ECE ILLINOIS researchers, together with collaborators from IBM and MIT, have developed algorithms that can discover new formulas for concrete that are stronger and reduce its carbon footprint by more than 40%. The innovation is a result of their AI algorithms that create novel and high-quality formulations of the man-made material. The findings could change the way concrete is formulated around the world.
ECE ILLINOIS Professor Lav R Varshney
and his team, led by ECE ILLINOIS alumnus Xiou Ge, made the breakthrough using computational creativity to choose proportions of ingredients that yield strong concrete but also reduce the amount of ingredients, like cement, that have the most environmental impact. This research is part of a larger program of study on computational creativity, as part of the Illinois-IBM Center for Cognitive Computing Systems Research (C3SR).
“With these results, a construction engineer may create a formula that meets structural needs and best addresses local environmental concerns,” Varshney explained.
Concrete is the most widely used engineered material in the world with more than 10 billion tons produced annually. However, that scale brings a significant burden to the environment in terms of energy, water, and release of greenhouse gases and other pollutants. The concern has increased interest around the world in creating concrete formulas that minimize that environmental burden while satisfying engineering performance requirements.
“One might have thought since this has been an active area of study for centuries, nothing more could be done, but in fact our neural network-based algorithms are able to come up with formulations that are stronger than existing formulations and yet can have 40% or more reduction in CO2 emissions (as measured through a lifecycle analysis), among other reductions in environmental impacts,” Varshney explained.
Recent advances in artificial intelligence have enabled machines to generate highly plausible artifacts, such as images of realistic looking faces. Semi-supervised generative models allow generation of artifacts with specific, desired characteristics. In this work, Varshney and his team use Conditional Variational Autoencoders (CVAE), a type of semi-supervised generative model, to discover concrete formulas with desired properties.
“We demonstrate CVAEs can design concrete formulas with lower emissions and natural resource usage while meeting design requirements. To ensure fair comparison between extant and generated formulas, we also train regression models to predict the environmental impacts and strength of discovered formulas,” Varshney said.
To prove the efficacy of their formulations, Varshney’s team enlisted Civil and Environmental Engineering Professor Nishant Garg. Garg made the formulations predicted by Varshney’s algorithms and tested their compressive strength. Their results demonstrate the concrete created from the AI-generated formulas are, in fact, stronger than existing concrete formulations.
Garg further suggested to Varshney the idea of using local materials as ingredients, rather than having to transport materials to the building site. “AI algorithms might enable the development of localized and/or personalized concrete formulations for such an application,” Varshney said.
One specific application Varshney’s team is considering is in building the foundations of cell phone towers to bring internet connectivity to half the world's population that is unconnected. “Perhaps contrary to what one might think, the foundation is the biggest bottleneck, as compared to the tower itself or the radio electronics.”
Varshney’s research team recently received funding from the Facebook Connectivity Lab to extend the work to specifically focus on building materials for telecommunications infrastructure in remote parts of the world, such as the mountains of Peru.
Additional Funding Sources: -
IBM T. J. Watson Research Center in collaboration with the MIT Concrete Sustainability Hub and the IBM-Illinois Center for Cognitive Computing Systems Research (C3SR).