CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Numeca Fine (

dimitris June 5, 2001 04:22

Numeca Fine

I was wandering if anybody knows why when running a problem with prescribed mass flow in inlet and outlet leads to divergence and why if i run the same problem with prescribing the velocity vector values in inlet and mass flow in outlet does not lead to divergence.Is there any way to make the code ''accept'' more numerical instabilities so as to accept the mass flow inlet and outlet boundary conditions? Furthermore the combination of prescribing velocity vectors in inlet and mass flow in outlet is correct or wrong even if i can calculate the mass flow in the inlet boundary condition from the velocity vectors?

Thanks for your time

John C. Chien June 5, 2001 12:28

Re: Numeca Fine
(1). There are two aspects of your question, namely, (a). Do you know the proper way to formulate your problem, ie, how to define your problem properly (mathematically), (b). Once the problem is defined properly (we don't know what your problem is, yet), do you know how to find a solver algorithm to find the converged solution? (2). These questions are essential. And if the problem (including boundary conditions) is not defined properly, and the solution procedure is not properly selected, then it is difficult to find a converged solution. (3). For the commercial codes, "who knows what is in the black box". For that part, you will have to discuss it with the vendor's support engineers. (4). It is very common to hear that a solution does not converge when using a commercial code. So, when you use a commercial code, follow the sample problems provided. If the sample prblems don't converge, then try something else. (5). You could send your suggestions to the vendor for the code you are using.

clifford bradford June 6, 2001 19:54

Re: Numeca Fine
Specifying inlet and outlet mass flow is probably not well posed. Are you solving an intertnal flow?

Fine is a time iterative scheme. the initial condition is specified somewhat arbitrarily. the only way you can specify the inlet and outlet mass flow is if you are certain that at every time step that the mass into any control volume in your domain is exactly equal to the mass flowing out. Of course this is isn't the case because the mass residual in a time iterative solution isn't zero initially (and at no time in the solution even when you get to "machine zero"). at least some portion of your boundary must be "free" to flow as necessary.

I. Dotsikas June 15, 2001 07:48

Re: Numeca Fine
Hi Mitso,

I try to understand your question. If you write down your problem, you have a set of equations. So far so good. In order to be able to solve your equations and obtain reasonable results, you have to define boundary conditions. That is nothing new. If you specify your Problem you get a set of boundary conditions, that actually specify your problem and so they depend on the grad of your differential equations.

If you have the mass flow in the entrance and nothing more, you donīt spacify any problem. You specify a number of cases. You may have a number of flow fields that all have the same mass flow in the entrance and a different velocity distributions, not only in the entrance region. You may get a recirculation flow, a fully distributed flow comming from a plug flow ... All these flows have in common the same mass flow in the entrance. (The mass flow in the outflow boundary is usually the same as in the entrance, unless other supplenentary in or outflow regions are availlabe <= Answer to your second question!!!). In this case is your problem not set properly. You need a set of boundary contitions in order to have a "closed Problem", a well defined problem and not a number of problems. If not, the poor computer may try to find a solution that satisfies your equations. It depends on how sofisticated is your code. Usually divergense is the end of the story. Having a grid you may define the mass flow to any inflow cell, that would work. but definining the mass flow all over your cells that would not work, because this mass may be distributed in many ways and not in a unique one.

If you try to make your code accept more instabilities that it is supposed to do, then you donīt do right. Anyway, the answer your code gives you can be as good as your question. If you give carbage in, you get carbage out, with a fast computer you may get you carbage quite soon, as John would say. The difference between an experienced enginner an a beginner is that the experienced poses the proper questions.

hope i helped,


clifford bradford June 16, 2001 05:05

Re: Numeca Fine
Dimitris, has your question been sufficiently answered. I'll be happy to discuss it further.

dimitris June 17, 2001 11:43

Re: Numeca Fine
Thx everyone for their help i will try to follow the advices.

All times are GMT -4. The time now is 18:42.