Parameter And Load Identification Methods For Nonlinear Structures Based On UKF

Engineering structures commonly perform nonlinearly during earthquakes and other disasters.The parameter identification of the structures is necessary and important for the imergent condition and safety assessment post-earthquake.The external excitation environment is complicate,which makes it difficult to obtain the information for all the external excitations.The response of the structure is determined by the excitation and structural property,and the identification of only one aspect may cause the error in the estimation result.Therefore,the accurate identification of parameter and excitation for nonlinear structure is a challenging subject in the field of civil engineering.In this thesis,identification methods of load and parameter are developed considering the nonlinear property of structures.The improvements to the identification results are discussed.Simplified lumped mass model,finite element model and shaking table testing are utilized to validate the proposed methods and improvements.The main works and results of this thesis can be concluded as follows.1.A load identification method based on the orthogonal decomposition is proposed for nonlinear structures.A regression formulation to determine the order of the decomposition is developed based on the frequency analysis of structural response.Two improvments to the identification results with iteration strategy and time window are used.The method and improvements are validated with the numerical simulations by the cases of periodic load,white noise excitation,earthquake excitation and impact force.2.A load identification method with the augmented state variable in state space is proposed.With this method,the discrete external force in time history is augmented to the state variable as unknowns,which can reduce the number of unknowns in state variable and improve the computational efficiency compared with orthogonal decomposition methods.A practical method for the setting of covariance in the identification process is proposed.The proposed method is validated by the numerical simulation studies of single-degree-of-freedom system and multi-degree-of-freedom system.It is demonstrated by the simulatin studies that the proposed method can accurately identify the external load even considering the measurement noise and model errors.3.Simultaneous identification methods are proposed with the orthogonal decomposition method and direct augmented state variable method,respectively.The simulateous identification method can update both the structural parameter and external excitation.The robustness of the proposed methods is discussed considering measurement noise and model error in different levels.It is illustrated from the results that the direct augmented simultaneous identification is more efficient in computation which is suitable to the complex finite element model based indentification while the orthogonal decomposition based simultaneous identification with more unknowns in state variable but would perform more accurately in identification which is fit for the simplified model based identification.4.A dynamic inverse analysis platform is constructed based on the softwares of OpenSees and MATLAB in which the nonlinear structural identification can be implemented with unscented Kalman filter(UKF)or other identification tools.In OpenSees,the nonlinear finite element model can be built while the UKF or other optimization tool can be implemented with MATLAB.Sensitivity analysis of the response with respect to the structural parameter is conducted to determine the parameter to be identified.The other parameters can be taken as the initial value.A five-storey reinforced concrete building is numerically studied and the identification methods proposed in this thesis are validated in OpenSees-MATLAB simulation platform.5.A shaking table test of a seismically isolated concrete frame is experimentally studied to validate the proposed methods and investigate the damage pattern of isolated buildings.It is demonstrated from the shaking table test that the isolated frame meet the seismic precautionary object.Based on the sensitivity analysis of the structural response with respect to the parameter of the isolated frame,the parameter and ground motion are taken as the unknowns augmented to the state variable.In the platform of OpenSees-MATLAB,the structural parameter and ground motion are estimated with the proposed methods in this thesis.The identified results well coinside with the testing results,which also validates the setting method for the UKF parameters applied in the identification process.