Identifying Environmental- and Operational-Insensitive Damage Features

Structural health monitoring (SHM) detects damage in structures using online, in situ, monitoring. In practice, structures are affected by factors that make it difficult to discern damage from environmental and operational (E&O) variability. Therefore, an improved process for identifying features that are sensitive to damage while insensitive to E&O effects is needed. In this study a SHM approach that utilizes causality metrics is proposed. The assumption is made that under E&O variability, the structure will remain linear while in its undamaged state, and linear changes in the structural properties will not affect the causal relations between sensor readings. Furthermore, the structure will exhibit an observable nonlinear response when damage is introduced to the system and that response will cause changes detected by the three proposed measures of causality: granger causality, mutual information, and coherence. This paper aims to evaluate the candidacy of causality measures in detecting nonlinearities introduced by damage while remaining insensitive to linear changes to the structural properties caused by E&O variability. This approach is evaluated using a numerical model simulating damage and E&O variability, an experimental dataset looking at the vibrational response of a concrete column with the introduction of damage, and an experimental dataset looking at the vibrational response of a bridge in different environmental conditions. Each measure was found to suggest an ability to detect damage and remain insensitive to E&O variability.