This article introduces a new methodology for the detection of structural changes using a statistical data-driven modeling approach by means of a distributed piezoelectric active sensor network at different actuation phases. The three main features that characterize the proposed methodology are (a) the nature of the data used in the test since vectors of principal component analysis projections are used instead of the entire measured response of the structure or the coefficients of an AutoRegressive model, (b) the number of data used since the test is based on two random samples instead of some characteristic indicators, and (c) the samples come from a multidimensional variable and therefore a test for the plausibility of a value for a normal population mean vector is performed. The framework of multivariate statistical inference is used with the objective of the classification of structures in healthy or damaged. The novel scheme for damage detection presented in this article —based on multivariate inference over the principal component analysis projections of the raw data—is applied, validated, and tested on a small aluminum plate. The results show that the presented methodology is able to accurately detect damages, that is, for each actuation phase, a unique and reliable damage detection indicator is obtained no matter the number of sensors and/or actuators. It is worth noting that a major contribution of this article is that there exists an entire range of significance levels where the multivariate statistical inference is able to offer a correct decision although all of the univariate tests make a wrong decision.

Peer Reviewed