Clamping both ends (dynamic.py)

[barancimen]

Is it possible to solve this convergence issue when I clamp both ends?

t_root = tangent(coords[0])
n_root = normal(coords[0])  
t_tip = tangent(coords[-1])
n_tip = normal(coords[-1])
w_tip = coords[-1]

bc_t = DirichletBC(V.sub(0), Constant((t_root[0], t_root[1], t_root[2])), root)
bc_n = DirichletBC(V.sub(1), Constant((n_root[0], n_root[1], n_root[2])), root)  
bc_w = DirichletBC(V.sub(4), Constant((0., 0., 0.)), root)
bc_v = DirichletBC(V.sub(5), Constant((0., 0., 0.)), root)

bc_t2 = DirichletBC(V.sub(0), Constant((t_tip[0], t_tip[1], t_tip[2])), tip)
bc_n2 = DirichletBC(V.sub(1), Constant((n_tip[0], n_tip[1], n_tip[2])), tip)  
bc_w2 = DirichletBC(V.sub(4), Constant((w_tip[0], w_tip[1], w_tip[2])), tip)
bc_v2 = DirichletBC(V.sub(5), Constant((0., 0., 0.)), tip)

bc = [bc_t, bc_n, bc_w, bc_v, bc_t2, bc_n2, bc_w2, bc_v2]

I get Newton solver did not converge because maximum number of iterations reached. at the first time step and playing with parameters doesn't help.

Thanks

[mou3adb]

Hi @barancimen,

That's right, though I guess the convergence problem is due to the boundary conditions. I had the same trouble before when I applied a load force on the clamped end, seems that using UserExpressions and point-wise force assignment, Fenics doesn't accept forces applied on points having already boundary conditions. I added in line 242 in dynamic.py another condition and not(np.equal_array(coords[-1], x)) in order to avoid to apply force on the tip. But this didn't help much...

I think Fenics checks all the unknowns if they have a corresponding boundary condition. In your try there isn't any bc for Omega and Fint. Another thing is that there is a surplus of bcs, 8 vs only 6 unknowns to solve for. I suspect that the lagrangian itself should be modified, particularly L_dFds and L_dMds. I'd be curious to see a solution for this issue, despite the scarcity of 1D element Fenics simulations on the web...

Mouad

[barancimen]

I'll let you know when I fix it. Thanks