Fig. (4) Four FE models with growing computational effort due to mesh refinement: going from (a) to (d) models.