This is a read only copy of the old FEniCS QA forum. Please visit the new QA forum to ask questions

How to "manully" control the boudary value?

0 votes

Hi,

I am implementing Gauge Uzawa method,
and I find the velocity does not satisfy the boundary conditions, for example: u=0;
The velocity correlation step as follows,
u = \hat{u} - \grad(rho);
where, \hat{u} = 0, on boundary, and
\rho is calculated by poisson equation,
\laplace(\rho)=\div(\hat(u)), with neumann condition.
The relative paper:
http://www.math.umd.edu/~rhn/papers/pdf/gau_uza.pdf

Now, I want to control the boundary condition of u = 0.
One way, I try to control the value of u on boundary, directly.
Similar, questions:
https://fenicsproject.org/qa/12578/simply-assign-value-function-unique-vertex-with-coordinate
https://fenicsproject.org/qa/11551/how-to-attach-value-of-number-to-a-cell-of-the-mesh
The seconde way, if we can make sure \grad(\rho) = 0, on the boundary, then, u = 0, naturally.
Some topics;
https://fenicsproject.org/qa/12184/prescribed-normal-velocity-and-tangential-traction?show=12184#q12184
https://fenicsproject.org/qa/8543/tangential-gradient-boundary-conditions

So, I have to find the boundary points, and then, attach values of them, how about high order elements? and the idea is suitable for this problem?

Best Regard!

asked Mar 4, 2017 by Hamilton FEniCS Novice (500 points)

1 Answer

0 votes

Hi,

unfortunately I can not open the paper, so I have to try to answer it without. I am not entirely sure what the structure of your problem is, but if you have Neumann boundary data for example you will not match it exactly.
I had a similar problem and my experience with setting the boundary values manually was not very good and I suspect this is because they will not fit to the rest of the numerical solution anymore.
This approach can be done for P1 easily for example, but I am pretty sure that it will be very nasty for higher order methods, simply because the function value of a certain point on the boundary will be influenced by more than one coefficient then.
I hope that helps you a bit, I am only trying to learn Fenics aswell.

Best,
Sebastian

answered Mar 7, 2017 by Sebi9496 FEniCS Novice (290 points)

Thanks for your comments. Your experience is valuable for me .

...