Robust methods of inclusive outlier analysis for structural health monitoring

Abstract The key novel element of this work is the introduction of robust multivariate statistical methods into the structural health monitoring (SHM) field through use of the minimum covariance determinant estimator (MCD) and the minimum volume enclosing ellipsoid (MVEE). In this paper, robust outlier statistics are investigated, focussed mainly on a high level estimation of the “masking effect” of inclusive outliers, not only for determining the presence or absence of novelty-something that is of fundamental interest but also to examine the normal condition set under the suspicion that it may already include multiple abnormalities. By identifying and detecting variability at an early stage, the prospects of achieving good generalisation and establishing a correct normal condition classifier may be increased. It is critical to highlight that there is no a priori division between the damaged and the undamaged condition data when the algorithms are implemented, offering a significant advantage over other methodologies. In summary, this paper introduces a new scheme for SHM by exploiting robust multivariate outlier statistics in order to investigate if the selected features are free from multiple outliers before such features can be selected for either supervised or unsupervised analysis.