Cement production accounts for about 8% of global CO2 emissions—more than the entire aviation industry produces. Researchers at the Paul Scherrer Institute (PSI) have developed an artificial intelligence system that accelerates the search for new cement formulations, enabling the preservation of material quality while simultaneously reducing its carbon footprint.

The Emissions Challenge in the Cement Industry

At cement plants, rotary kilns are heated to 1400°C to produce clinker, the main ingredient in cement. Most CO2 emissions are linked not only to fuel combustion but also to the chemical decomposition of limestone at high temperatures. One strategy to reduce emissions is to replace part of the clinker with alternative binding materials.

Artificial Intelligence for New Formulations

The interdisciplinary PSI team has developed a machine learning-based approach that allows for modeling and optimizing cement compositions to lower CO2 emissions while maintaining mechanical properties. This model rapidly generates practical recipes, significantly speeding up the process of discovering new formulations.

Using artificial neural networks trained on data from GEMS software and experimental studies, scientists can quickly assess the mechanical properties and CO2 emissions of various cement recipes. This approach has accelerated calculations by about a thousand times compared to traditional modeling methods.

Genetic Algorithms for Optimization

To find optimal compositions, the team used genetic algorithms, which help identify recipes with maximum mechanical strength and minimal CO2 emissions. This targeted search avoids blindly testing countless combinations and focuses on the most promising options.

Outlook and Future Development

Among the discovered formulations, there are already promising candidates that require further laboratory testing. The research demonstrates that mathematical modeling can effectively identify efficient cement recipes. In the future, the tool could be expanded to consider additional factors such as raw material availability and material operating conditions.

The project was carried out as part of the SCENE (Swiss Centre of Excellence on Net Zero Emissions) program, which aims to develop solutions for reducing greenhouse gas emissions in industry and energy sectors.