Solving the heat equation by FEniCS

Question

I am facing a coding error in FEniCS when I run the heat equation code it’s fail with,

error = np.abs(u_e.vector().array() - u.vector().array()) AttributeError: 'dolfin.cpp.la.PETScVector' object has no attribute 'array’

I believe the idea is to calculate the difference between u and the exact solution for each time level and put them in array then we going to take the maxium value.(if not, correct me please!)

the Original code for computing the error is

u_e = interpolate(u_D, V)
error = np.abs(u_e.vector().array() - u.vector().array()).max()
print('t = %.2f: error = %.3g' % (t, error))

Where u_D is the boundary condition.

I am not sure how I can make it works so any advice from you would appreciated.

Thanks in advance

I found out the answer and I would like to share it with you, just in case any one interested. array() was replaced with get_local(). — roby, Sep 15 '18 at 17:32
Thanks @roby. I had similar error, replacing array() with get_local() solved my problem too.. :) — tired and bored dev, Apr 01 '19 at 05:36

score 0 · Answer 1 · answered Jul 25 '19 at 01:20

Per the comment of roby, array() has been replaced by get_local(). So in lines 62-65 of the third fenics tutorial, the following works:

    # Compute error at vertices
    u_e = interpolate(u_D, V)
    error = np.abs(u_e.vector().get_local() - u.vector().get_local()).max()
    print('t = %.2f: error = %.3g' % (t, error))

score 0 · Answer 2 · answered Apr 02 '20 at 08:38

I suggest to employ the conversion from the numpy module, in case you desire a lighter notation:

u_v = numpy.array(u.vector())

Or if you prefer to mantain the structure of the original variable

u_v = u.vector()

I prefer the first one since I am more confortable with a MatLab-style usage of vectors.

Solving the heat equation by FEniCS

2 Answers2