Assign solution from previous problem to new problem

[mircea...@gmail.com]

Hi everyone!

I want to use the solution of one pde that I solved in fenics as initial conditions for the Newton solver on another problem.  

So, for the first problem I defined a mixed function space :

mesh = UnitSquareMesh(32, 32)

# Define function spaces

P1 = VectorElement("CG", mesh.ufl_cell(), 2)

P2 = FiniteElement("CG", mesh.ufl_cell(), 1)

element = MixedElement([P1,P2])

W = FunctionSpace(mesh, element)

After setting the boundary conditions and formulating the problem I solved it and saved the result in a h5 file format

w = Function(W)

solve(a == L, w, bcs) #bcs are the boundary conditions

(u, p) = w.split()

#Save result

stokes_file = HDF5File(MPI.comm_world,"data/stokes.h5", "w")

stokes_file.write(u, "velocity")

stokes_file.write(p, "pressure")

stokes_file.close()

Now, I would like to use this saved file for my new problem, which is defined in the following space

mesh = UnitSquareMesh(32, 32)

P1 = FiniteElement("CG", mesh.ufl_cell(), 1)

P2 = VectorElement("CG", mesh.ufl_cell(), 2)

P3 = TensorElement("CG", mesh.ufl_cell(), 2, symmetry = True)

element = MixedElement([P1, P2, P3])

W = FunctionSpace(mesh, element)

So I read the file with the solution of the variables

V = VectorFunctionSpace(mesh, "CG", 2)

Q = FunctionSpace(mesh, "CG", 1)

stokes = HDF5File(MPI.comm_world,"data/stokes.h5", "r")

w = Function(W)

(q, v, psi) = TestFunctions(W)

u = Function(V)

p = Function(Q)

#load result from Stokes here

stokes.read(u, "velocity")

stokes.read(p, "pressure")

#Create Tensor

sigma = grad(u)

sigma = project(sigma, W.sub(2).collapse())

But when I run the Solver for nonlinear variational problems I get this error:

Error:   Unable to successfully call PETSc function 'MatSetValuesLocal'.
*** Reason:  PETSc error code is: 63 (Argument out of range).

I also tried to assign those values to my function, like this:

assign(w, [p,u,sigma]) 

But I get the same error. Can someone point me towards a solution to this problem?

Thank you!

[Jan Blechta]

I would be skeptical relying on the symmetry=True option. There are some
bug reports on the FFC and/or DOLFIN issue tracker.

Jan

[mircea...@gmail.com]

Yes, I know, I read the bug reports. But I get the same error with or without the symmetry argument. I think I am not properly assigning my functions onto the new function space. Is there some documentation on how to use functions from a h5 file format as initial values for the variational problem?