Yerinde uygulamalar için makine öğrenimi algoritmaları kullanılarak sürdürülebilir yapı malzemelerinin reolojik özelliklerinin tahmini

Given that machine learning (ML) has gained significant acceleration in terms of computational power, this study investigates the prediction of rheological properties of cementitious materials using ML algorithms. The rheological properties, static yield stress (SYS), dynamic yield stress (DYS), and viscosity (VIS) play a crucial role both in material behavior and construction practices. In this context, by incorporating regression models of multiple linear (MLR), decision tree (DT), random forest (RF), and gradient boost (GB), the target properties of SYS, DYS, and VIS were estimated. In this study, an experimental dataset was generated through laboratory testing of various cementitious mixtures, composed of CEMII cement, fly ash, calcium oxide (CaO), water, certain superplasticizers (SP), and viscosity-modifying agents. Including a wide range of material compositions ensured that the contribution of each material to the ML models could be accurately interpreted. The study revealed that SP, CaO, and water-to-cement ratio are critical for estimating the target properties. Among the models, MLR and GB models both provided reliable predictive capabilities for the rheological properties. Overall, the findings suggest that the ML techniques carry significant potential in predicting the properties of cementitious materials, ultimately leading to more efficient and sustainable practices. Integrating additional materials and using more advanced models could further benefit the development of mixtures designed for specific performance requirements.