Particle Swarm Optimization for damage identification in beam-like structures

Abstract

The main objectives of Structural Health Monitoring (SHM) are the characterization and the assessment of the health condition of structural systems. Combined with appropriate Damage Identification (DI) strategies, SHM aims to provide reliable information about the localization and quantification of the structural damage by using an inverse formulation approach, with the damage parameters being estimated from parametric changes in dynamic properties. Mathematically, an inverse problem consists of the optimization of a function which represents the "distance" between the experimental and the numerically-simulated features of the system. Such process requires the development of a mock-up numerical model fairly representative of the system and iteratively updated until a response, as close as possible to the experimental one, is provided. The minimization of the difference between measured and predicted features' values is the objective function, whose global minimum corresponds to the best adjustment of the model variables. Metaheuristics represent a large class of global methods for optimization purposes able to outperformtraditional methods in the following aspects: ease of implementation, time consumption, suitability for non-linear phenomena, black-box and high-dimensional problems. The present paper analyses, through a numerical experimentation approach, the suitability of one of the best-known metaheuristics, i.e. the Particle Swarm Optimization (PSO) algorithm, for DI of beam-like structures. Modal properties are used to define the objective function and various algorithm instances are tested across different problem instances to assess robustness and influence of the algorithm parameters.