Researchers at Ogarev Mordovia State University (MRSU) have pioneered a high-performance, self-compacting, shrinkage-free fine-grained concrete, marking a breakthrough in construction materials. This advanced concrete, boasting physical and mechanical characteristics 2−4 times superior to traditional heavy concrete, was engineered using artificial intelligence and machine learning algorithms.

The innovation is a dry construction mix that combines Portland cement, quartz sand, and a complex organic-mineral additive. The additive integrates surfactants, metallurgical waste, and commonly available sedimentary rocks. This multifunctional component enables the cement system to self-compact and counteracts shrinkage deformation when water is added.

«The material is efficient across all stages of its lifecycle—from preparation and placement to curing and long-term structural performance. Its low permeability, high strength, and corrosion resistance mean it can last 150−200 years under standard non-aggressive or mildly aggressive conditions. Additionally, the mix is 10−15% less costly than current alternatives, thanks to its reduced cement content and the use of locally sourced materials and industrial by-products,» explained Artemy Balykov, an Associate Professor at MRSU’s Department of Structural Engineering.

This new concrete is applicable for a wide range of construction purposes, from residential to industrial and transportation infrastructure. It’s especially promising for complex structures like stadiums, bridges, overpasses, airfields, and energy platforms, including those in the oil, gas, thermal, and nuclear sectors.

Machine learning, a key field within artificial intelligence, was instrumental in the concrete’s development. The research team created and analyzed an extensive database containing over a thousand formulations of cement materials to fine-tune the mix.

«Our intelligent approach involved creating multifactorial models that allow the computing system to predict material properties and optimize the composition with precision. Using ensemble algorithms and neural networks, these machine learning models can monitor predictive factors and continuously improve with new data, providing insights into how various formulations and processing parameters affect the structure and performance of fine-grained concrete,» noted Elena Kaledina, an Associate Professor at MRSU’s Department of Applied Mathematics.

These machine learning models offer a flexible toolset for accurately selecting compositions and predicting the characteristics of high-quality concrete. The system is expected to become a valuable resource for companies in the construction materials industry.

«The material, developed under the leadership of Dr. Tatiana Nizina, has successfully completed laboratory testing and is ready for adoption in the construction sector. The next steps involve adapting the technology for specific production environments. Industrial partners of MSU have already shown interest, and we are in talks with regional enterprises that manufacture ready-mix concrete, dry mixes, and reinforced concrete products to bring this innovative material to market,» stated MRSU Rector Dmitry Glushko.

This breakthrough in cement materials, along with the machine learning-based design system, was developed as part of the Russian Ministry of Science and Higher Education’s «Priority 2030» program under the strategic project «New Generation Materials and Energy Efficiency.» The project also received support from the Russian Science Foundation under initiative No. 21−73−228, titled «Designing Compositions, Modeling Structure and Properties of High-Strength Lightweight Concrete with Machine Learning.»