Modeling Techniques for Evaluating the Redundancy of Composite Steel Tub Girder Bridges Subjected to Brittle Fracture

Abstract Twin-girder bridges have historically been classified as nonredundant due to the perceived inability of the bridge to support the required vehicular loading if one of the girders were to fail. Although this classification is likely appropriate for twin I-girder bridges, twin-tub girder bridges have been shown to possess significant redundancy even with the failure of a girder. Advances from previous studies have induced the Federal Highway Administration to permit redundancy to be demonstrated through analysis in the failed state. The analysis requires a simulated fracture in a girder with applied loading consisting of the bridge self-weight and live load. Bridge models in the failed state require a relatively fine mesh and complex loading application, demanding significant computational resources and analysis run times that can require weeks to complete even on a powerful computer system. This paper outlines the development and validation of computational modeling methods for evaluating the redundancy of twin-tub girder bridges subjected to brittle fracture. The computational framework employs implicit and explicit algorithms to capture the respective cases of the construction sequence, fracturing of a girder, and application of vehicular live loads. While an implicit algorithm is advantageous for evaluating behavior during the construction phase, the adoption of an explicit algorithm is indispensable for analyses involving fracture, since implicit algorithms are not well-tailored for cases with severe material and geometric nonlinearities. Ensuring accuracy in both modeling and analysis techniques requires validation, which was completed using full-scale experimental results from a twin-tub girder bridge fracture test. Redundancy analyses are demonstrated on three existing bridges that include both straight and curved geometries. Modeling efficiency is improved and demonstrated by varying multiple parameters. In particular, results elucidate the efficacy of mass scaling and damping in reducing the computational effort. The techniques and recommendations dramatically improve the efficiency of the analyses.