This thesis deals with the development of a strategy for the optimal design of modular systems. A modular system is a system composed by "elementary units" (the modules) where each module is characterised by the same vector of unknowns that can get different values for each module. The optimisation of a modular system is a hard task that can be mathematically formalised as a non-classical optimisation problem. When dealing with this kind of problems the goal is to optimise the system in terms of the number of modules N and in terms of the constitutive parameters of each module. From a mathematical point of view, this means to look for a global optimum configuration of the system over a search space having a variable dimension Nvar. Therefore, we need to conceive a procedure that simultaneously includes the number of modules N among the design variables of the problem and that has to be able to deal with design variables of different nature. Considering all the previous aspects, we decided to develop a numerical strategy in the framework of genetic algorithms (GAs). In this thesis we introduce the concept of species and we develop new genetic operators allowing the reproduction among individuals of different species. In the second part of the thesis, the algorithm has been applied to the solution of some problems: the design of laminates with the least number of layers satisfying some given requirements, the design of stiffened composite structures having the least weight, the design of hybrid elastomer/composite laminates and of laminates with bonded elastomer patches for maximising damping, the identification of the constitutive properties of piezoelectric devices.
