| With the building of large yet flexible structures from long-span bridges to high-rise buildings, improved understanding of dynamic characteristics become essential for design and analysis. Because of the complexity of modern structures, full-scale measurement is the only assured way to determine model parameters. One of the most popular techniques is the ambient vibration survey in which ambient sources such as wind and traffic loadings serve as the excitation. Common problems encountered in the ambient vibration survey are that the input excitation is generally unknown and the measurement of structural response is corrupted with ambient noise. Certain analysis assumptions including locally white noise input, have been made. Improper treatment of input excitation, however, may strongly affect the accuracy of system parameters identification, and the excitation itself (e.g., wind excitation on a long-span bridge) may be of interest in the general context of structural dynamics.;System identification techniques have been recently applied to structural identification problems. The Kalman filter, which incorporates both model and measurement uncertainties to achieve optimal estimate of state variables with minimum error covariance matrix, becomes a natural choice to address the problem of simultaneous identification of system and excitation characteristics from measurement alone. An extended Kalman filter algorithm is implemented for structural identification problems formulated in the frequency domain. The new algorithm is formulated for both discrete and continuous systems. Numerical simulations, laboratory experiments as well as field measurements are used to demonstrate the accuracy, reliability and robustness of the proposed method. Parameter studies have also been conducted to reveal the effect of Kalman filter parameters on the performance of the algorithm. |
