Road surface degradation manifests in various forms, ranging from subtle irregularities like surface roughness to more prominent damage such as interconnected alligator cracks. Among the numerous factors contributing to this deterioration, vehicle weight stands out as a particularly significant burden on infrastructure longevity.
In a groundbreaking research paper that earned top honors, Fengdi Guo, a doctoral student at the MIT Concrete Sustainability Hub, has unveiled new insights into the complex relationship between traffic weight and pavement deterioration. His innovative machine learning approach has demonstrated that heavy vehicles specifically accelerate damage in asphalt surfaces, while concrete pavements show remarkable resilience to increased weight loads.
The award-winning paper, titled "Assessing the Influence of Overweight Vehicles on Pavement Performance," secured first place in the prestigious Aramis López Challenge Category of the LTPP Analysis Student Contest. This competition, jointly organized by the Federal Highway Administration (FHWA) and the American Society of Civil Engineers' Transportation and Development Institute, recognizes outstanding student research in pavement engineering. Guo is scheduled to present his findings at the upcoming 2021 Transportation Research Board annual meeting.
Accurately predicting pavement deterioration plays a vital role in maintaining efficient transportation networks. Traffic patterns—particularly the cumulative weight loads over time—significantly influence how rapidly road surfaces degrade.
"The cumulative traffic weight represents two key factors: traffic volume, measured by annual average daily truck traffic (AADTT), and the actual weight of vehicles," explains Guo. "Our analysis indicates that increasing AADTT by just 1,000 trucks on a road segment could reduce the time between necessary maintenance interventions by approximately five months."
Given current transportation trends, truck traffic weight presents an escalating challenge. According to projections from the U.S. Energy Information Administration, heavy- and medium-duty vehicle traffic is anticipated to surge by nearly 40% by 2050, substantially outpacing growth in passenger vehicle traffic.
Addressing this substantial increase in heavy truck traffic demands more sophisticated analytical tools, especially since the precise relationship between vehicle weight and pavement deterioration has remained insufficiently understood until now.
"While it's well-established that heavier traffic accelerates asphalt pavement deterioration, the specific types of damage caused have been unclear," notes Guo. "Multiple variables, from precipitation levels to pavement layer thickness, can significantly alter how road surfaces respond to vehicle weight."
Traditionally, researchers and engineers have relied on either complex mechanistic models or data-driven approaches. Mechanistic models focus narrowly on pavement mechanical properties, require substantial computational resources, and prove impractical for analyzing entire pavement networks. Conversely, while data-driven models can be applied to networks, they typically cannot incorporate a pavement's unique maintenance and deterioration history.
In his innovative research, Guo expanded the capabilities of data-driven models by directly incorporating historical factors into his calculations rather than merely estimating them.
His methodology employs a recurrent neural network (RNN)—an artificial intelligence technique that loosely mimics the brain's neural connections to solve complex problems. He developed three specialized RNN models to predict roughness, rutting, and alligator cracking in asphalt pavements—performance metrics his research identified as particularly sensitive to traffic weight variations.
To construct his neural network, Guo designed a framework with input layers supplying relevant data (such as pavement structures and freeze index), hidden layers processing and correlating this information, and output layers presenting the final calculations. Unlike conventional feed-forward neural networks commonly used in pavement engineering, the hidden layers in RNN models can retain crucial historical information about pavement deterioration patterns.
After developing these models, Guo input road quality data from the FWHA's Long Term Pavement Performance (LTTP) database. His analysis revealed a clear correlation between traffic weight and specific forms of pavement damage.
"My models demonstrate that increased traffic weights on asphalt pavements accelerate deterioration rates for roughness by 1.3%, rutting by 7%, and alligator cracking by 3.7% for a representative asphalt pavement," Guo explains.
Due to dataset limitations, Guo's model couldn't fully determine traffic weight's impact on other forms of asphalt pavement damage. Moving forward, he plans to utilize more comprehensive datasets to investigate these additional potential effects. He also aims to explore the nationwide economic implications of overweight vehicles on infrastructure costs.
Until now, the specific impacts of traffic weight on road quality have represented a significant knowledge gap in the transportation community. While some questions remain unanswered, Guo's models have helped illuminate this persistent problem and opened new avenues for productive research in infrastructure management.
This research was conducted through the MIT Concrete Sustainability Hub with support from the Portland Cement Association and the Ready Mixed Concrete Research and Education Foundation.
