[kriver]

In FLUENT 6.3 User's Guide,it's said that transient simulations for fully-developed fluid flow are not valid with translational periodic flow.I wonder if that means fluent can not do transient simulations with periodic boundary conditions.Can any one help me?Thanks a lot!
-------------------
Kan R

[sbaffini]

I don't think it's correct because i'm actually performing unsteady simulations with streamwise periodicity. A different question is "what are you simulating with such a simulation?" but it doesn't means it is imposssible to do.

[Volker P.]

Originally Posted by sbaffini

I don't think it's correct because i'm actually performing unsteady simulations with streamwise periodicity. A different question is "what are you simulating with such a simulation?" but it doesn't means it is imposssible to do.

Hello Paolo,

together with my working student we are planning to do the same like "kriver" with a translational periodic element of a heat exchanger. We have red the same limitations in the user 's Guide and are now wondering whether a transient or unsteady laminar simulation (up to DNS) is still possible. Do you have further information?

Volker

[sbaffini]

Hi,

the limitation on the use of periodic boundary condition is just physical. We can define 3 classes of incompressible flows:

1) Adiabatic flows: in this case the zero heat flux on the whole boundary makes the periodicity a feasible choice; in fact this prevents the energy of the flow to blow up. So, periodicity = feasible

2) Fixed temperature on the whole boundary: in this case the periodic condition is feasible because the flow temperature will change in time until the temperature is uniformly distributed.

3) Fixed non zero heat flux on the boundary: in this case the continuous heat transfer will make the energy of the flow to blow up without bounds. In this case the periodic condition is not feasible.

The obvious reason for which the former reasoning is true is because with a periodic streamwise boundary condition the flow exiting the domain will actually enter again from the inlet and if at every passage the energy will grow because of the fixed heat flux...well, it's easy to see that the energy will grow without bounds. In the same way follows the explanation for the other cases.

In my case, i'm performing a simulation of an unsteady turbulent incompressible adiabatic channel flow so the periodic condition is feasible because there is no energy flux in or out of the domain.

In your case, the heat exchanger, the periodic condition seems not feasible (unless it doesn't exchange at all ) and a classical velocity-inlet/outflow combination should be the best choice.

hope it helps

EDIT:

I'd like to add two more points:

1) The former reasoning is still valid also in the case of a steady simulation...in this case, if the periodic condition is not physically sustainable the solution will blow up in the same way
but in the "framework" of the algebraic solver

2) I'm not sure but, if your periodicity is in the spanwise (3D case) direction then it is going to work. In this case the flow will definitely exit the domain and the periodic condition will not give rise to any anomalous behaviour

[Volker P.]

Thanks a lot for your answer,

fortunately our case has fixed temperature boundaries (a fixed bulk temperature at the periodic "inlet" and fixed wall temperatures). Hence according to your explanations this periodic condition is feasible. The Fluent users's guide tells the same.

Our main problem is the Users's Guide statement, that streamwise periodic flow AND transient mode do not give valid results! But you said s.th. different and this is encouraging. I have posed the question to the fluent staff today and I am looking forward to an answer.

Volker

[sbaffini]

There is no reason for this to be unfeasible. It's just a matter of what you want to simulate. In this case you can't actually fix a temperature at the inlet because it is periodic. You will actually need to define an initial value of temperature in the whole domain and the temperature on the wall, then just start the simulation.

However you should be aware of what you're simulating in this case. Periodic boundary conditions will be such that you will follow the same mass of fluid along the streamwise direction while it is interacting with the fixed temperature of the wall. So, it is inevitable, at some time the whole fluid will have the same temperature everywhere.

what i mean is that this kind of boundary conditions will not give you a boundary layer like solution growing in the streamwise direction with fixed inlet and outlet heights, but just the same mass of fluid passing over and over through the domain until it will reach a steady thermal state. If, for example, you are interested to know at which position streamwise the fluid will reach some temperature this will not be immediately achievable but should be somehow extracted from the time evolution.
In this sense the periodic boundary condition is not physical for time dependent flow (but still feasible).

In fact, in my case the flow is unsteady (it's a large eddy simulation) but will reach a time in which it is in a statistically steady state. For example if i had the two parallel walls at two different temperatures, even at the steady state i had a physically relevant case (to study, for example, the scalar mixing at the inteface of the two temperatures).

I don't know if i've been clear enough

[nenazarian]

Hi

I've been reading your conversation and its been really helpful.
I am running a similar case, trying to simulate the urban environment and for now Im doing that on a uniformly spaced rows of buildings with same height with periodic boundary condition in both stream-wise and span-wise direction. ( so that it would form an unlimited rows and columns of buildings)

right now Im only concerned with the flow as the 1st step of simulation.
I tried both RANS and LES methods and LES has the best results so far. (and its the method I want to use in the end)

I want to move to the next step which is simulating heat transfer and now Im concerned that the energy of the flow might blow up.
especially since I will be using solar calculator in FLUENT...

Do you think my case would be feasible with streamwise periodic boundary condition?

I appreciate your help.

Cheers
Negin

[sbaffini]

Dear nenazarian,

i worked with LES/URANS on the (conjugate) heat transfer in a matrix of surface mounted cubes. This is an ERCOFTAC test case which, i presume, is very similar to your case; in my case i had streamwise-spanwise periodicity, two parallel adiabatic walls and a single, internally heated (fixed temperature) copper cube on one of the walls.

In my case, among the infinite number of cubes, only one was heated so the solution was to define a very thin zone at the inflow of the domain where the periodic temperature was overwritten by a fixed temperature which was known from the experiment (in Fluent you can fix variables in your defined zones).

If you are interested in the heat transfer for all the cubes then probably you can skip the fixed temperature zone and go for straight periodicity, however i suggest you to carefully read the paper (and the manual) concerning the way periodicity is handled in Fluent when heat transfer is involved to check if it is suitable for your case:

http://dx.doi.org/10.1115/1.3450666

I don't know about the solar calculator but i guess you can still find any limitation in the manual

[nenazarian]

Dear Paolo

Thank you for your response.
My only concern is that in your case, wall condition was either isothermal or zero-flux, while my case temperature of the wall is to be determined using heat transfer simulation and the fluxes on walls are not zero. from your earlier discussion, Im concerned that it might not be possible with periodic boundary condition...

[sbaffini]

Dear nenazarian,

i guess you already figured it out but, as long as fixing the temperature in the inlet zone is relevant for your case (the flow can be conveniently assumed periodic but you have fixed inlet temperature) then no difference exists between temperature and heat flux b.c.s; whatever you do, everything should be fine.

[LuckyTran]

Originally Posted by sbaffini

Dear nenazarian,

i guess you already figured it out but, as long as fixing the temperature in the inlet zone is relevant for your case (the flow can be conveniently assumed periodic but you have fixed inlet temperature) then no difference exists between temperature and heat flux b.c.s; whatever you do, everything should be fine.

There is substantial difference between the way fluent treats a periodic heat transfer prob. The way the applied wall temperature and applied heat flux boundary conditions are nothing alike! Be careful!

For temperature, the exit temperature profile is rescaled and applied at the inlet after subtracting the change in bulk temperature.

For applied heat flux, a sink term is added to the energy equation to maintain a constant bulk temperature across the cross section.

If possible, I recommend using an applied heat flux in the case of periodic b.c.'s. The setup of the prob is more robust. The way the temperature profile is rescaled is rather arbitrary for the applied wall temperature case.

[sbaffini]

"There is substantial difference between the way fluent treats a periodic heat transfer prob. The way the applied wall temperature and applied heat flux boundary conditions are nothing alike!"

Well, not to be annoying, but i was not contradicting this (which would have no sense, also because i cited the reference above. By the way, see also: Mathur, Murthy - Periodic flow and heat transfer using unstructured meshes). What i literally said is:

"as long as fixing the temperature in the inlet zone is relevant for your case [...] then no difference exists between temperature and heat flux b.c.s; whatever you do, everything should be fine"

What i meant, in case it was obscure, is that you can run in both modes and still get significant results (as long as fixing the temperature at inlet is significant; otherwise i don't know if in unsteady mode everything is still fine with any of the two methods). Actually, the "should" was not arbitrary because i never used the fixed inlet condition with heat flux b.c.(besides now, as i tested it in steady mode and it seems feasible) so i was not really sure, exactly because i don't know the way this source term is handled in the simulation and how can affect it.

Moreover, if you want to be more comfortable with that, i guess you can still apply a fixed temperature B.C. with a fixed inlet temperature and then use an UDF to convert the heat flux in a boundary temperature; in this case you are sure there is just a temperature scaling (no weird source terms), which you kill with the fixed inlet temperature, and you're happy with the heat flux at boundary. Also, with this method, you could try use both kind of b.c. together (which is not allowed by fluent of course). However, i never tried this.