A robust damage-detection technique with environmental variability combining time-series models with principal components

AbstractIn this paper, a novel output-only damage-detection technique based on time-series models for structural health monitoring in the presence of environmental variability and measurement noise is presented. The large amount of data obtained in the form of time-history response is transformed using principal component analysis, in order to reduce the data size and thereby improve the computational efficiency of the proposed algorithm. The time instant of damage is obtained by fitting the acceleration time-history data from the structure using autoregressive (AR) and AR with exogenous inputs time-series prediction models. The probability density functions (PDFs) of damage features obtained from the variances of prediction errors corresponding to references and healthy current data are found to be shifting from each other due to the presence of various uncertainties such as environmental variability and measurement noise. Control limits using novelty index are obtained using the distances of the peaks of the PDF curves in healthy condition and used later for determining the current condition of the structure. Numerical simulation studies have been carried out using a simply supported beam and also validated using an experimental benchmark data corresponding to a three-storey-framed bookshelf structure proposed by Los Alamos National Laboratory. Studies carried out in this paper clearly indicate the efficiency of the proposed algorithm for damage detection in the presence of measurement noise and environmental variability.Keywords:: structural health monitoringdamage detectiontime seriesprincipal component analysisAR–ARX modelsenvironmental variability AcknowledgementsThis paper is published with the permission of the Director, CSIR-Structural Engineering Research Centre (SERC), Chennai.Notes1. Email: lakshmik@serc.res.in2.http://www.lanl.gov/projects/ei/shm/index.shtml