Structural Nonlinear Behavior Identification Based On Kalman Filter And Partial Measurements

Structural damage identification and performance evaluation based on structural dynamic responses play key roles in structural health monitoring.Structural nonlinear behavior widely exists in engineering structures,especially when structural damage occurs.Identification on structural nonlinear behavior is helpful for the direct description on the initiation and development of structural damages and the quantitative evaluation of energy consumption of structures during vibration.Due to the uncertainty of the parametric model of nonlinear restoring force(NRF)of structures and mass and the unavailability of dynamic response measurement at all degrees of freedom(DOFs),the development of structural nonlinear behavior identification approach without the use of parametric NRF model with limited dynamic measurements is important.In this study,by combining Kalman filtering and other algorithm theories,structural nonlinear behavior identification approaches using partial dynamic response observation with unknown parametric model and mass are proposed and validated numerically and experimentally.The main contents of the research are as follows:(1)Kalman filtering,identification based on least square estimation(LSE)and double Chebyshev polynomials model(DCP)are introduced in approach deduction,and their advantages have been analyzed.(2)By combining extended Kalman filtering(EFK)and parameter identification approach based on LSE,structural NRF identification with unknown mass and incomplete response measurement is carried out.Numerical simulation validation with nonlinear 4-DOF chain structures equipped with Duffing oscillator and Bouc-Wen magnetorheological damper to mimic NRF is carried out.The effect of different initial mass errors and measurement noise is considered.(3)By combining the unscented Kalman Filter(UKF)suitable for strong nonlinear system with equivalent linear theory and double Chebyshev polynomials model,a free-model approach based on partial observation is proposed for nonlinear structure identification with unknown mass.A nonlinear numerical model equipped with Bingham MR damper is used to validate the performance of the proposed approach considering the effect of different excitation location,dynamic response observation locations and noise levels.(4)In order to experimentally validate the performance of the proposed approaches,dynamic test on a four-layer steel frame nonlinear structure model with a SMA damper is conducted.The feasibility of model-free NRF identification approach is validated by comparing the dynamic response and the restoring force identification results with the measurements.