Damage detection in steel structures using Bayesian calibration techniques

Development and implementation of new and innovative damage detecting techniques is at the forefront of civil engineering advancements. One such technique, presented herein, is formulated to identify damage in steel frames. This study utilizes modal analysis and Bayesian inference model calibration techniques to detect connection damage in steel frames, and is illustrated on a two-bay, two-story scaled steel model. Rotational stiffness coefficients for the connections of the finite element model of the undamaged frame are calibrated through the comparison of model predictions with experimental data. Damage in the form of the removal of bolts at the base connection of one column is then introduced into the experimental frame and features are extracted through dynamic testing. Modal features are used to identify damage, and the calibration of connection spring stiffnesses is used to identify the location of the damage. The shift in natural frequencies of the first four modes indicates damage (i.e. loss of stiffness). The reduction in the calibrated rotational stiffness parameters for the base connection indicates that the damage is present in the base connections of the structure.