Structural Time-Varying Parameter Identification Based On Strong Tracking Unscented Kalman Filter

Recently, Structural health monitoring technology has gained rapid development along with the growing demands for the safety of the buildings. As one of the important contents of structural health monitoring, structural parameter identification could know clearly the structural parameters and their change under the work conditions and provide a basis for structural damage identification and state evaluation. During construction and operation period, structures will show nonlinear and time-varying characteristic because of the extreme load such as earthquake and strong wind. As a consequence, structural time-varying parameter identification becomes a research focus.Unscented Kalman filter(UKF) is an effective method for nonlinear structural parameter identification, and it has higher identification accuracy than extended Kalman filter. But unscented Kalman filter is effective only in the case of timeinvariant parameters, once structural parameters change, the identification results will have a very low accuracy, or even diverge. To solve the problems mentioned above, strong tracking principle is introduced to unscented Kalman filter, and a strong tracking unscented Kalman filter(STUKF) is put forward to identify the timevarying parameters of structural system. An intensive study is conducted to the proposed STUKF, and the main research contents are as follows:The limitation of extended Kalman filter(EKF) in nonlinear structural parameter identification is introduced. Faced with the questions above, the unscented Kalman filter is studied and its identification process is introduced. A simulation example shows that the UKF could identify the nonlinear structural parameters effectively, and it has higher identification accuracy and stability than the EKF.To solve the problems that the UKF could not identify the structural time-varying physical parameters, the strong tracking principle is introduced to the unscented Kalman filter, thus a STUKF method is put forward. In this method, a fading factor matrix which is suitable for structural systems is constructed, and the solution of fading factors is given based on the orthogonality principle. The equivalent expressions of the Jacobian matrices are derived and used to replace the old ones, to avoid the complex calculation of the Jacobian matrices of the nonlinear functions. The identification process of the STUKF is given, and numerical simulations are carried out to verify the validity of the proposed STUKF method. Simulation examples include two-DOF and five-DOF linear storey shear models, a SDOF nonlinear hysteretic model, a three-DOF nonlinear model with one hysteretic storey, and a four-DOF nonlinear model which all storeys are hysteretic nonlinear. The displacements, velocities, hysteretic curves, the structural parameters and their changes are identified with the STUKF method. In the simulation examples, different parameters, different locations, different magnitude of abrupt changes are considered, and the influence of noise level to the effect of time-varying parameter identification of linear and nonlinear structures are studied. Numerical simulation results show that the proposed method could track the change of the structural parameters effectively, and it has good anti-noise performance.Shaking table test of a two-story steel frame is conducted. In this experiment, the lumped masses of the two storeys are set as the time-varying parameters. The change of each lumped mass is identified based on the STUKF, only using the measured excitation and acceleration response of the model. This experiment results show the validity and reliability of the STUKF in structural time-varying parameter identification.