Bridges over busy waterways carry thousands of automobiles every day while sitting in the middle of complex maritime shipping lanes. Many of these structures are aging, were designed for very different traffic volumes than they see today, and were not built with modern vessel sizes or security threats in mind. Traditionally, agencies have relied on visual inspections and design checks focused mainly on natural hazards like wind, earthquakes, and floods. The Key Bridge collapse showed that this approach is not sufficient.
We need tools that can both track how bridges are deteriorating over time and evaluate how exposed they are to threats that fall outside traditional engineering models, especially collisions, intentional attacks, and cyberattacks on monitoring systems. This research area tackles both sides of that challenge through two complementary projects.
I. Monitoring Bridge Health
Most bridges in the United States are inspected on a regular cycle, but those inspections offer only periodic snapshots of their condition. This project uses machine learning to turn decades of inspection and environmental data into continuous, forward-looking health forecasts.
The team built a nationwide dataset of thousands of long-span steel truss, suspension, and cable-stayed bridges, including structures similar in design to the Francis Scott Key Bridge. That dataset combines:
- Federal inspection and inventory records tracking condition ratings over time
- Bridge geometry and design details such as span length, material type, and structural configuration
- Traffic demand indicators including average daily traffic and truck share
- Local climate data (temperature and precipitation) from NOAA
Using this information, the team computes and forecasts a Bridge Health Index (BHI): a single score that reflects the condition of a bridge’s key components and its effectiveness. Several machine learning models were trained and compared to predict current and future BHI values, quantify uncertainty in those predictions, and identify which factors most strongly drive deterioration. This approach enables:
- Early warning when a bridge is deteriorating faster than expected
- Peer comparison placing the Key Bridge alongside structurally similar bridges nationwide to understand how its health evolved relative to comparable structures
- Risk-based maintenance planning, highlighting which bridges are likely to need attention soon and which can safely remain on longer inspection cycles
- Understanding what accelerates wear, such as heavy truck demand and harsh climate exposure
For agencies managing large bridge inventories, this shifts the paradigm from “inspect and react” to “predict and plan”.
Essential Bridge Assets
- Superstructure — The deck, beams, and cables that carry traffic across the span
- Substructure — Piers and foundations that transfer the bridge’s load to the ground
- Pier Protection — Fenders, dolphins, and other energy-absorbing structures that shield piers from vessel impact
- Navigational Clearance — The horizontal and vertical space that allows vessels to pass safely beneath the span
- Traffic-Carrying Function — The bridge’s capacity to handle its planned volume and vehicle load range
- Surveillance & Monitoring — Cameras, sensors, and structural health systems that track conditions in real time
- Lighting & Alarms — Navigation lighting and emergency alert mechanisms that support safe operations day and night
II. Threat & Vulnerability Risk Assessment
Most existing risk frameworks for bridges focus on natural hazards. Human-induced threats like large-vessel collisions, intentional attacks, arson, vandalism, cyberattacks on control systems are often handled separately, if they’re assessed at all.
SMARTER researchers developed a Threat and Vulnerability Risk Assessment (TVRA) framework specifically for bridges over navigable waterways in the Mid-Atlantic region, drawing on lessons from the Key Bridge collapse and other high-profile incidents. The framework works in three steps:
Step 1 — Break the bridge into its critical assets. Each bridge is decomposed into key components: superstructure (deck, beams, cables), substructure (piers and foundations), pier protection systems, navigational clearance, surveillance and monitoring systems, and lighting and alarm infrastructure. Each asset is scored for importance based on its role in structural safety, economic significance, traffic disruption potential, and impact on navigation and emergency response.
Step 2 — Identify the threats. The framework considers a defined set of human-induced threats:
- Vessel collisions
- Terrorist attacks and sabotage
- Arson and intentional fires
- Vehicle crashes or overloading events
- Unauthorized access and vandalism
- Cyberattacks on control or monitoring systems
- Drone interference and other emerging technologies
For each bridge, threats are matched to the specific assets they could affect.
Step 3 — Quantify impact and vulnerability. For every asset–threat combination, the framework evaluates impact and vulnerability . These scores combine into a risk rating that highlights which assets are most critical, which threats pose the highest risk, and where mitigation investments would have the largest effect.
Key Insights
Aging and exposure matter. Bridge age, length, truck demand, and climate exposure all have measurable effects on the health and deterioration of bridges. These machine learning models can quantify them in ways that support concrete decision-making.
Risk is complex and multi-dimensional. Structural qualities, maritime traffic, climate, and other risk vectors must be considered holistically. A bridge that looks acceptable under a traditional inspection may appear high-risk once human-induced threats and projected deterioration are factored in.
Vulnerabilities are often clustered. The TVRA case studies show that risk is often concentrated in a handful of asset–threat pairs. For example, specific piers exposed to vessel impact, or critical spans with limited structural redundancy. That makes targeted protection investments highly cost-effective.
Mitigation is often targeted, not blanket. Upgrading pier protection at a few critical locations, improving monitoring systems, or adjusting vessel traffic operating procedures can significantly reduce system-level risk without requiring massive across-the-board spending.
