Abstract
Additives from man-made waste are currently a fairly useful ingredient in the composition of preparatory building mixes, providing both chemical and physical characteristics of the plaster mixture. This study describes the developed comprehensive approach to assessing and predicting the adhesive strength of plaster mixtures containing fly ash and metallurgical slag. The study used microstructural analysis of the plaster mixture after 7, 14, 21 and 28 days, hardening depending on the humidity regime and machine learning methods using various models. Based on the SEM/EDS data, the parameters of the interphase transition zone, porosity, the proportion of C– S–H / C–(A)–S–H gel phases, and the degree of reactivity of technogenic additives were quantified. In the course of the experiments, it was found that a decrease in the porosity of the interfacial transition zone from ~20% to less than 6% provides a twofold increase in adhesive strength, while a transition from adhesive to cohesive fracture occurs. It is shown that gradient boosting models (CatBoost, XGBoost) provide high accuracy of adhesion prediction (R2 > 0.95; RMSE< 0.05 MPa). The scientific novelty of the work is to establish a quantifiably interpretable relationship between the SEM parameters of the interphase transition zone and adhesive strength using SHAP analysis, which allows us to move from an empirical assessment to a computer study of building mixes with manmade additives. The practical significance lies in the possibility of optimizing the hardening modes with a reduction in the volume of experimental tests and the detection of various pathologies at early stages.