Abstract
The relevance of the study is due to the formation of significant amounts of manmade waste from the energy and metallurgy of the Pavlodar region and the need for their rational involvement in the production of building mixes. The effect of the dispersion of fly ash and metallurgical slag on the physical characteristics of dry building mixes is studied using machine learning methods. The aim of the study is to develop a scientific approach to determine the optimal and critical range of particle sizes of man-made ingredients that ensure the qualitative characteristics of the building mix. The novelty of the research lies in the complex combination of experimental data and machine learning models (Gradient Boosting, Random Forest, polynomial regression) to predict the strength, water absorption, frost resistance and workability of building mixes for various compositions (fly ash 15% (20%), slag 20% (15%)). It was found that the optimal dispersion range is 30-40 microns, with a strength of 34-36 MPa, while water absorption does not exceed 4.2%, frost resistance up to 100 cycles, and workability is characterized by stable and maximum values. The accuracy indicators of the models are characterized by RMSE values 0.8 –1.2 and a coefficient of determination R2 up to 0.96, confirming the high reliability of forecasting. The practical significance of the research lies in the possibility of applying the developed scientific approach to optimize the processes of grinding and designing building mixes using manmade waste, contributing to the reduction of man-made waste in the Pavlodar region.