Bridge Risk Index for Freight Corridor Resilience: A Non-Parametric Machine Learning and Threat Modeling Approach

Bridges are critical nodes in freight networks, yet limited funding prevents agencies from maintaining all structures in good condition. This creates the need for a transparent and scalable method to identify which bridges pose the greatest risk to supply chain continuity. This study develops a bridge risk index using the threat–vulnerability–consequence (TVC) framework and validates its components with machine learning. Threat is defined as per-lane average daily traffic, vulnerability as effective bridge age (epoch), and consequence as detour distance, with traffic also contributing to disruption magnitude. The methodology applies log transformation and normalization to construct an interpretable multiplicative index, then classifies risk using Jenks natural breaks. The results show that epoch dominates vulnerability, detour distance amplifies consequence, and their interaction explains most of the risk variation. Specifically, effective age explains over three times more variation in bridge condition than any other attribute. The vulnerability–consequence interaction dominates with an R2 = 0.98. The highest-risk bridges are concentrated in rural areas and near major freight gateways where detour options are limited. The proposed TVC index provides a transparent, data-driven decision-support tool that agencies can apply nationwide to prioritize investments, safeguard freight corridors, and strengthen supply chain resilience.