The reliability implications of the upper bounded bridge traffic load model

The safety of bridges is increasingly evaluated using reliability calibration and assessment. An important part of reliability analysis is the accurate probability modeling of all random variables, including the distribution of bridge traffic loading. Recent studies supported by decades of research have shown that the distribution of bridge traffic loading is upper-bounded, reflecting the fact that traffic is a physical process subject to physical limits. Nevertheless, most reliability-based studies continue to adopt unbounded probability models, which implicitly allow for loads beyond what is physically possible. This paper assesses the impact of adopting a physically consistent, upper-bounded traffic load model on bridge safety. First, the typical shape parameter ( ξ ) of the distribution of traffic load is evaluated through a literature review covering 348 bridges in Europe and Australia. Next, both unbounded and upper-bounded traffic load models are compared. Theoretical analyses are conducted to examine the influence of distributional parameters on safety, and the results are verified across the typical parameter space. The implications are demonstrated through code reliability calibration and the reliability assessment of a representative bridge under growing traffic load. It is found that the incorrect use of an unbounded traffic load model leads to the underestimation of safety, whereas the adoption of physically consistent upper-bounded models reveals hidden safety reserves in existing codes and structures. An exploratory study in the Australian context found that the live load partial factor of safety can be decreased by up to 0.25, while in assessment for increased load access, approximately 45% of bridges are found to be safe using the upper-bounded model, compared to only 30% of bridges using the unbounded model. Finally, it is recommended that a shape parameter of ξ = − 0.05 , or equivalently, a centered upper bound of 16.1, be adopted in future probabilistic modeling of bridge traffic loading, unless a more accurate site-specific model is available.