Meter's latest AI discovers stronger, greener concrete formulas | Technology
|Meter's latest AI discovers stronger, greener concrete formulas|
Meter's most recent AI identifies concrete formulas that are both stronger and greener
While AI/ML systems can't shout "Eureka!" like humans, they've showed enormous promise in the field of compound discovery, whether it's combing through mountains of data to locate novel medicinal compounds or creating new compositions based on the flavour profiles of the components. Now, Meta AI has developed an AI that can conceive and modify formulae for progressively high-strength, low-carbon concrete in collaboration with experts at the University of Illinois, Urbana-Champaign.
Traditional concrete manufacturing methods that produce billions of tons each year are not environmentally friendly. They generate around 8% of the world's total annual carbon dioxide emissions. The concrete industry has made progress in reducing CO2 emissions in recent years (also making the material stiffer, more flexible and able to charge electric vehicles), but overall its production is modern and the most carbon-intensive.
Reducing the carbon content in concrete is as easy as changing the material in the concrete. These ingredients consist of four basic components: cement, aggregate, water, and a mixture (which acts as a dopant). Since cement consumes the most carbon of the four, research is underway to reduce the amount of cement required by supplementing with low carbon elements such as fly ash, slag and ground glass.
READ MORE: Intel provides AI with in-depth education on DIA imaging cardiac ultrasound analysis techniques | Technology
Likewise, composite materials such as crushed stone, crushed stone and sand can be replaced with recycled concrete. The problem is that there may be dozens of components available, and the proportions of all components interact to affect the structural profile of the concrete. In short, researchers have every possible combination for testing, selection, and improvement. Processing these myriad options at human speed will go on forever. That's why Meta people train AI so much faster.
Lav Varshney, prof of electrical and computer engineering department and Prof Nishant Garg team first worked with the Department of Civil Engineering at the University of Illinois at Urbana-Champaign to train the model using a dataset of compressive strength of concrete. The collection includes more than 1,000 hard formulas and their structural characteristics, including 7-day and 28-day compressive strength data. The team used the Cement Sustainability Initiative's Environmental Product Declaration (EPD) tool to determine the carbon footprint of the resulting concrete mix.
From the resulting list of possible sources, the research team selected the five most promising options and repeated until they met or exceeded the 7-day and 28-day energy targets, reducing carbon demand by at least 40 percent. .. The refining process takes only a few weeks, and when 50% of the required cement is replaced by fly ash and slag, it is possible to produce concrete formulations that exceed all these requirements. Later, Meta teamed up with Ozinger, a concrete company that recently built Meta's newest data center in Illinois, to further refine the formula and conduct practical experiments.
Going forward, Metateam will further improve the formula's 3-day and 5-day intensity distribution (basically, make it dry quickly so the rest of the build can go quickly) and various weather conditions. I would like to understand how it applies. Air, high humidity, etc.
Source: A. Tarantola, Engadget, Direct News 99