Local modal filters for automated data-based damage localization using ambient vibrations

Abstract

a b s t r a c tThe motivation of the paper is to develop a fully automated data-based technique for damage localization using in-service ambient vibrations. The idea is an extension of the modal filtering technique previously developed for damage detection. A very large network of dynamic strain sensors is deployed on the structure to be monitored and split into several independent local sensor networks. Simple and fast signal processing techniques are coupled to statistical control charts for efficient and fully automated damage localization. The efficiency of the method is demonstrated using time-domain simulated data on a simply supported beam and a three-dimensional bridge structure. The method is able to detect and locate very small damages (2% stiffness reduction in an area corresponding to 1/100th of the length of the structure) even in the presence of noise on the measurements and variability of the baseline structure.

Experimental Localization of Small Damages Using Modal Filters

“…More explanation on the rank deficiency can be found in [5]. The alternative adopted by the authors to regularize the inversion of [C] T consists of applying a singular value decomposition (SVD) of [C] T : (2) is given by…”

“…Previous works pointed out that the filtering is the most efficient if only two mode shapes (nm = 2) are taken into account (both in numerical [5] and experimental tests. )…”

“…This indicator should be sensitive to the shape of the function (appearance of spurious peak), but not to its level (change of the level of the excitation). Previous works considered an indicator called peak indicator [5,6], but recent researches pointed out that this indicator presents some drawbacks. On the other hand, using the moment of order 4 which quantifies the flattening of a probability density function is a very interesting alternative.…”

“…Recently, a very simple non-model based output-only method using modal filters has been developed for damage detection [1]. By considering strain measurements as well as control charts [9], the idea has next been extended successfully to automated damage localization in many numerical studies [5]. However, with the exception of the experimental study led by Mendrok et al using accelerations [8], the modal filters technique using strain sensors to locate damages has never been tested experimentally so far.…”

Automated data-based damage localization under ambient vibration using local modal filters and dynamic strain measurements: Experimental applications

“…In order to avoid the high computational costs associated to model-based methods and to take into account the usual lack of training data from damaged civil engineering structures, a new fully automated data-based unsupervised technique for damage localization was developed in [22]. The method is based on an automated feature extraction process using the so-called modal filters [23] which is computationally very cheap.…”