MOSCOW, 13 Jul. To create concrete resistant to aggressive influences for construction in any weather and climatic conditions, a machine learning model that selects the components of a concrete mixture will allow, SibFU scientists believe. According to them, with the help of the development of the university, it is possible to produce more durable building materials for use in areas of extreme temperatures. The results of the study are published in the Journal of Building Engineering.
For the construction of residential buildings and industrial facilities in regions with extremely low temperatures or in aggressive operating conditions, for example, on saline soils and with significant temperature differences, special concrete is needed. It must have high strength and resistance to possible climatic anomalies, the Siberian Federal University (SFU) said.
To improve the properties of concrete, various additional components are used. To build a machine learning model that determines the most important rules and recipes for the best samples, more than 40 concrete mix compositions with different aggregates from 5 different deposits with different quality, types and compositions were used.
However, additives include more than 10 different types of mineral and chemical components. This model was able to take all this into account and highlight the main components that affect the resistance and durability of concrete. This is especially true for hard-to-reach and remote regions, including the northern territories, where high-quality materials are usually not available, but are still used in modern industry, the university noted.
Specialists stressed that it is important to build equally high-quality buildings throughout Russia. However, trial and error cannot guarantee that the desired quality will be achieved everywhere. SibFU scientists have proposed a solution to this problem — using the model they created for selecting the optimal composition of concrete mixtures for construction work in the Far North and other regions with extreme weather conditions.
“The modeling process selects the best combination, analyzes the data in the selected areas, and provides useful recommendations, including “recipes” for concretes with the required properties. Such a model is a valuable assistant for materials scientists, since it takes into account all the factors that cannot be expected from a person in due to limitations in experience and available time,» explained Maxim Molokeyev, Associate Professor at the Institute of Engineering Physics and Radioelectronics of the Siberian Federal University.
The scientist added that the model works on the basis of machine learning, a type of artificial intelligence that is improved by solving tasks by comparing and analysis of many possible options. That is, it selects the best combination of components, taking into account their availability and availability in the regions.
Also, one of the main advantages of the development is the accurate prediction of the properties of the compositions proposed by it and their adaptability to specific weather and climatic conditions. Of course, frost resistance and high strength will be the main criteria, Molokeyev emphasized. /20220621/dgtu-1796808375.html» data->
«The program allowed us to find an approach to eliminate problems with low quality aggregates, optimize the composition of the mixture, and showed the factors associated with chemical and physical processes in concrete, which with a high degree of probability determine how this composition will be frost-resistant and applicable, say, in Norilsk , for construction on saline soils, the runway of airfields or any other objects in specific conditions,» added Irina Yendzhievskaya, head of the testing laboratory for building materials and chemical analysis of water at SibFU.
Artificial intelligence evaluated the contribution of all parameters to the composition optimization. In particular, it was believed that «excess» air unambiguously reduces the strength of concrete, but, according to the model calculations, within certain limits it increases the resistance of the material without reducing strength. And this conclusion has already been confirmed by testing in practice, the scientist drew attention. /20230620/nauka-1878701991.html» data->
The researchers also plan to develop competitive cementless concrete using machine learning methods. Its main advantage is the reduction of carbon dioxide emissions during production. In addition, according to the plans of scientists, it will use activated industrial waste — ash and nepheline sludge, which will make production more profitable, said Yendzhievskaya. br />