Vibration-based Structural Identification:Global and Local Dynamic Characteristics
Date Issued
2015
Date
2015
Author(s)
Chan, Chun-Kai
Abstract
In civil engineering, health monitoring and damage detection are mostly carried out using a dense array of sensors. Typically, most methods require global measurements to extract the properties of structures. However, some sensors such as linear variable differential transformers (LVDTs) cannot be used due to in situ limitation. Thus the global deformation remains unknown. Therefore, it is necessary to develop algorithms to identify the physical features such as permanent deformation for structural health monitoring. In this study, signal processing techniques and nonlinear identification methods are used and applied to the responses of test specimens subjected to different level of earthquake excitations. Both modal-based and signal-based system identification and feature extraction techniques are used to study the nonlinear inelastic response of the test specimens using both input and output response data or output only measurement. From the signal-based feature identification method, which include: (1) enhancement of time-frequency analysis of acceleration responses, (2) Hilbert marginal spectrum, (3) estimation of permanent deformation using directly from acceleration response data, (4) instantaneous phase difference, and (5) damage indices. For the modal-based system identification method, structural system parameters are identified, which include: (1) natural frequency, (2) damping ratio, and (3) modes shape. The extracted features are used to compare with the local information measured from the optical sensors. For the local measurements, some physical features are extracted, which include: (1) strain time history, (2) curvature time history, and (3) stress distribution. Two experiments are used to demonstrate the proposed algorithms: a one-story two-bay reinforce concrete frame and a three-story steel frame under weak and strong seismic excitation. The analysis results show that the identified global features are related to the local features and the proposed methods are capable for system identification and damage detection.
Subjects
structural health monitoring
seismic response processing
wavelet packet transform
singular spectrum analysis
multivariate autoregressive model
permanent deformation
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-104-R02521248-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):9ceae706368a03121d5dd1598f412548