Applying a modified Particle Swarm Optimizer to sizing and topological optimization of steel framed structures

As one of the most successful global optimization tools the particle swarm optimizer (PSO) has been designed for most complex, multimodal optimization problems. It belongs to the classes of evolutionary and stochastic algorithms and does not need gradient information derived from the error function. It is now widely applied to various kinds of optimization problems especially of nonlinear, non-differentiable or non-concave types. The original particle swarm optimization algorithm has two obvious shortcomings: firstly it sometimes converges to local minima like the premature problem in Genetic Algorithm (GA); secondly it needs mass computations especially when large amount of particles are adopted. In order to overcome them, a modified guaranteed converged particle swarm algorithm (MGCPSO) is proposed in this paper.