AI Accelerates Development of Low-CO2 Cement Formulations, Reducing Industry's Carbon Footprint

The cement industry, responsible for about eight percent of global CO2 emissions—more than the entire aviation sector—is a significant contributor to environmental degradation. Addressing this issue, researchers at the Paul Scherrer Institute (PSI) have pioneered an AI-based model that accelerates the discovery of eco-friendly cement formulations, maintaining material quality while reducing carbon footprint. Traditional cement production involves heating limestone in rotary kilns to 1,400 degrees Celsius, a process that releases chemically bound CO2, accounting for over half of the industry's emissions. This makes altering the cement recipe a crucial strategy for emission reduction. An interdisciplinary team at PSI's Laboratory for Waste Management, led by mathematician Romana Boiger and including Nikolaos Prasianakis and John Provis, focused on enhancing the cement formulation process. Rather than conducting extensive lab experiments or using resource-intensive simulations, they leveraged machine learning to optimize cement recipes. Their AI model, powered by a neural network trained on data generated using PSI's open-source thermodynamic modelling software GEMS, can rapidly predict the mechanical properties and CO2 emissions of different cement formulations. Boiger explains that the AI model can generate practical recipe suggestions within seconds, a dramatic improvement over the time typically required for traditional methods. This not only speeds up the development cycle but also reduces the computational burden and experimental costs. Prasianakis adds that the vast range of material compositions makes the AI-driven approach particularly useful for identifying promising candidate recipes for further experimentation. Industrial by-products like slag and fly ash are already used to replace some clinker, which cuts CO2 emissions. However, global cement demand necessitates more scalable and reliable alternatives. Provis highlights the importance of finding the right combination of materials that can be sourced in abundance and produce high-quality cement. The complexity of these combinations stems from the geochemical processes involved in cement hardening, which are computationally intensive and expensive to model. To optimize the cement formulations, the researchers formulated the problem as a mathematical optimization task, aiming to maximize mechanical properties while minimizing CO2 emissions. They integrated genetic algorithms, which mimic natural selection, to identify formulations that best meet these dual criteria. This "reverse approach" allows for targeted recipe development, focusing on specific desired outcomes instead of testing numerous combinations randomly. Some of the AI-identified formulations show great promise, balancing CO2 reduction, material quality, and practical feasibility in production. However, these candidate recipes need to be validated through laboratory testing before widespread adoption. Boiger emphasizes that the AI tool is a proof of concept, demonstrating the ability to identify promising formulations through computational means. The team plans to expand the tool to include additional parameters such as raw material availability and usage environments, like marine or desert conditions. The success of the project hinges on the interdisciplinary collaboration among experts in cement chemistry, thermodynamics, and AI. Prasianakis credits the integration of diverse skills and knowledge, along with the support of other research institutions like EMPA, within the SCENE project. SCENE, the Swiss Centre of Excellence on Net Zero Emissions, aims to develop solutions for significant greenhouse gas reductions in industries and energy supplies. Industry insiders and experts have applauded the potential of this AI-driven approach. Romana Boiger notes the substantial time savings offered by the workflow, which can be adapted for various material and system designs. John Provis sees this as a critical step towards achieving sustainable cement production, highlighting the practical and environmental benefits of the AI model. The team's innovative method not only accelerates the discovery of climate-friendly cement recipes but also opens the door to more efficient material design across multiple industries. PSI is a leading research institution focused on the development and application of advanced scientific and technological solutions. The SCENE project, supported by PSI, is a testament to the institute's commitment to addressing major environmental challenges. The integration of AI into materials science showcases a promising pathway towards a more sustainable future, leveraging cutting-edge technology to tackle one of the most polluting industries.