[sepidehkavousi]

hi every body
I am writing a code for flow in micro channels, but as boundary condition I do not have any information about velocity, I just know I have input pressure 150000 (pa) and output pressure atmosphere (100000 pa). I use finite volume method and simple algorithm.
In Versteeg's book it is defined az physical boundary condition bud I have some problems . I don't get convergence for my code and I think it is something about my boundary condition defining. I wonder if anyone could help me.
my Du and Dv and so for most nodes pressure corrections too small to correct velocity or pressure .
I don't know what to do about under relaxation factor, does it have such important effect on code convergence?
please help me

[sara1987]

is it steady state?
have you done grid size independence check?
have you initialized it correctly?
can u tell me more details about the problem u wanna solve

[rmh26]

Can you assume the flow is undisturbed near the boundaries. ie just a posiuelle type flow. If so then as a reasonable approximation you can set the tangential part of the flow to zero and set the derivative of the normal part of the flow to zero. There are some better approximation you can use if the flow isn't so regular.

"Open and Traction boundary conditions for the incompressible navier stokes equations" Jie Liu JCP 228, 2009

[mihaipruna]

can you try setting gradient limiters?

[sepidehkavousi]

Originally Posted by sara1987

is it steady state?
have you done grid size independence check?
have you initialized it correctly?
can u tell me more details about the problem u wanna solve

Dear sara...yes it is steady state ...at first I make my model from a paper and I used the same grid size and channel size and I give the initial condition near the real results but it does not get convergence....it is about 3d flow in a micro channel with two fixed boundary condition in inlet and outlet and no slip condition...water is flowing in the channel, I did every thing exactly written in Versteeg for fized boundary condition, but I can not find the problem

[sepidehkavousi]

Originally Posted by mihaipruna

can you try setting gradient limiters?

Hi...I couldn't understand you...can you tell me more about this?

[sepidehkavousi]

Originally Posted by rmh26

Can you assume the flow is undisturbed near the boundaries. ie just a posiuelle type flow. If so then as a reasonable approximation you can set the tangential part of the flow to zero and set the derivative of the normal part of the flow to zero. There are some better approximation you can use if the flow isn't so regular.

"Open and Traction boundary conditions for the incompressible navier stokes equations" Jie Liu JCP 228, 2009

thank you my dear friend...I would study the paper and ask you If i had problem...

[mihaipruna]

Originally Posted by sepidehkavousi

Hi...I couldn't understand you...can you tell me more about this?

well, when I had this problem in OpenFOAM, I used one of the following settings. I think it gives alink to the code, maybe you can implement it in your program. good luck!

http://openfoamwiki.net/index.php/Op...guide/Limiters

[sepidehkavousi]

Originally Posted by mihaipruna

well, when I had this problem in OpenFOAM, I used one of the following settings. I think it gives alink to the code, maybe you can implement it in your program. good luck!

http://openfoamwiki.net/index.php/Op...guide/Limiters

Hi
I am writing my code with fortran language...didn't you understand exactly what OpenFOAM did in that case?

[mihaipruna]

I haven't reviewed the source code for OpenFOAM, but for your program, you should be able to calculate the difference of inflow and outflow in a cell, for any quantity/ direction, divide by cell size in that direction and if the result is larger than a certain number, repeat the iteration with a smaller time step, add relaxation factors maybe, or simply set that gradient to the maximum allowable value and proceed from there, I think this is what OpenFOAM does and although it introduces some accuracies (not much in my case, few cells have that problem), it allows the solution to converge.