A multilevel genetic algorithm for optimization of geometrically nonlinear stiffened composite structures

Abstract.

A multilevel genetic algorithm aiming the global optimization of beam reinforced composite structures with nonlinear geometric behaviour is proposed. A unified approach based on load-displacement control for buckling and first ply failure analysis is adopted. The Newton-Raphson iterative scheme and the arc-length method are used for tracing the equilibrium path and later for updating the critical values. The proposed genetic algorithm performs several sequences of two optimization levels resulting from the decomposition of the original optimization problem. Independent genetic searches are implemented for each level where different fitness functions and sub-populations are considered. The genetic operators selection and crossover supported by an elitist strategy are used while the diversity of the sub-populations is guaranteed based on implicit mutation. A genetic material exchange between levels is performed using clones and so the offspring of matured sub-populations is guaranteed. To improve the efficiency of the multilevel genetic optimization a niche of population is induced after the first stage at both levels.