CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Steady and Unsteady nature (http://www.cfd-online.com/Forums/fluent/116475-steady-unsteady-nature.html)

Natali_Beginner April 19, 2013 19:25

Steady and Unsteady nature
 
Hello,
I am very beginner in CFD, sorry that I ask this question.
I am wondering how to compare a problem in steady and unsteady.
Because I think any problem that is steady can be solved unsteady as well.
However, I am running a case, but heat transfer coefficient in steady state is much much larger than that of unsteady. The problems were runned for some hours and when I see the transient solutions, it seems they reached to a steady state.
So, if the transient solution reached to steady state, why the results of steady and unsteady heat transfer coefficients are not the same?

Again, sorry if my question was childish but I really need your experts.
Natali

agustinvo April 19, 2013 19:46

Hi Natali;

I'm not an expert, but the difference between unsteady and steady is that you are not studying the flow at the same time: steady, the flow becomes "fix"; meanwhile unsteady, the flow changes along the time, until it arrives the time unsteady flow approachs the steady conditions.

For this reason, unsteady and steady heat transfer coefficients are not the same: you're not studying the same flow conditions.

I hope you understood my explanation.

Natali_Beginner April 19, 2013 20:16

Thanks for your answer,
But, after some times, the unsteady flow should be steady. Am I right?
So, why at that time the characteristics are not the same? like heat transfer coefficient?

oj.bulmer April 20, 2013 01:17

Not all the flows are steady "after some time". eg formation of vortices behind the bluff body. These vortices will keep changing their positions and hence this flow can never be steady.

Even unsteady flow can be categorized into, periodic and chaotic. Periodic flows have a pattern of repetition after certain period and thus has a certain dominant frequency. In chaotic flow, there is no single dominant frequency but there are set of frequencies. Then you need to use statistical techniques like FFT to figure out the relatively dominant frequencies from the flow characteristics.

OJ

Far April 20, 2013 10:03

I agree with comments of OJ.

Quote:

I am wondering how to compare a problem in steady and unsteady.
Compare time average flow parameters from unsteady simulation with steady flow solution and you will notice the difference (if there is any) .

Quote:

However, I am running a case, but heat transfer coefficient in steady state is much much larger than that of unsteady.
How did you observe this? did you compare the instantaneous value with steady case? If so you may be get large difference at some instant.

Quote:

So, if the transient solution reached to steady state, why the results of steady and unsteady heat transfer coefficients are not the same?
This may be the convergence problem with your steady or unsteady problem. All other settings are same? did you make the grid independence study?

Natali_Beginner April 20, 2013 17:15

Quote:

Originally Posted by Far (Post 421947)
I agree with comments of OJ.



Compare time average flow parameters from unsteady simulation with steady flow solution and you will notice the difference (if there is any) .



How did you observe this? did you compare the instantaneous value with steady case? If so you may be get large difference at some instant.


This may be the convergence problem with your steady or unsteady problem. All other settings are same? did you make the grid independence study?

Thanks for your answer,
Yes, all the conditions are the same. I did grid study as well. So, there is no place to consider difference between the results. I saw that the flow characteristics remain unchanged for a long time in a transient solution. Then I decided to call the flow steady. So, I expected to see the same results from the transient simulation at this time and the steady sate simulation. But, the results are much different.

blackmask April 20, 2013 23:38

Maybe your steady case has not converged yet. Have you tried to switch from unsteady to steady simulation with the case&data you obtained from long-time transient simulation?

Natali_Beginner April 21, 2013 00:23

Quote:

Originally Posted by blackmask (Post 422029)
Maybe your steady case has not converged yet. Have you tried to switch from unsteady to steady simulation with the case&data you obtained from long-time transient simulation?

I didn't try that but for sure the steady state solution is converged. I checked that by observing several flow characteristics, as well as checking the changes in mass flow rates, friction coeff, heat transfer coeff, etc.

Far April 21, 2013 02:21

can you show the convergence plots of both cases? Did you check the solution at lower time step ?

Natali_Beginner April 21, 2013 11:33

Quote:

Originally Posted by Far (Post 422038)
can you show the convergence plots of both cases? Did you check the solution at lower time step ?

The time-step I am using is 1e-6 s. In fact, I always use lower time-step to be safe.
Regarding the convergency, I just check the changes in flow parameters and since for a long time no changes happen for their values, I decided that the flow became steady. Is that enough to check the convergency? Or do I have to check anything else?

Thanks for all your times and care,
Natali

Far April 21, 2013 11:41

What if you use 0.1 s time step and it should not make any difference as case is steady state!

Natali_Beginner April 21, 2013 11:47

Quote:

Originally Posted by Far (Post 422127)
What if you use 0.1 s time step and it should not make any difference as case is steady state!

Sorry,
I am confused. I don't get what exactly you say.

Far April 21, 2013 11:51

I am saying if your case is steady state then it should give same results with steady, unsteady with 0.1 s time step, unsteady with 0.005 s time step and 1e-06 s time step. With 1e-06 time step you may need longer time to reach steady state while with larger time step you may get convergence in less time.

Natali_Beginner April 21, 2013 11:56

Quote:

Originally Posted by Far (Post 422129)
I am saying if your case is steady state then it should give same results with steady, unsteady with 0.1 s time step, unsteady with 0.005 s time step and 1e-06 s time step. With 1e-06 time step you may need longer time to reach steady state while with larger time step you may get convergence in less time.

oh. That is exactly a confusing stuff that I am struggling with that. I had tried to run the case with 1e-6 s and 1e-2 s, before. However, it seems many phenomena are ignored when time-step is large. That seems reasonable. That's why I tried to use small time-step. In fact, based on Courant number and some other factors, I think working with small time step is more safe, especially in combustion problems.
I really appreciate you correct me wherever I am wrong.
Thanks,

Far April 21, 2013 12:40

Quote:

I am wondering how to compare a problem in steady and unsteady.
Compare some time averaged parameters with steady state solution.

Quote:

However, it seems many phenomena are ignored when time-step is large
This may be the reason that your steady and unsteady results are different

Natali_Beginner April 21, 2013 12:43

Quote:

Originally Posted by Far (Post 422135)
Compare some time averaged parameters with steady state solution.



This may be the reason that your steady and unsteady results are different

Yes, actually in my problem there are different time-scales and that's why I have to use very small time-step.


All times are GMT -4. The time now is 22:45.