NON-STRUCTURED GRID REFINEMENT USING GENETIC ALGORITHMS

Resumo

Refinement of non-structured grids is an important topic related to the accuracy of numerical solutions
of the most used methods to solve engineering problems, such as the finite elements method and the finite volumes
method. Considering this fact, the current paper presents a methodology to optimize triangular non-structured
meshes using a floating point genetic algorithm to choose the best position for grid vertices. Such position is
based on a fitness function evaluated to each internal grid vertex; it is based on the average of the quality of
volumes to which the vertex belongs to. The algorithm is executed to a grid firstly generated with the Delaunay’s
triangularization and then disturbed to obtain a non-optimal grid, which needs to be improved. Results show that
the quality parameters were improved, although the method needs high computational efforts.