Genetic optimization of stiffened plates without the FE mesh support

This paper deals with the optimization of beam sti eners on plates by varying their positions in order to minimize the maximum de ection of the plate. Genetic algorithms are used. The research strategy consists in substituting, for nite element calculations in the optimization process, an approximate response of a neural network, or an approximate response from a Ritz method. This paper shows how it is possible to deal with more important problems and to envisage more general baseline design for sti ened plates by adding the orientation angle of each sti ener as a design variable. This paper also shows how one can free oneself from the FE mesh, i.e. how one no longer needs the support of the mesh for the sti eners