For weeks, the lab had crowds with scribals, diagrams and chemical formulas in the whiteboard. A research team in the Olivetti Group and MIT Concrete Sustainability Hub was working intensely on a significant problem: how can we reduce the amount of cement in concrete to save cost and emissions?
The question was certainly not new; Fly ash, a by -product of coal production, and slag, a by -product of steelmaking, have long been used to replace some cement of concrete. However, the demand for these products is beating the supply as the industry appears to reduce their climatic effects by expanding their use, making the search for the option necessary. The challenge that the team gave was not the lack of candidates; The problem was that there were many through it.
On 17 May, Postdock Sorus Zaubi-led team published an open-access paper in Nature Communication material Underlining their solution. “We realized that AI was the key to moving forward,” Mahjobi noted. “There are too much data on potential materials – hundreds of thousands of pages of scientific literature.
With large language models, many of us uses like chatbots, the team created a machine-learning framework that evaluates and ordering the candidate content based on its physical and chemical properties.
“First, there is hydraulic reactivity. The reason is that the concrete is strong that cement – ‘glue’ that keeps it together – is harsh when exposed to water. Therefore, if we change this glue, we need to ensure that the option reacts equally,” Mahajabi explains. “Second, there is posholenity. This is when a material reacts with calcium hydroxide, when cement is found in water, a subport is made, so that the concrete can be made hard and strong over time. We need to balance hydraulic and posalanic materials in the mixture so that solids perform the best.”
Analyzing scientific literature and more than 1 million rock samples, the team used framework to sort the candidate content into 19 types, including biomass to mining sub -products from demolished construction materials. Mahjubi and his team found that the appropriate material was available globally – and, more effectively, many can only be grinded and included in concrete mix. This means that it is possible to extract emission and cost savings without much additional processing.
“Some of the most interesting materials that can change a portion of cement are ceramic,” said Mahjabi. “Old tiles, brick, pottery – all these ingredients can have high reaction.
The ability of everyday materials, such as earthen vessels and industrial materials such as mining is an example of how materials like concrete can enable a circular economy. Otherwise, by identifying and re -introducing the materials ending in the landfill, the researchers and industry can help these materials a second life as part of our buildings and infrastructure.
Further, the research team is planning to upgrade framework, and to be able to assess even more materials, while experimentally validate some of the best candidates. “AI Tools have removed this research in a short time, and we are excited to see how the latest growth in the big language model enables the next stages,” says Professor Elivetty, MIT Department of Materials of Materials Science and Engineering Work and Member Professor Elivetty, Elsa Olivetti. She serves as a MIT Climate Project Mission Director, a CSHUB principal investigator and leader of the Oliveti Group.
“The concrete is the backbone of the atmosphere,” says Randolf Kirchen, co-writer and director of CSUU. “By implementing data science and AI devices for material design, we expect to support industry efforts to create more continuously, without compromising power, safety or durability.
In addition to Mahzobi, Olivetti, and Kirchain, co-writers at work include MIT Postdock Vinath Venugopal, IPEC Bansu Manav SM ’21, PhD ’24; And CSHUB Deputy Director Hasm Azari.
The research was conducted through the MIT concrete sustainability hub, supported by the Concrete Advancement Foundation. This work also received money from MIT-IBM Watson AI Lab.
