Element level time domain system identification techniques with unknown input information

A finite element based linear time domain system identification algorithm is proposed to estimate the stiffness and damping coefficients of structures at the element level using response data alone without using information of excitation measurements. The unknown input excitation could be applied at any location of the structure including at the ground level representing the seismic excitation. The proposed method is an Iterative Least-Squares with Unknown Input (ILS-UI) procedure. The element-level structural parameters can be identified directly by using proposed ILS-UI procedure. No information of the modal properties is required. The efficiency and robustness of the proposed algorithm is illustrated by numerical examples. For verification purposes, both noise-free and noise-included output responses are considered in numerical examples. The applications of three types of structures, i.e., shear-type buildings, trusses, and frames, are considered in this dissertation. In all examples, the identified results indicate that the proposed ILS-UI method identified the structural parameters very well. For the successful implementation of the proposed method, only a small number of sampling time points are required, and a long time duration of responses is not necessary.;For a large system, since it is practically impossible to measure responses at every dynamic degree of freedom, the absence of some observation points of responses and its effect on the proposed system identification technique must be studied. Based on the above ILS-UI procedure, a new technique combined with the Kalman filter technique is developed to identify all element-level structural parameters using measuring responses at several optimal locations only. The optimal numbers and locations of measurement points required to identify uniquely the system using this proposed ILS-EKF-UI technique are determined. Again, numerical examples with two special cases are used to illustrate the applications of this new technique. The results of numerical examples indicate that this new system identification technique is very economical, simple, and robust, since the input is not required to be measured and only several observations are required.