Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6679563 | Advanced Engineering Informatics | 2018 | 13 Pages |
Abstract
Natural hazards result in ill-conditioned structures with unfavorable damage. To early recognize damage existence, structures can be screened by damage detection methods after a critical hazard event. These damage detection methods are often developed based on a centralized acquiring and computing system that challenges the feasibility of deployment in a large-scale structure. Decentralized damage detection methods alter a single system to multiple subsystems that allow spatially distributing in a structure and yield comparable performance with the centralized approach. In this study, a decentralized damage detection method based on modal prediction errors via multiple banks of Kalman estimators is proposed. First, a sensor network is comprised of multiple subsystems over a structure of which the subsystems have overlapped sensing nodes. These subsystems are individually identified by an input-output frequency-domain system identification method under ambient vibrations. The identified models are then converted into several banks of Kalman estimators, and the estimators generate the estimation of structural modal responses. The prediction errors are calculated from the differentiation between measured and estimated modal responses, and the accumulated standard deviations of modal prediction errors serve as the damage indices for recognizing the damage occurrence, locations, and levels. A numerical example is introduced to demonstrate the proposed method as well as to evaluate the detection effectiveness. Moreover, the proposed method is also experimentally verified by a scaled twin-tower building using shake table testing. The experimental results indicate that the proposed method is quite effective to inform damage of structures in terms of damage occurrence, locations, and levels.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Artificial Intelligence
Authors
Jau-Yu Chou, Chia-Ming Chang,