity to timestep. SO you can sloly start increasing the timestep.Ofcourse there would eb some max limit for this increase. this will depend upon problem to problem

for reactive flows very small timesteps of 1e-3 to 1e-4 should be used. The timestep to be used will depend upon the physics.

You neednot always provide the initial guess files for a transient case with a steay state result.You will have to give the initial guess. For eg if there is no flow at time t=0, give the initial condition as velocity=0

For a steady state problem the inital guess provided is such that it is more closer to the steayd state value. It is not the same for a transient case problem


mefpz March 5, 2004 04:34

You are great!See you later!
 


All times are GMT -4. The time now is 15:27.