Nonparametric Identification Of Nonlinear Restoring Force For MDOF Structures Under Partial Measurements

It is the major task of structural health monitoring(SHM)and post-disaster condition evaluation for engineering structures to evaluate their safety,residual load-carrying capacity and residual life after the application of strong dynamic loads based on the damage location and severity identification.The structural restoring force is the direct representation of the initiation and development of structural damage and nonlinear behavior and can be used to quantitatively determine the energy dissipation of structures during the vibration excited by dynamic loadings.Based on the identification result of structural restoring force of different structural members at different time instants,the damage initiation and the development failure patterns at different time instants can be identified,which is helpful for the understanding of the structural failure pattern transformation during the application of dynamic loadings.However,due to the diversity of different construction materials and structural forms in various civil engineering structures,it is always a hard task to describe the nonlinear restoring force(NRF)of an engineering structure with an accurate parametric model in prior in practice.Meanwhile,the limited availability of structural dynamic response measurement at all degrees of freedom(DOFs),excitation and mass information lead to difficulties in the identification of nonlinear behavior of engineering structure.Therefore,it is of great significance to develop a general structural damage identification method using only partial dynamic measurement such as acceleration in a nonparametric way where no parametric numerical model for the nonlinear restoring force is required.Therefore,in this study,a general nonparametric identification method for structural NRF and mass is developed using acceleration response measurement at limited DOFs and excitation information.The developed approach provides an effective method for damage identification ofengineering structures,especially for post-disaster evaluation where the nonlinear behavior of structures should be considered.The main contents of this study are as follows:1.Based on an extended Kalman filter with memory fading technique and global iteration method(EKF-MF-GI)and power series polynomial(PSP),a nonparametric identification method of NRF and mass is proposed under unknown mass and partial acceleration response measurements.After explaining the theoretical background and the proposed identification approach,numerical simulation verification is carried out using multi-DOFs nonlinear chain-like structure involving piecewise linear and Magneto-Rheological damper(MR).The identified NRF and mass are compared with the theoretical values and the effects of different initial mass errors and measurement noise on the different results are discussed to verify the effectiveness and robustness of the developed approach.2.Following the numerical simulation verification,the feasibility of the proposed approach for structural NRF and mass identification was illustrated with a four-story steel frame structure equipped with a MR damper employed to mimic structural nonlinearity using limited acceleration measurements.The comparison of the identification restoring force provided by the MR damper and mass of the structure are compared with the measurement and that identified with free vibration test.Results show that the proposed approach can identify the restoring force of MR damper and the mass of the test model with acceptable accuracy.3.Aiming at the post-earthquake structural damage identification problem,a nonparametric nonlinear behavior,earthquake and mass identification method is proposed for multi-DOFs structures excited with unknown earthquake using limited dynamic response measurements.The proposed approach is numerically with a 7-DOF chain-like structure equipped with different number of MR dampers.The identified MR damper force,earthquake acceleration and the mass are compared with the theoretical values and the effects of the measurement noises on the identification results and theconvergence of the proposed approach are discussed.