Solving process: principles


The finite element numerical solution of a problem is always reduced to resolving a system of equations (linear or not, symmetrical or not). These systems are generally characterized by big size matrices.

We will not intend to give here a detailed presentation of the solving methods. Only the elements necessary to understand the principle of the solving process, and a description of the various solvers proposed in Flux are presented in this section.

Parallel computing, enabling to improve the solving process executing tasks simultaneously with multicore postprocessors, is presented.

Error criterion, enabling to identify areas where the mesh quality can be improved, is also presented.