A civil engineering assistant professor at The University of Texas at Arlington is working to better understand a bridge’s structural health by combining machine learning with traditional monitoring measurements,
The 18-month, $122,000 grant to Dr. Suyun Ham of the Civil Engineering department is part of UTA’s membership in the Transportation Consortium of South-Central States (Tran-SET), a U.S. Department of Transportation Center administered by Louisiana State University. He will test his models in Dallas and Fort Worth.
The systems in place to monitor bridges today are weight-in-motion systems with sensors that measure vibrations, strain, and deflection. Measuring the bridge’s response to those elements present a picture of the bridge’s structural health. But the sensors do not take into account different types of trucks, multiple lanes, times of day and traffic congestion.
Ham is trying to create a system to supplement the weight-in-motion sensors currently in place with machine learning. The hope is that the resulting data would give transportation departments more accurate load parameters for bridges and a better picture of a structure’s overall integrity.
“We are combining a physics-based model with artificial intelligence, because the more a computer learns, the better information you get,” Ham stated. “Ultimately, the addition of machine learning allows us to accurately determine multiple conditions.”
Ham is also working on related research with a Texas Department of Transportation grant to use a non-contact testing system to make faster, easier, and more accurate determinations about when and where bridge repairs are needed.
Similar work is going on at the University of Surrey and King’s College London, where a team has developed a new machine learning algorithm to help monitor major infrastructure—such as dams and bridges.
In a paper published by the journal Structural Health Monitoring, researchers from Surrey and Kings detail how they created an AI system named SHMnet to analyze and assess the condition of bolts in metallic structures under stress conditions (DOI: 10.1177/1475921719881237).
Built on the foundations of a modified AlexNet neural network, the research team set up an impact hammer test under lab conditions. SHMnet was challenged to accurately identify the subtle condition changes of connection bolts on a steel frame under 10 scenarios where damage was inflicted.
The team found that when SHMnet is trained using four repeated datasets, it had a flawless (100%) identification record in their tests.
Dr. Ying Wang, the corresponding author of the paper and Assistant Professor at the University of Surrey, stated, “The performance of our neural network suggests that SHMnet could be incredibly useful to structural engineers, governments, and other organizations tasked with monitoring the integrity of bridges, towers, dams and other metal structures.”
“While there is more to do, such as testing SHMnet under different vibration conditions and obtaining more training data, the real test is for this system to be used in the field where a reliable, accurate, and affordable way of monitoring infrastructure is sorely needed,” Dr. Wang stated.
A technical paper on Bridge Structural Health Monitoring Aided by Big Data (BD) and AI was recently published in the Journal of Structural Engineering. While structural health monitoring (SHM) techniques have been widely-used in long-span bridges, the massive SHM data is not well interpreted, states the abstract.
“Big Data and AI techniques are seen as promising ways to address the data interpretation problem,” the authors write. “This paper aims to clarify the scope of BD and AI techniques on what and how regarding bridge SHM.”
The paper’s lead author is Prof. Limin Sun of the Department of Bridge Engineering at Tongji University, Shanghai, China. He studies the theory and technology of beam structure vibration, bridge structure health monitoring, and condition assessment and bridge inspection technology. His recent research topics have included: research on key technologies for health monitoring and online evaluation of large and complex bridge structures, and for highway bridge health monitoring based on the priorities of management.
Here, AI is used to process unstructured data for visual inspection and time series for structural damage detection. “The paper offers meaningful perspectives and suggestions for employing BD and AI techniques in the field of bridge SHM,” the authors state.
Comments