Redundant pressure boundary condition in Stokes demo

In demo-stokes_taylor_hood.py, the Stokes problem is solved with the following boundary conditions

$u = u_0 $ on $\Gamma_D$ and $\nabla u\cdot n + p n = 0$ on $\Gamma_N$

These are reflected in the weak form as neither the bilinear nor the linear forms include any surface terms. The velocity boundary conditions are set strongly with

bc0 = DirichletBC(W.sub(0), ...)
bc1 = DirichletBC(W.sub(0), ...)

Later in the code, pressure is enforced strongly on $\Gamma_N$ with

bc2 = DirichletBC(W.sub(1), zero, sub_domains, 2)

and the problem is solved with all three boundary conditions. I believe there is no need to use the pressure boundary condition as the stress prescibed on $\Gamma_N$ is already included weakly in the form. The problem solves fine without it

# Output without pressure bc
Norm of velocity coefficient vector: 82.6968599618892
Norm of pressure coefficient vector: 4147.69457577137

# Output with pressure bc
Norm of velocity coefficient vector: 82.699892467548
Norm of pressure coefficient vector: 4146.57536645252

Note that if this is found a bug, it should be fixed also in the remaing Stokes demos, stokes-iterative, stokes-mini and stokes-stabilized.

Comments (6)