Please use this identifier to cite or link to this item: https://scholarbank.nus.edu.sg/handle/10635/183150
Title: APPLICATION OF SYSTEM IDENTIFICATION TECHNIQUES FOR STRUCTURAL APPRAISAL
Authors: SEE LIN MING
Issue Date: 1994
Citation: SEE LIN MING (1994). APPLICATION OF SYSTEM IDENTIFICATION TECHNIQUES FOR STRUCTURAL APPRAISAL. ScholarBank@NUS Repository.
Abstract: This thesis presents several system identification techniques developed for the parameter determination of structures which have fairly high degrees of freedom. First, a substructural identification method is formulated. By decomposing the structural system into several smaller subsystems (substructures) for identification purpose, the numerical efficiency in terms of convergence, speed and accuracy is greatly improved compared to the identification of the complete system. The extended Kalman filtering (EKF) is applied to determine the parameters of each substructure in the time domain based on some measured responses due to dynamic excitations. Two versions of the substructural identification method are derived: one with and the other without overlapping members. The superiority of the substructural identification techniques over the conventional method is illustrated by numerical simulation examples. The effects of input and output noise and response measurement locations are investigated. From computational point of view, the substructural identification method is a highly concurrent procedure. It is thus advantageous to employ the relatively new concept of parallel processing to further improve the computational speed. The fine grained and coarse grained techniques are implemented on a PC-base transputer system to convert a sequential algorithm to a parallel one. The coarse grained approach exploits natural parallelism of substructures at a system level, whereas the fine grained approach exploits natural parallelism is matrix operations involved in the EKF. The algorithm efficiency achieved for the whole identification process is about 70% and about 90% for the fine grained approach and the coarse grained approach, respectively. As for the local damage detection of buildings, an "improved condensation" method is developed to quantify the changes in storey stiffness of multi-storey frame buildings. The complete building model is reduced to a condensed model with significantly fewer degrees by static and kinematic condensations. The ensuing modelling error is minimized via a simple remedial model, of which the parameters are identified by the EKF. The modelling error is successively reduced by computing stiffness correction factors from the remedial model and then updating the complete model. Upon convergence, the improved condensation method yields integrity indices which reflect the soundness of the storey stiffnesses. The efficacy of this method is illustrated numerically by a twelve-storey plane frame building with various noise levels and substantiated experimentally by laboratory tests of a six-storey steel frame with artificially introduced column damages. Taking into consideration the torsional response, the method is extended to study three-dimensional asymmetrical buildings. A numerical simulation example of three-storey building which has a total of 270 degrees of freedom is presented. Lastly, a numerical procedure is proposed to determine the uncertainties of identified parameters. An adaptive filter is derived and incorporated in the EKF to determine the system noise covariance by continuous feedback from the residual to the Kalman gain. As such, a statistically consistent estimate of the error covariance matrix is obtained, thereby providing a means to gauge the reliability of the identified parameters. Three numerical examples, including one involving the use of the improved condensation method, are presented to illustrate the application potential of the adaptive procedure in the safety evaluation of structures.
URI: https://scholarbank.nus.edu.sg/handle/10635/183150
Appears in Collections:Ph.D Theses (Restricted)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
b19530006.pdf15.82 MBAdobe PDF

RESTRICTED

NoneLog In

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.