Report No.: CCEER-89-2-A

Title: Dynamic Response Analysis of the Dominion Road Bridge Test Data

Authors: James Richardson and Bruce Douglas

Date: March 1989

Performing Organization:
Department of Civil Engineering/258
University of Nevada, Reno
Reno, NV 89557


The three-dimensional dynamic response of an existing highway overpass is analyzed and compared with a computer model of the bridge. Well-defined lateral, vertical and torsional modes of vibration are identified and presented. A computer-implemented bridge response model is fit to the experimental frequencies and mode shapes using a system identification procedure. The final model agreed well with the measured response; it also accurately predicted responses not included in the system identification procedure.

The work is based on the data collected from the full scale bridge test of the Dominion Road Overpass in Auckland, New Zealand. The Dominion Road Overpass is a ten-span. reinforced concrete box girder bridge which follows a 70 degree curve along half of its length. In the bridge test, large-amplitude static loads were applied laterally to the bridge superstructure, (using the snap-back and quick-release method), and the free vibration response was measured. Accelerations measured on the bridge superstructure were of similar magnitudes as those expected due to a moderate magnitude earthquake. Time histories for as many as six acceleration components (three translational and three rotational) were measured on both the superstructure and the foundations of the bridge.

Natural frequencies and three-dimensional mode shapes of the bridge are extracted from the time history data using unique applications of traditional Fourier Transform methods. Also, a method to separate the responses of vibrations modes closely spaced in frequency is developed.

Finally, a computer model of the bridge response is constructed. The values of the boundary element springs used to represent the foundations were calculated based on geotechnical methods. The model parameters most affecting the model response were identified and their values adjusted in order to bring the model response into reasonable agreement with the measured response. The optimal values of these model parameters were determined using a system identification algorithm (Abstract by authors).