Data Science and Machine Learning Drive Innovation in Civil Engineering at Stevens
Stevens professors are utilizing new data science techniques and machine learning to accelerate research in various civil engineering disciplines, from developing smart infrastructure to predicting storm surges
Machine learning—fueled by data analytics, data mining—is one of the most pervasive and applicable technologies on the planet, touching nearly every area of modern society. Professors at Stevens Institute of Technology’s Department of Civil, Environmental, and Ocean Engineering (CEOE) are harnessing these cutting-edge technologies to advance the design, construction, and maintenance of smart, sustainable, and resilient infrastructure systems, as well as mitigating the impact of natural disasters.
Diving into the Subsurface
When a large underground highway or sewer pipe needs to be built, civil engineers use tunnel boring machines (TBMs) to cut through rock, clay and other soil types. These machines, also called moles, can be as tall as a building and span more than 300 feet in length.
“The single most significant challenge facing underground construction is not knowing with a fair degree of certainty what lies underneath the surface, which is especially true in tunnel construction,” said Rita Sousa, an assistant professor in CEOE who studies tunneling and subsurface soil conditions. Using data collected from sensors placed in the cutting head of TBMs, Sousa applies machine learning to extract knowledge about the subsurface and predict ground conditions ahead of construction. This type of predictive modeling can result in more efficient, faster, and safer subsurface construction operations.
“Using data science in civil engineering is relatively new,” said Sousa. “It has become more mainstream and presents exciting opportunities to address many challenges in the field of civil engineering.”
Valentina Prigiobbe, CEOE assistant professor, uses machine learning for the calibration and optimization of statistical models to identify aging sewer pipes impacted by groundwater flooding (or infiltration).
“This approach provides a tool for strategic intervention for sewer repair and sewer flood mitigation. Its implementation is flexible and it can be extended to other types of infrastructure,” Prigiobbe said.
Prigiobbe also uses advanced techniques such as density functional theory (DFT) - a quantum mechanical modeling method often used in computational chemistry and physics - to look at how the radioactive element radium moves through soil. With her students, she is complementing this approach with data science tools to uncover many complexities associated with transport processes in the subsurface, she said.
Monitoring Bridges and Infrastructure
Kaijian Liu, CEOE assistant professor, is interested in harnessing the power of big data and machine learning to better develop and manage smart built environments.
“A fundamental research question is how to extract actionable knowledge from the large, scattered and heterogeneous data that are increasingly available in order to support enhanced decision making towards smart buildings, civil infrastructure systems and communities,” said Liu.
In addressing this question, Liu and his research team developed a novel smart data analytics framework that is able to extract, fuse, and analyze bridge data from multiple sources for enhanced bridge deterioration prediction and maintenance decisions. The framework uses novel machine learning methods to allow the utilization of both structured and unstructured data for prediction, which goes beyond the state-of-the-art where the two kinds of data are mostly handled separately. This framework has the potential to transform the way decision makers use the scattered and heterogeneous data in an integrated, holistic, and analyzed manner towards bridge management. Currently, Liu and his cross-institutional collaborators are creating a Civil Infrastructure Systems - Open Knowledge Network (CIS-OKN) to enable data-driven civil infrastructure decision making for improved safety and accessibility, and to increase economic opportunity for the public.
Mohammad Ilbeigi, CEOE assistant professor, also focuses on developing machine learning-based methods to analyze uncertainties in infrastructure development. Recently, Ilbeigi and his research team developed bridge deterioration forecasting models and optimal bridge inspection procedures. This can help transportation agencies assign their inspection and maintenance resources more efficiently and reinvest millions of dollars into much needed infrastructure development projects.
“Today, we have access to an unprecedented amount of data in infrastructure systems,” said Ilbeigi. “Using the available data in conjunction with high-performance machine learning methods, we can eventually solve problems that we could not address for decades due to lack of data and computational power.”
Predicting Storms Surge
The data available to civil engineers will grow much larger in the future, said Muhammad Hajj, CEOE chair and Director of the Davidson Lab.
“They will be expected to use these data to make better-informed decisions, and our department is at the forefront of implementing data science in civil engineering applications” said Hajj.
Hajj, his colleague Reza Marsooli, an assistant professor in the department, and their students are using machine learning to advance the capability of coastal flood hazard prediction induced by extreme wind events in a changing climate. In a recent study, they showed that physics-based simulations can be made more effective, when combined with machine learning, towards predicting low-probability, high-consequence storm surge events.
“What would take months or years using only high-fidelity physics-based computational simulations could be reduced to hours or days when these simulations are combined with machine learning, which is very significant when trying to determine low-probability storm surge event, for example, a 100-year return period,” said Hajj.
Learn more about the Department of Civil, Environmental, and Ocean Engineering at Stevens: