A SWARM BASED MEMETIC EVOLUTIONARY ALGORITHM FOR MULTI-OBJECTIVE OPTIMIZATION OF LARGE STRUCTURES

In this paper a swarm based version of the M-PAES is proposed to solve structural design optimization problems with high number of design variables. The M-PAES is a well-known memetic evolutionary algorithm for multi-objective optimization. In the present algorithm, all the produced individuals in each local search phase of the M-PAES are stored in a swarm and then search process is conducted using the provided information by members of the swarm. In other words, the obtained fitness function of all the individuals in a swarm, gives an approximation of the gradient of that limited area of the search space in which the swarm is located. Then this vector is utilized to guide the next cycle of individual creation in the process of local search. Additionally, in order to escape from getting stuck in local Pareto optimums, a combination of random and directed search is employed. Furthermore, in the present algorithm, the crossover process is performed by differential evolution (DE) algorithm, where each new individual gains its characteristics from three parents, the feature that provides a powerful global search ability for the algorithm. Finally, we present some numerical results on three sets of problems, indicating the strength of the idea employed in the proposed algorithm.