CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Duct Flow (

omar December 6, 1999 22:28

Duct Flow

I am modeling a duct flow with one inlet and four outlets. The duct cross section is non-circular and very irregualr. The objective of the simulation is to predict mass flow distribution at the outlets for a given mass inlet. The flow is incompressible and turbulent. I am imposing a velocity inlet and a pressure outlet ( outlets were placed 10 diameters downstream to avoid reversed flow).

I was able to obtain a solution where the residuals reduced to 1.e-5 smoothly, but the mass flow rates at the outlets were oscillating. In other words, outlet mass flow rates are changing with iteration.

I have tried different boundary conditions and different turbulent models with no success. I need your comments on why outlet mass flow rates keep changing with iteration even when residuals are very low .


Mohammad Kermani December 7, 1999 01:52

Re: Duct Flow
I would increase the domain of computation to let the flow to better develope before exititing the computation domain. Also I would let it run for a quite number of iteration more, even there are not much reduction in residuals or even the level of residulas have become flat.

Good luck.

I am intersted to know what happened.

Regards. MJ

John C. Chien December 7, 1999 03:58

Re: Duct Flow
(1). It is likely that your problem is outside the normal scope of the commercial code you are using. (2). The current knowledge to solve a fluid dynamic problem is still very limited. Each case is a special case. If it is not solved validated before, there is no guarantee that the commercial code will produce converged solutions. (3). This is the reason why various test cases were developed to test the capability of CFD codes. And, even for a very standard test case, various codes and methods are being developed and tested over and over again. (4). My feeling is, the configuration of your problem with four outlets is probably too difficult for the code, and one way to find out is to start with just a couple of outlets first. And make sure that you can get converged mesh independent solution with two outlets. (5). Once you have learned how to do it, or once you have demonstrated that it can be done with this code, you can then add one more outlet to your problem. (6). You also need to know whether the boundary condition you use will generate a unique and stable solution with four outlets. (you will have to do this with the help of the support engineer, I hope) (7). Creative thinking is not very productive when using a commercial code. (8). Based on my experience, even if the code is designed to solve a particular problem, it does not mean that it will work for the problem with different set of conditions.

omar December 7, 1999 21:40

Re: Duct Flow
Thanks for your comments. I will block some of the outlets and see how it works. I will also try with a different commercial solver.

Thanks again..

David F. December 9, 1999 13:43

Re: Duct Flow
Perhaps the physical solution is truly unsteady, and your solver cannot converge on a steady solution because one does not really exist. To verify this you could use an unsteady method.

All times are GMT -4. The time now is 08:59.