An EKF-based Study On The Identification Of Substructure System

For large-scale structures,frequency-domain damage identification methods are often not sensitive to local damage.The direct identification in the time domain usually involves too many unknown quantities resulting in low calculation efficiency and poor convergence.With the usage of limited time-series observations,extended Kalman filter(EKF)technique is capable of identifying structural parameters by treating them as the extended structural states.However,a significant identification difficulty will encounter if such technique is directly used for large-scale structures with many unknowns involved.To circumvent this limitation,the idea of substructure is developed.Nonlinearity exists widely in engineering structures.These nonlinear behaviors are generally described by their nonlinear restoring force(NRF).However,it is often difficult to precisely describe a specific nonlinear phenomenon by a predefined NRF model.Although some non-parametric methods are available for NRF identification,the rationality of the mathematical model should be guaranteed.An EKF-based method was previously proposed by our research team for the identification of structural parameters and unknown inputs.However,in this method,the acceleration responses at the locations of the external excitation should be measured.Moreover,the direct usage of this method for the identification of large-scale structures may encounter the problems of low identification accuracy and poor convergence.Considering the aforementioned problems,the following research work has been conducted and given in this thesis.All of the proposed approaches are validated by several numerical examples.The details are described as follows:(1)With the usage of weighted global iteration procedure,an EKF-based modelfree identification approach is proposed.The NRF is treated as ‘unknown fictitious input’,and identified without any assumptions for the its model.(2)An improved EKF-based approach is proposed for the simultaneous identification of structural parameters,unknown inputs and NRF.As compared with the previously proposed approach,the measurements of acceleration at the locations of external excitation are not required.Moreover,the low frequency drift is avoided by data-fusing the displacement responses.(3)An EKF-based substructural identification approach is proposed for simultaneously identifying the parameters and interface forces of the concerned substructure as well as the unknown excitation applied to it.The unknown interfacial forces are viewed as ‘unknown fictitious input’,and then a revised observation equation is obtained by using projection matrix.(4)The acceleration responses at the interfacial degrees-of-freedom are required in the previous approach mentioned in content(3).Thus,to circumvent this limitation,by combining the technique described in the content(2),an improved EKF-based substructural identification approach is further presented for simultaneously identifying the substructural parameters,interface forces,the unknown excitation and NRF.