CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Problem with periodic boundary layers (

Phil September 25, 2000 12:12

Problem with periodic boundary layers
I try to investigate a flow (water) in a duct steady state, laminar with exchange of heat. I have set periodic boundary conditions for the in and outflow so I try to simulate fully developed flow. To reach faster convergence I try to solve the problem by a step to step solution process. I am using the segregated solver so I turned of turbulence (k-epsilon RNG) and energy. After convergence of the laminar flow I turned the laminar equations off and solved only the turbulent equation to get a good initial guess for k and epsilon.

**Here I have the first question, why must I not make an initial guess for k and epsilon. When I simulate without periodic boundary layer I have to give at the inlet and outlet (in the boundary condition panel) an initial guess for k and epsilon, or alternatively the hydraulic diameter and the turbulence intensity.

Anyway I solved the turbulence equations and turned on after a while the flow field. After a while I had convergence and the results seemed to be in agreement with what I have expected. Then I turned of the flow field and the turbulence, and turned on the energy. I'm using constant wall temperatures and made a guess for the bulk temperature in the periodic boundary condition panel. But with the default values for the relaxations factors for the energy field the calculation didn't converge and the temperatures were limited in nearly all cell volumes to a value about 1 Kelvin, which seems to be a little bit too cold. ;-) When I decrease my relaxationfactor for energy from 1 to 0,1 the value of cells where the Temperature is limited is less and the iteration seemed to converge but the results weren't impressing. Perhaps I have to wait longer because I was far away from the convergence (Residual about 1e-3) but it seems to be a general problem.

Anybody has got an idea?????

Thank you Phil

John C. Chien September 25, 2000 13:55

Re: Problem with periodic boundary layers
(1). For laminar flows, you don't need to solve the k-epsilon equations. (2). For fully developed flows, just run a calculation for a long duct, with regular inlet and exit conditions. You don't have to use the so-called periodic boundary conditions. (3). You are solving the problem in a very strange way. My suggestion is: solve the flow field and the energy equation at the same time with a long duct and regular boundary conditions.(velocity inlet and downstream exit conditions).

Oliver September 25, 2000 14:59

Re: Problem with periodic boundary layers
Hi Phil

1. When using the energy equation, Fluent recommend that the residual sum be set to 1e-6 so all equations have to be computed in tandem. I would even recommend an order of magnitude higher. So you solution is no way near convergence. 2. If the flow is laminar because the critical Reynolds number says so, why are using a turbulence model?. 3. You can use that data you have already obtained for the laminar case as initial guess for the final solution.

Regards Oliver

Sandeep September 25, 2000 16:14

Re: Problem with periodic boundary layers
Yep, use a residual conv. criterion of 1e-09 or something for energy. The default from fluent doesnt seem to work well in all cases. I had similar problems.

Phil September 25, 2000 16:17

For John Chien
I think you misunderstood me because I only started to tackle the problem laminar, then I turned on the turbulence equations with the calculation of k and epsilon. In the Fluent manual this is called step by step solution. Then after solving my turbulent flow I turned on my energy equation. So what I definetivly have is a turbulent flow!! The problem is that the energy equations don't converge. Hope I could explain my problem more clearly than before. So any ideas why I have Temperatures in the range of 1 Kelvin????

Thank yopu for your help

Phil September 25, 2000 16:20

For Sandeep
The proble is not the convergence criterium, it's the result with a wall temperature about 1 Kelvin!! Thank you for your help

Regards Philipp

Phil September 25, 2000 16:47

Sorry I ment turbulent instead of laminar flow!!!!
I only started with a laminar solution finally I want to solve the turbulent problem with turbulent equations. Sorry for this mistake

Regards Philipp

John C. Chien September 25, 2000 17:07

Re: For John Chien
(1). Well, so the flow is turbulent. And I think, it is all right to use the laminar flow solution as the initial guess for the turbulent flow equations. (2). My guess is that if you use the periodic condition, the inlet and the exit temperatures will be the same. But it will be floating. (3). So, you need to use the regular inlet condition to specify the temperature, and the velocity.

Oliver September 26, 2000 15:51

Re: Sorry I ment turbulent instead of laminar flow
4 questions for you.

1. Whats the temp. of the wall? 2. Whats the temp. of the fluid? 3. How fast is the fluid moving? 4. Is the temperature relative?


Phil September 27, 2000 06:30

For Oliver
Answers for you: Temperature of the Wall 350 K Temperature of the Fluid 300 K velocity 0,4 m/s (normal to the boundary) I don't know what you mean with relative temperature, i think the temperatures i gave for the wall and the fluid are the total temperatures!

Thank you for your help


Chetan Kadakia September 28, 2000 06:09

Re: Problem with periodic boundary layers
Obtaining 1K for the fluid temperature when you're walls are at a constant temperature of 350K and the fluid is initially at 300K is very odd. I suggest that you verify that you have set your initial conditions (Solve-->Initialize...) Also allow your solution to converge to a higher energy equation criterion (perhaps 1e-6). If the solution coverges and the Temperature flow field is reasonable (between 300K and 350K), then switch over to the turbulence model. Otherwise, review your B.C.'s, problem set up, and set a higher convergence criterion. If your problem persists, you can zip and mail the mesh and case file to: Cannot promise any answers but I'll give it a shot.

Younes Khamliche September 28, 2000 11:16

Re: Problem with periodic boundary layers

I agree with John Chien, when using periodic BC's all variables at those boundaries are equal. You might need to change your BC's.

Phil September 29, 2000 08:31

Re: Problem with periodic boundary layers
But imagine you have a tube and the wall has got a constant temperature and you want to simulate the fully developed flow which is possible with periodic boundary conditions, so that the flow leaving from the outflow is linked to the inflow. This is possible in Fluent and they describe a solution process in the manual where the first solve the turbulent flow and then turn of the flow and turbulent equation and turn on the energy equations. The same thing I did and the results are rubbish. I solved the problem in destroying the periodic boundary and switched of the flow and turbulence equations. For the inflow I took a pressure boundary inlet and for the outflow a pressure outlet, with the initial pressure values found by the isothermal simulation. I solved the energy equations and assume that a change of 2 to 4 Kelvin in my average flow temperature doesn't influence dramaticly the flow.

Thank you for your help


Oliver September 30, 2000 11:31

Re: For Oliver

There is a 50 degree temperature gradient across the thermal boundary layer, which implies that you're B.C are incorrect, even for a relatively coarse grid. Check out the tutorial guides.

Regards Oliver

Phil October 1, 2000 09:28

Re: For Oliver
I can't change this bc because this is what happens in my heat exchanger that my fluid enters with a temperature of 300 K in my geometry and is heated by a nearly constant wall temperature of 350 K. Do you think there is a problem??

Regards Phil

pavel October 2, 2000 14:56

Re: Problem with periodic boundary layers
I am not sure but those problems depend on the laws of conservation on the each cell and for whole domains (fluid, solid, conducting walls). The very cold (;-), very hot or both temperatures show the unbalance of energy equation in the domains. Fluent sets those temperature and the energy balance is fulfiled. The steps below can help to disappear the unrealistic temperature sometimes: 1. Adaptation the mesh by iso-values. Choose only cells with unrealistic temperature (very often one or two cells only). 2. Again use the adaptation by iso-value, but don't adapt !!. This I use for the select of cells with unrealistic temperature. 3. Patch on selected cells the REALISTIC temperature (300K or 350K) by SOLVE, INITIALIZE, PATCH and select REGISTER TO PATCH (no Zones to Patch !!!) 4. Start the solution, maybe a few iteration and check the temperature field again. When you see the "my" unrealistic temperature to jump on 1. Two, three, ..., tentimes. 5. Sometimes the points 2. and 3. help only.

Good luck Pavel

All times are GMT -4. The time now is 08:09.