High-Dimensional Model Approach for Stochastic Response of Multispan Box Girder Bridges

Abstract

Seismic safety assessment of reinforced concrete box girder bridges has received considerable attention as they are increasingly popular in modern highway systems. However, the detailed seismic assessment of box girder bridges requires a consideration of the material nonlinearity and uncertainty in geometry, material properties, and loading. This paper presents the development of metamodels for selected seismic response parameters of a box girder bridge considering the uncertain input variables using high-dimensional model representation, a relatively new nonstatistical metamodel-based approach. The seismic responses at the sampling points of high-dimensional model representation are evaluated through detailed finite-element analysis. The developed metamodels are further utilized to generate seismic fragility curves that are found to be much simpler than traditional fragility analysis. The results from this approach are found to agree with that of other popular response surface methods (using Central Composite Design and Box Behnken Design) with significantly fewer simulations. High-dimensional model representation (HDMR) provides a failure metamodel that includes all related random variables, but at the same time significantly reduces the computational effort of fragility analyses.