Multi-objective optimisation of geometrically nonlinear composite structures using memetic algorithms

A multi-objective optimisation framework for the sizing and anisotropic topology of geometrically nonlinear beam-stiffened shell structures made of fibre-reinforced plastics is proposed. The objective functions for the proposed robust design approach are the minimisation of mass and the strain energy’s mean value, standard deviation and coefficient of variation. Their trade-offs are optimised while adhering to stress, displacement and buckling constraints. The design variables are the ply angle of each layer of each shell laminate, the width of each layer of each beam laminate and the thickness and material of each layer in the structure. The proposed multi-objective memetic algorithm (MOMA) exploits the synergy between multiple Lamarckian and Baldwinian learning procedures, whose relative performance determines their selection for future offspring generation. Design space decomposition is achieved through the use of multiple sequentially-evolving subpopulations, while a virtual population with age-dominance dual nature updates the Pareto front and accelerates the global search. The concept of species serves as an abstraction of the search space discontinuities due to the presence of integer-type design variables. The MOMA is validated through its application to two bi-objective and two tri-objective optimisation problems.