So if I'm understanding this correctly, for the flux F , testfunction phi and a boundary \Omega_h, I'll have \int phi F .n ds(\Omega_h) on the boundary to impose the boundary conditions weakly. So if i include known information as the numbers and unknown information as the variables, then this would fulfil my Dirichlet bcs. In the case of non linear shallow water, then it would be +(hu*phi)*ds(1) + (hu*phi)*ds(2) # boundary flux for height +(0.5*g*1+ hu*hu/1)*xi*ds(1)+(0.5*0.12*0.12 + hu*hu/0.12)*xi*ds(2) #boundary flux for momentum hl = 1, hr = 0.12 As the height and not the momentum are known on the boundary. Will ________________________________________ From: firedrake-bounces@imperial.ac.uk <firedrake-bounces@imperial.ac.uk> on behalf of Lawrence Mitchell <lawrence.mitchell@imperial.ac.uk> Sent: 23 February 2016 12:45 To: firedrake@imperial.ac.uk Subject: Re: [firedrake] Facetnormal compile error On 23/02/16 12:23, William Booker wrote:
Dear all,
I now get the following error when attempting to do 1 application of the nonlinear solver.
We had not anticipated the use of geometric boundary conditions on interval meshes, hence your error. If you raise an issue on github we can endeavour to fix this. However, once again, I recommend that if you have a discontinuous discretisation like this, to apply your boundary conditions weakly: when you integrated by parts you must have obtained some ds integrals which you have eliminated from your form by strongly imposing boundary conditions on the ds(1) and ds(2) integrals. You could instead enforce those conditions weakly. Cheers, Lawrence