Mitigation of environmental variabilities in damage detection: a comparative study of two semi-supervised approaches

Vibration-based structural health monitoring (VSHM) employs vibration signals asobservables from which inferences are made concerning the integrity of structural systems.More specifically, the premise of this work is to detect damage through changes in a set of fea-tures extracted from the vibration signals. A major challenge in this regard is that false positivesmay arise due to the influence of environmental and operational variabilities (EOVs). Environ-mental variabilities, e.g. shifts in temperature and humidity, introduce changes in mechanicalproperties. These changes are reflected in the vibration response and can reduce the probabilityof detecting damage in a structure. This paper conducts a comparative study between a novelsemi-supervised damage detection approach and a well known cointegration-based scheme todeal with EOVs. The novel approach uses the pattern recognition capability of an artificialneural network (ANN) to learn how EOVs affect a Mahalanobis distance-based damage indexin a reference state. The cointegration-based scheme seeks to mitigate the EOVs by comput-ing stationary linear combinations of non-stationary output response signals. The merits ofthe damage detection methods are examined in the context of a mass-spring system, which isexposed to a simulated temperature field that renders the output response non-stationary. Thesystem is analysed in a reference state and a perturbed state in which damage is emulated byreducing a single spring stiffness by 2%. Both methods are evaluated with the area under thecurve (AUC) for receiver operating characteristic (ROC) and the false alarm rate. The resultsshow that the ANN-based damage detection approach outperforms the cointegration-based onein this particular example.