A two-stage approach for structural damage detection using a damage localization indicator vector and cluster particle swarm optimization

Abstract

A two-stage approach based on the finite element model is proposed for structural damage detection. In the first stage, the damage identification problem is transformed into an -minimization issue related to the natural-frequencies-based sensitivity matrix, which can be quickly solved using the alternating directional multiplication method. Based on this, a damage localization indicator vector (DLIV) is proposed for dimensionality reduction. In the second stage, a particle swarm optimization (PSO) algorithm with four-cluster topology, namely Cluster PSO, is employed to identify the location and severity of the actual damage after dimensionality reduction. Therefore, the proposed two-stage approach can be denoted as the DLIV-based Cluster PSO. In the numerical simulations, the performance of the proposed DLIV-based Cluster PSO is evaluated for single and multiple damage detection of a cantilever beam. The simulation results demonstrate that the proposed DLIV-based Cluster PSO approach provides faster convergence, less computational time and better noise tolerance.