Help with modification of step-15

[Ali Seddiq]

Hello,

I'm  very new to Deal.ii and trying to understand it through modifying step-15. However I get the following error. Although it is descriptive  I'm unable to figure out the problem:

"error: no match for ‘operator*’ (operand types are ‘const dealii::Tensor<1, 2>’ and ‘__gnu_cxx::__alloc_traits<std::allocator<dealii::Vector<double> > >::value_type {aka dealii::Vector<double>}’)
                        * old_solution_values[q] "

I am not sure if I understand it, but my initial guess is that it might be related to the sizes of vectors and tensors (for the case of gradients). I tried to take care of them to the best of my understanding.

My bilinear form:

((grad. phi_i, grad.u*phi_j)+(grad. phi_i, u*grad. phi_j))*del.Uj = (grad. phi_i, u*grad.u)

 I have modified/added the following  lines:

std::vector<Tensor<1, dim>> old_solution_gradients(n_q_points);

std::vector<Vector<double>> old_solution_values(n_q_points, Vector<double>(dim -1));

fe_values.get_function_gradients(present_solution,  old_solution_gradients);

fe_values.get_function_values (present_solution, old_solution_values);

Assembly:

for (unsigned int q = 0; q < n_q_points; ++q)
          {
            for (unsigned int i = 0; i < dofs_per_cell; ++i)
              {
                for (unsigned int j = 0; j < dofs_per_cell; ++j)
                  cell_matrix(i, j) +=
                    (((fe_values.shape_grad(i, q)     
                       *old_solution_gradients[q]
                            * fe_values.shape_value(j, q)) 
                          +                          
                      (fe_values.shape_grad(i, q)
                              * old_solution_values[q]
                              * fe_values.shape_grad(j, q)))  
                          * fe_values.JxW(q));            

                cell_rhs(i) -= (fe_values.shape_grad(i, q)
                        * old_solution_values[q]
                                * old_solution_gradients[q]
                                * fe_values.JxW(q));       
              }
          }

I assume old_solution_values[q] should be the correct variable for "u" but when I use present_solution[q]  instead (in assembly part, for curiosity ), it is compiled but in the running  stage I get the following error message:

"Additional information:
Iterative method reported convergence failure in step 337. The residual in the last step was 22.0585.

This error message can indicate that you have simply not allowed a sufficiently large number of iterations for your iterative solver to converge. This often happens when you increase the size of your problem. In such cases, the last residual will likely still be very small, and you can make the error go away by increasing the allowed number of iterations when setting up the SolverControl object that determines the maximal number of iterations you allow.

The other situation where this error may occur is when your matrix is not invertible (e.g., your matrix has a null-space), or if you try to apply the wrong solver to a matrix (e.g., using CG for a matrix that is not symmetric or not positive definite). In these cases, the residual in the last iteration is likely going to be large."

While the second problem might not be the case could anyone offer some suggestions or explanations on both cases?

Thank you!

Ali

[Wolfgang Bangerth]

Ali,

> I'm  very new to Deal.ii and trying to understand it through modifying
> step-15. However I get the following error. Although it is descriptive  I'm
> unable to figure out the problem:
>
> "error: no match for ‘operator*’ (operand types are ‘const dealii::Tensor<1,
> 2>’ and ‘__gnu_cxx::__alloc_traits<std::allocator<dealii::Vector<double> >
> >::value_type {aka dealii::Vector<double>}’)
>                         * old_solution_values[q] "

What the error message is telling you is that you are trying to take the
product (with operator *) between a Tensor<1,2> and a Vector<double> object,
but that that is not defined. This is not defined even if the Vector has length 2.

I believe that the error comes out of this statement:

>                       (fe_values.shape_grad(i, q)
>                               * old_solution_values[q]
>                               * fe_values.shape_grad(j, q)))

Here you multiply tensor times vector times tensor. It's not clear to me what
that is supposed to represent, but that's at least where the error is from.

Best
W.

--
------------------------------------------------------------------------
Wolfgang Bangerth email: bang...@colostate.edu
www: http://www.math.colostate.edu/~bangerth/