Characterization of Steel Bridge Superstructure Deterioration through Data Mining Techniques

Abstract

With a significant number of steel bridges approaching the end of their service life, understanding deterioration characteristics will help bridge stakeholders better prioritize bridge maintenance, repairs, and rehabilitation as well as help with budget planning. This paper applies data mining techniques including logistic regression, decision trees, neural networks, gradient boosting, and support vector machine to the United States’ national bridge inventory to estimate the probability of steel bridge superstructures reaching deficiency. A focused subset of data was created based on the defined scope of the research: design material (steel), type of design (stringer/multibeam or girder), and deck type (cast-in-place concrete). The predictors of the model include age, average daily traffic, design load, maximum span length, owner, location, and structure length. The magnitude that these factors contribute to the likelihood of a steel bridge superstructure’s deficiency was identified. Outcomes of the analysis afford bridge stakeholders the opportunity to better understand the factors that are correlated to steel bridge deterioration as well as provide a means to assess risks of superstructure deficiency for the sake of prioritizing bridge maintenance, repair, and rehabilitation.