Data-Driven Predictive Modeling of Reinforced Concrete Bridge Deck Performance Considering Chloride-induced Corrosion (RIP2026-00048)

Mehran Mozafarjazi, Qindan Huang

Marquette University

Abstract
Chloride-induced corrosion at rebar is believed to be the primary degradation mechanism for reinforced concrete (RC) bridge deck damage, and the damage degree determines both the timing and extent of required patching repairs. However, the literature shows that the models that solely based on chloride-induced corrosion mechanism fails to predict bridge deck deterioration extent.
The research adopts data-driven approaches which utilize not only chloride-related quantities but also other influencing factors including bridge features, traffic volume, and environmental factors to capture the complex degradation processes involved in the RC deck deterioration conditions. In particular, the patching quantities over more than 300 bridges recorded by Minnesota department of transportation (MnDOT) are used as the damage level, and the possible explanatory variables include chloride-related quantities derived based on chloride diffusion models, structural variables extracted from the National Bridge Inventory (NBI), and environmental variables obtained from long term bridge programming (LTBP). Different statistical and machine-learning methods including logistic regression, multivariable linear regression, machine learning approaches, and gamma-distribution based models are explored using training and testing data sets, and model performances are evaluated and compared.
The research shows that chloride related quantities such as chloride level at time of patching play a significant role in predicting deck damage levels, but their impact is modulated by multiple accompanying factors. Variables (such as deck age, rebar type, cover depth, bridge features, and weather conditions) significantly influence deck deterioration progression and repair needs. These models consistently revealed that the deck patching quantities arise from a combined corrosion–structural–environmental interaction rather than from chloride penetration alone.
The developed models provide reliable deck degradation predictions, which enable state DOT to optimize maintenance planning while maximizing infrastructure lifespan and environmental sustainability. The models also provide valuable insights into the quantitative impact of various actions/design options on RC deck degradation.