Quantifying the Effects of Modeling Simplifications for Structural Identification of Bridges

Abstract

Several long-span, prestressed, segmental box girder bridges were built in the early 1980s and many of them are affected by long-term residual deformations. Although some models have been proposed to describe their structural behavior, several uncertainties remain. This paper examines the effects of errors introduced by model simplifications on predicted values. The results are used to improve the estimation of parameter values using model-based data-interpretation strategies. The procedure is illustrated for the Grand-Mere Bridge located in Canada. This bridge is affected by excessive long-term vertical displacements. Model simplifications such as in its degree of complexity are found to have an important influence on prediction errors. Representing these errors by zero-mean independent Gaussian noise does not adequately describe the relationships among the errors observed in this case study. Estimated errors are used in the interpretation of the ambient vibration acceleration data recorded on the structure. The interpretation approach employed is based on error-domain model falsification. This study provides ranges of parameter values that can be used subsequently to characterize more accurately aspects such as long-term creep and shrinkage behavior.