Blending the virtual and physical worlds
Civil and environmental engineering faculty have developed a testbed to make AI-enabled digital twins more accessible in the curriculum. Originally built for a new online certificate program, the testbed is also being used in active research projects to strengthen our infrastructure systems.
Technological advancements of the past few years have brought new norms to civil engineering and academia. Students are increasingly interested in online education and enjoy the convenience of completing courses at their own pace. At the same time, civil engineers parse through an array of new tools, such as artificial intelligence (AI) and digital twins, that are set to transform the field as they know it.
But when remote learning and civil engineering intersect, how are concepts that are—literally—concretely rooted in society’s physical structures translated to asynchronous forms of thinking and learning?
When instructors in Carnegie Mellon’s Department of Civil and Environmental Engineering set out to develop their new online certificate program, AI Engineering – Digital Twins and Analytics, these were the challenges they were tasked to solve. For a program designed for practicing engineers looking to upskill, remote learning was necessary to make the curriculum flexible to a full-time work schedule. Additionally, because a physical counterpart is so crucial to a digital twin’s virtual model, they needed to capture the substance of real-world data in a remote setting.
A key process in digital twin modeling is closing the loop between simulating the behavior of an infrastructure system and controlling it in the real world. So we felt strongly that we needed to give students the opportunity to interact with a physical structure instead of just models.
Mario Bergés, Professor, Civil and Environmental Engineering
“Digital twins are very new to our field so there isn’t a lot of precedent set on how to teach them, especially in an online class,” said Mario Bergés, professor of civil and environmental engineering. “A key process in digital twin modeling is closing the loop between simulating the behavior of an infrastructure system and actually controlling it in the real world. Because of this, we felt strongly that we needed to give students the opportunity to interact with a physical structure instead of just models.”
Student tests various bridge thicknesses, materials, and loads.
To accomplish this, the team enlisted help from an undergraduate student in the Department of Mechanical Engineering to construct a testbed that bridged the gap between the virtual and physical environments. Consisting of a train locomotive equipped with sensors navigating tracks that traverse a bridge, the 20-foot scaled-down model sits in a Carnegie Mellon lab but is continuously streamed to the classroom web portal.
Students can log-in from their computers, input the variables they’d like to test—such as speed of the vehicle or the sampling rate—and watch the train run live from wherever they may be. The data from the test is sent directly to their email, where they can analyze their results and adjust as needed, identifying problems like abnormal vibration patterns between the train and tracks that could indicate defective railway sections and use control algorithms to stabilize motion and maximize passenger comfort.
The testbed’s train car which is equipped with sensors for data collection.
“A problem we see in digital twin adoption is that most people are only dealing with one aspect of them. They differ from standard modeling technologies in that the virtual and physical environments are constantly informing and improving on each other,” said Bergés. “The testbed will give students their own physical structure to manipulate and ensure they understand how each aspect of the digital twin is integrated.”
Bergés and Pingbo Tang, associate professor of civil and environmental engineering, emphasize these points in the online certificate courses, Principles of Digital Twins and Digital Twins and AI for Predictive Analytics, first teaching students how to use digital twin models, then how to capture and analyze the data they generate. Both courses use the train testbed as an integrated project on which to apply the course material. Students who complete the program are equipped with new AI skills to bring back to their organizations and make better engineering decisions.
Side-view schematic of the testbed, vehicle, and pulley system.
Though originally built to advance digital twin education, the testbed has also proved useful in active department research projects and is directly applicable to work being done in the industry, explained Katherine Flanigan, assistant professor of civil and environmental engineering.
The testbed serves as an active research site, offering a scaled-down, hands-on environment to test, collect, and analyze data, and to make informed decisions to improve our infrastructure systems.
Katherine Flanigan, Assistant Professor, Civil and Environmental Engineering
“The testbed serves as an active research site, offering a scaled-down, hands-on environment to test, collect, and analyze data, and tomake informed decisions to improve our infrastructure systems,” Flanigan explained. “With this at our disposal, we can explore new configurations safely and conduct experiments at a much faster rate than would be possible in real-world settings.”
Faculty look forward to using the testbed in the lab and classroom, setting a new standard for how civil engineers integrate cutting-edge technologies into their field.