Structural damage detection: comparison between GA and PSO techniques

Abstract

This study compares optimization techniques of Particle Swarm Optimization (PSO) with Genetic Algorithms (GA), both techniques employed for the implementation of a system intended to detect and diagnose structures failures by using a modal response. Different noisy and noiseless damage events (simple and multiple damages) enable the simulation of actual conditions of a beam structure and a framework structure, which are used to determine detection and diagnosis behaviors of proposed systems. Additionally, both proposed systems are evaluated by modifying the input amount, that is to say, the number of vibration modes. The structure dynamic response, under normal or failure condition, is obtained by using the OPENSEES® free-download tool, aswell as the optimizationalgorithms PSO and GA, both implemented under Matlab®® environment. Behavior comparison between the two techniques, as far as detection and diagnosis abilities are concerned, are carried out for a supported beam segmented in ten sections and a framework structure with 13 elements. Experimental results showed the effectiveness and robustness of proposed systems for the determination of system conditions, at different noise environmental levels and with different amount of inputs. However, the performance varies in accordance with the controlled system and the objective function, which are employed.