Damage detection in frame structures using noisy accelerometers and Damage Load Vectors (DLV)

In the domain of damage detection, many notable methods have been introduced over the past years. Damage Load Vectors (DLV) is among the most powerful methods, which computes a set of load vectors from variations in exibility matrices of a frame in undamaged and damaged conditions. These exibility matrices are derived from acceleration responses of the frame that can be captured using accelerometers. The DLV method then scrutinizes this shift among exibility matrices, which ultimately enables locating the damaged member(s). This study holistically conducted seven experimental tests with seven damage scenarios of a test frame installed on a semi-harmonic shaking table. The DLV method was subsequently employed to locate the damaged members using recorded frame vibration data that were obtained from 'noisy' accelerometers positioned on the frame at eight predefined locations. The Eigen Realization Algorithm (ERA) alongside Pandy's recommendations was adopted herein to facilitate the generation of accurate exibility matrices derived from the noisy accelerometers. The outcome is very encouraging with the accurate identification of damaged members in all seven damage scenarios without any 'positive-false' and 'negative-false' findings. Additionally, there is a decrease (from 0.045 to 0.289) in the accuracy of Weighted Stress Indices (WSI) index when the number of damaged members is increased. © 2020 Sharif University of Technology.