Modeling Hydraulic and Structural Serviceability Hazards of Reinforced Concrete Culverts

This paper presents a data-driven dual-hazard framework for estimating hydraulic and structural serviceability hazard rates in reinforced concrete culverts, which serve both traffic support and water conveyance functions. A network-level dataset of 2,190 reinforced concrete culverts in Ethiopia was used to model two types of serviceability hazards. Hydraulic hazards, driven by sedimentation, overtopping, and scouring, were modeled as functions of discharge depth, barrel slope, and barrel length. Structural hazards, resulting from fatigue cracking and chemical degradation, were modeled using average daily truck traffic, soil pH, and soil salinity. To identify the best-fitting model, six survival models were compared using Akaike Information Criterion, loglikelihood, and residual diagnostics. The Γ-frailty Cox proportional hazards model provided the best fit for both scenarios and effectively captured unobserved heterogeneity. Hydraulic risk decreased significantly with steeper barrel slopes, while discharge depth and barrel length had no significant effects. Structural risk increased with higher 𝐴𝐷𝑇 𝑇 and was mitigated by more alkaline soils. Soil salinity was excluded due to violations of model assumptions. Hazard rate distributions showed greater variability in hydraulic risk across the network, while structural risk was relatively uniform with isolated extremes. Hydraulic failure tended to occur earlier in the lifecycle, particularly in flatter and longer culverts. Structural deterioration was driven by traffic-induced fatigue and moderated by soil alkalinity. Covariatestratified survival curves and relative hazard distributions identified critical subpopulations for targeted inspection. This framework supports optimized maintenance planning by prioritizing hydraulic assessments for flat, long culverts and structural evaluations for high-traffic, acidic sites.