Damage detection under environmental and operational variability using the cointegration technique
Marco Pirrò, Sérgio Pereira, Carmelo Gentile, Filipe Magalhães, Álvaro Cunha
Abstract
Within Structural Health Monitoring (SHM) framework based on Operational Modal Analysis (OMA), novelty analysis aims to detect structural anomalies by inspecting the time evolution of the modal characteristics of a structure subjected to ambient (unknown) excitations. However, in real applications, modal parameters are affected not only by structural changes, but also by environmental and operational variability (EOV), thus it is necessary to remove EOV effects before approaching any novelty analysis. The paper describes the vibration-based SHM of Baixo Sabor dam (Portugal) involving the application of automated tools for modal identification during the first 2 years of monitoring. Within this context, two different novelty analysis procedures are adopted: (a) the well-known Hotelling T 2 control chart based on PCA regression and (b) the less diffused Cointegration technique. Furthermore, several damage scenarios are simulated in a FE model of the dam, mainly with the purpose to compare the ability in the damage detection provided by the two different procedures of novelty analysis. Compared to traditional methods, Cointegration relies on more general operations and also permits to considerably reduce the training period where the structure is supposed to be in an healthy state. • Novelty analysis based on PCA and Cointegration is presented. • Selected results of a 2-year dynamic monitoring on a concrete dam are presented. • EOVs are mainly due to reservoir water level and temperature. • Numerically simulated damages are introduced for damage detection purposes. • Cointegration method provides advantages with respect to PCA-based novelty analysis.