Uncertainty Analysis and Quantification of the AASHTO Live Load Distribution Model for Moderately Buried–Reinforced Concrete Box Culverts

Abstract

RC box culverts are buried structures prevalent in transportation systems. The existence of backfill brings in additional challenges to estimate load demands in these structures. The AASHTO LRFD specifications present a practical method to model live load distribution within the backfill and onto the top slab of culverts. However, it is widely acknowledged that this AASHTO live load distribution model may considerably overestimate load demands in culverts, resulting in overly conservative designs or unnecessary load postings. The present study analyzes the model uncertainties associated with this AASHTO model in moderately buried RC box culverts. These uncertainties are characterized by the ratio of load effects obtained from two-dimensional frame and three-dimensional finite-element analyses, respectively. Herein, a moderate backfill refers to a depth that yields consistent model biases not sensitive to the backfill depth. Focusing on the bending moment at barrel midspan, Bayesian regression is conducted to quantify the uncertainty in the model bias. The quantified uncertainty is then compared among culverts with different barrel spans as well as the model uncertainty in conventional girder bridges. The results indicate the need for, and provide the basis for, a reassessment of the structural reliability of moderately buried RC box culverts.