ARX model-based damage sensitive features for structural damage localization using output-only measurements

The study proposes a set of four ARX model (autoregressive model with exogenous input) based damage sensitive features (DSFs) for structural damage detection and localization using the dynamic responses of structures, where the information regarding the input excitation may not be available. In the proposed framework, one of the output responses of a multi-degree-of-freedom system is assumed as the input and the rest are considered as the output. The features are based on ARX model coefficients, Kolmogorov-Smirnov (KS) test statistical distance, and the model residual error. At first, a mathematical formulation is provided to establish the relation between the change in ARX model coefficients and the normalized stiffness of a structure. KS test parameters are then described to show the sensitivity of statistical distance of ARX model residual error with the damage location. The efficiency of the proposed set of DSFs is evaluated by conducting numerical studies involving a shear building and a steel moment-resisting frame. To simulate the damage scenarios in these structures, stiffness degradation of different elements is considered. It is observed from this study that the proposed set of DSFs is good indicator for damage location even in the presence of damping, multiple damages, noise, and parametric uncertainties. The performance of these DSFs is compared with mode shape curvature-based approach for damage localization. An experimental study has also been conducted on a three-dimensional six-storey steel moment frame to understand the performance of these DSFs under real measurement conditions. It has been observed that the proposed set of DSFs can satisfactorily localize damage in the structure.