Subspace-based damage detection in engineering structures considering reference uncertainties and temperature effects

Automated vibration-based damage detection is of increasing interest for structural health monitoring of engineering structures. In this context, stochastic subspace-based damage detection (SSDD) compares measurements from a testing state to a data-driven reference model in a statistical framework. In this thesis theoretical developments have been proposed to improve the robustness of SSDD for realistic applications conditions. First, a statistical test has been proposed considering the statistical uncertainties about the model obtained from the reference data. This leads to a precise description of the test's distribution properties and damage detection thresholds. Second, an approach has been developed to account for environmental effects in SSDD. Based on reference measurements at few different environmental conditions, a test is derived with respect to an adequate interpolated reference. The proposed methods are validated in numerical simulations and applied to experimental data from the laboratory and outdoor structures.