Report No.: CCEER-95-9
Title: System Identification Studies on Cazenovia Creek Overpass
Authors: Bruce Douglas, Emmanuel Maragakis, and Shiping Feng
Date: October, 1995
Prepared for the: National Center for Earthquake Engineering Research
Department of Civil Engineering/258
University of Nevada, Reno
Reno, Nevada 89557
The actual system identification algorithms and programs for southbound Cazenovia Creek bridge are developed. These consist of the following:
- A finite element software package used in the mathematical approximation of the bridge's soil and structural parameters
- An error criterion function whose minimizing yields the best quality of fit between the computed and the measured data sets
- A pattern search algorithm which leads to the optimum solution by minimizing the above error criterion function through modification of the model's parameters within a bounded multi-parameter domain
- An algorithm for the numerical evaluation of the Hessian matrix and its eigenvalues to confirm the existence of any local minimum indicated by the convergence of the above pattern search algorithm.
The verification of all the algorithms and the confirmation of the validity of the methodology above were performed in order to address questions on their efficiency and limitations for the southbound bridge. A "fictitious" problem was constructed by assigning to the bridge model a set of known input variable values and then determining the model's computed or "exact" response corresponding to these input values. In these tests, a zero error (or zero noise) is computed when using the error criterion function. The data set in the fictitious problem consists of an "exact" copy of the model's computed response used in the place of the corresponding measured data sets in the field. The data sets of six, five and four variables were tested and the data set of four variables showed the satisfactory results. The four variables were the soil spring stiffness of abutment, soil spring stiffness of the pier, moments of inertia of the superstructure in both the transverse and vertical directions. Finally, the system identification using the measured data from the field was carried out in this study (Abstract by authors).