α-stable distributions for better performance of ACO in detecting damage on not well spaced frequency systems

Abstract

In this paper, the Ant Colony Optimization (ACO) algorithm is modified through α-stable Lévy variables and applied to the identification of incipient damage in structural components. The main feature of the proposed optimization is an improved ability, which derives from the heavy tails of the stable random variable, to escape from local minima. This aspect is relevant since the objective function used for damage detection may have many local minima which render very challenging the search of the global minimum corresponding to the damage parameter. As the optimization is performed on the structural response and does not require the extraction of modal components, the method is particularly relevant for structures with not-well spaced natural frequencies