COMPDYN 2009 ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering M. Papadrakakis, N.D. Lagaros, M. Fragiadakis (eds.) Rhodes, Greece, 22–24 June 2009 INVESTIGATION OF THE ROBUSTNESS OF PSO ALGORITHMS IN RELIABILITY OPTIMAL DESIGN PROBLEMS Christos K. Dimou 1 , Vlasis K. Koumousis 1 , 1 National Technical University of Athens Zografou Campus 15780 ckdimou@central.ntua.gr vkoum@central.ntua.gr Keywords: Particle Swarm Optimization, Reliability Based Optimal Design, truss structures, Repulsive PSO, craziness operator. Abstract. In this work, the robustness of two variants of the Binary Particle Swarm Optimiza- tion algorithm (ΒPSO) are examined, in the Reliability Based Optimal Design (RBOD) of de- terminate truss structures. Particle swarm optimization (PSO) [1] is a population based stochastic optimization method inspired from the social behavior of birds or insects. In the first variant, the Binary Repulsive PSO (BRPSO), the velocity vector uses information ob- tained from an optimal solution found by the neighbors of the individual, the individual’s memory, its current velocity vector and some “noise”. Repulsive PSO is considered robust in very complex search space with numerous local optima [2]. The second variant is the Binary PSO incorporating the “craziness” operator. When craziness is implemented a subset of the particles is selected, and their position and/or velocity vectors are re-initialized [3]. The ro- bustness of these variants is examined in the RBOD of a 25-bar truss and a 30-bar arch. The results of the investigated optimization schemes, for these two problems, are compared with the results obtained from the BPSO and the Standard Genetic Algorithm (SGA) and the rela- tive performance of all variants is discussed.