Structural Damage Detection via a Combination of Pso and Bayesian Reliability Analysis

Different kind of methodologies have been presented in the last decades and achieved broad applications in structural damage detection (SDD). Particle swarm optimization (PSO) based algorithms have been confirmed to be effective for SDD. The method avoids inversion computation which is prone to be ill-posed or ill-conditioning. However, the accuracy of optimization algorithm is affected by its randomness although it is the theoretical basis of algorithm. Repeated calculations are often performed to gain an average value of SDD results, but the computing cost increases simultaneously. In this study, a novel two-step SDD method is proposed via a combination of PSO and Bayesian reliability analysis. It consists of two major steps, i.e., SDD and Bayesian reliability analysis. Firstly, SDD on structures is achieved by the PSO – improved Nelder-Mead method (PSO-INM). a new objective function, so-called multi-sample objective function, is proposed based on Bayesian theory. The Bayesian reliability analysis is then performed for a further analysis, the most likely damaged elements are distinguished from the spurious ones. Finally, some numerical simulations on SDD of a 2-storey rigid frame are used to assess the effectiveness of the proposed method, some related issues are discussed as well.