That worked, i and j were the wrong way round. Thank you for spotting that. For some reason I also thought n was an inward pointing vector. ________________________________________ From: firedrake-bounces@imperial.ac.uk <firedrake-bounces@imperial.ac.uk> on behalf of skramer <s.kramer@imperial.ac.uk> Sent: 23 April 2016 13:50 To: firedrake@imperial.ac.uk Subject: Re: [firedrake] Pressure outlet boundary condition in plane poiseuille flow On 23/04/16 13:20, Fryderyk Wilczynski wrote:
Thank you all for the responses.
Colin: you're right no need to specify zero pressure strongly. Yes I'm checking against exact solution U_x = y(1-y), U_y = 0.
Stephan: I think I understand what you mean but can you just check below. I have tried specifying tangential velocity on the boundary strongly - that did not work (solution just didn't anything like the exact). So I tried adding the boundary integral term into the formulation weakly following the reasoning below:
[cid:ac887fc4-72ca-4e5f-b746-3bf596f810e0]
I think your i and j might be swapped: nabla_grad(v)[j, i] means dv_i/dx_j whereas your formula (correctly) has dv_j/dx_i. Also, I'm not entirely sure what you mean by "minus sign as n points inside the domain". Both your formulas and FacetNormal are using outward pointing normals Cheers Stephan _______________________________________________ firedrake mailing list firedrake@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/firedrake