Researchers at the University of Illinois Urbana-Champaign (UIUC), Meta, and concrete supplier, Ozinga, have formed a partnership with the aim of discovering better low-carbon concrete formulas using artificial intelligence (AI).
Early-stage results have found that the AI-powered formulas reduce the carbon footprint of the concrete by 40 per cent while maintaining strength and durability. Meta has begun testing the formulas on multiple structures at the company’s DeKalb data centre in Illinois, namely the floor slabs of the guardhouse and the construction management team’s temporary offices.
Concrete is the most popular building material in the world, with between ten and 30 billion tonnes used for construction each year. But the price of that progress is a cost to the environment: Cement, an essential ingredient in concrete, is responsible for eight per cent of global anthropogenic greenhouse gas emissions.
“We designed new formulations that nearly halve the carbon requirements of concrete yet are just as strong or stronger than traditional formulations,” said Lav Varshney, an associate professor of electrical and computer engineering at UIUC. “Given the popularity of concrete, there is a global scale of potential applications.”
With its glue-like properties, cement has historically been combined with other ingredients, such as water, sand, and coarse aggregates, to make concrete. But the manufacture of cement causes enormous amounts of carbon emissions, in part because of the fuels needed to heat some of the ingredients to 1,400 degrees Celsius. In addition, one of the key ingredients is limestone (or calcium carbonate), which releases carbon dioxide during calcination in the manufacturing process.
To replace cement in the concrete mix, researchers had to identify a formula that would be as strong, durable, and workable as the standard one.
Varshney and Nishant Garg, an assistant professor of civil and environmental engineering, trained a model using the Concrete Compressive Strength data set, which is openly available from the UCI Machine Learning Repository. This database has 1,030 concrete formulas along with their validated attributes, including seven-day and 28-day compressive strength data (i.e., how the concrete gained strength seven days and 28 days after pouring). The embodied carbon footprint associated with the concrete formulas was derived using the Cement Sustainability Initiative’s Environmental Product Declaration (EPD) tool.
EPDs are a standardised way of accounting for the environmental impacts of a product or material, including carbon emissions over its life cycle.
Using the input data on concrete formulas along with their corresponding compressive strength and carbon footprint, the Al model was able to generate several promising new concrete mixes that replaced cement with other supplementary materials, such as fly ash and slag. The final recipe was tested and further refined by Ozinga – taking into account several factors including expected cold weather conditions and material availability – before it was poured at the Meta DeKalb data centre.
“A lot of researchers are using AI for predictive purposes, in that you give them certain recipes and they can predict the strength or some other characteristic,” said Garg, who specialises in the chemistry and characterisation of construction materials. “But our approach is unique in that we leverage the best available data and use the model to generate the potential recipes based on our needs. It is tremendously useful.”
The model may also be useful in designing concrete formulas for places where building materials may be less readily available – for example, for constructing cell phone tower foundations in remote rural regions.
