[Mishra Apoorv]

I am new to the community and CFD itself and am working on my first CFD program in Scilab based on the SIMPLER algorithm i.e. modeling a flow through a rectangular cross section channel. I want to get a Poiseuilli kind of flow profile but I'm stuck at a couple of boundary conditions.

I have given a dirichlet type of Boundary condition i.e. constant velocity inlet at the face of the inlet and no slip at the 4 walls but I'm not understanding what should be the y and z (perpendicular to the length of channel) direction velocities (v & w respectively) at the inlet and outlet of the channel.


It's a 3D staggered grid arrangement like the image below (image just for illustration), the black points being the points for pressure.

https://drive.google.com/file/d/1oBm...ew?usp=sharing



Let's say the red velocities on the extreme left are the constant u inlet velocities and the right ones being the outlet. So, for such an arrangement, there exists a series of v velocities (green) on the left of the inlet and also on the right of the outlet out of the main domain. For the no slip boundary conditions at the walls I did,

v( :,:,1 ) = -v( :,:,2); &
v(:,:,npz+1) = -v(:,:,npz); (npz = (no. of pressure elements along z direction) + 1)

and similar for u & w.


But what should be the condition for v and w at the inlet?
I've tried v(1,:,: ) = -v(2,:,: ) & same for w since both should be 0 at inlet face but it doesn't work and the solution seem to diverge after just the first iteration.


Also I just want to give the constant inlet velocity as the initial condition so is there any need to specify pressure as boundary condition anywhere? Because as opposed to SIMPLE, SIMPLER requires an initial guessed velocity field and no pressure guess.

Any help will be appreciated.

[sbaffini]

Let's split what you want from how you implement it.

At inlet, assuming an incompressible flow, you should assign all the velocity components, and having v and w as 0 is probably what you want.

For the implementation,.it depends from what code you have. Is this a staggered finite difference code?

[Mishra Apoorv]

Thanks for taking time to respond.
I've edited the question with details.

[sbaffini]

Originally Posted by Mishra Apoorv

Thanks for taking time to respond.
I've edited the question with details.

I haven't used staggered stuff in more than a decade, so I can only say that what you have shown of your implementation seems correct to me. Yet, that's just the textbook part of a whole code, the error might be everywhere.

Have you tested the code in 2D, or even 1D using symmetry bc and just one cell in that direction? Have you tested the code on any other case?

[Mishra Apoorv]

Originally Posted by sbaffini

I haven't used staggered stuff in more than a decade, so I can only say that what you have shown of your implementation seems correct to me. Yet, that's just the textbook part of a whole code, the error might be everywhere.

Have you tested the code in 2D, or even 1D using symmetry bc and just one cell in that direction? Have you tested the code on any other case?

Yes I've tested the code for the 1D case and the problem is the same. I've managed to narrow the causes for it, like I think the code for pseudo velocity computation is the one which messes the whole solution though the algorithm is correct theoretically. As mentioned earlier, it gives satisfactory output (i.e. correctly applied boundary condition and roughly parabolic nature of velocity profile could be seen) at the first and second iteration but starts giving random values, even negative thereafter.


https://drive.google.com/file/d/1Rwg...ew?usp=sharing



Here I'm attaching my code for reference. If you have any insights or have worked with SIMPLER before, then please suggest how to troubleshoot such issues and if there's any good source to study the coding for CFD/FVM that you know of, then please mention it.



Thank you.

[sbaffini]

Originally Posted by Mishra Apoorv

Yes I've tested the code for the 1D case and the problem is the same. I've managed to narrow the causes for it, like I think the code for pseudo velocity computation is the one which messes the whole solution though the algorithm is correct theoretically. As mentioned earlier, it gives satisfactory output (i.e. correctly applied boundary condition and roughly parabolic nature of velocity profile could be seen) at the first and second iteration but starts giving random values, even negative thereafter.


https://drive.google.com/file/d/1Rwg...ew?usp=sharing



Here I'm attaching my code for reference. If you have any insights or have worked with SIMPLER before, then please suggest how to troubleshoot such issues and if there's any good source to study the coding for CFD/FVM that you know of, then please mention it.



Thank you.

I can't really help you here, as I am not really acquainted with the SIMPLER method. Have you tried with other SIMPLE variants? Clearly, if it doesn't work in 1D, where you should get a constant velocity, there must be a problem in something larger than simply bc.

The way these things are debugged is to focus on the most simple problem that doesn't work, possibly on the coarsest possible grid, and then analyze the algo step by step, comparing at each step what you get with what you expect. At some point you must find something that doesn't work, then you investigate its input. If it is legit than you have found your bug, otherwise you proceed backward until the input to a line is legit but the output isn't.

[FMDenaro]

Originally Posted by Mishra Apoorv

But what should be the condition for v and w at the inlet?
I've tried v(1,:,: ) = -v(2,:,: ) & same for w since both should be 0 at inlet face but it doesn't work and the solution seem to diverge after just the first iteration.


Any help will be appreciated.

Why? You should suppose that this is a Dirichlet condition corresponding to a flow having only one velocity component, the one in streamwise direction,