Application of tuned random forests model on cement paste including fly ash and MgO expansive additive

The limited level of hydration and challenges in managing delayed expansion provide obstacles to the exclusive use of fly ash (FA) and Magnesium Oxide (MgO) expansive additive (MEA) in significant quantities, respectively. Despite challenges, combining fly ash (FA) and MgO additive (MEA) is common in hydraulic mass concrete, yielding appropriate outcomes. Machine learning models assess the volume expansion (Ve) of cement paste, validated with 170 experimental findings. Random Forests (RF) algorithm is employed, incorporating various input variables such as Portland cement, FA, MEA, and sample age. Overall, the study's innovation lies in its advanced methodological approach, combining machine learning with sophisticated optimization techniques to tackle the complex problem of predicting volume expansion in cement paste. By validating the models, the study ensures that the models are grounded in empirical data, enhancing their applicability in actual hydraulic mass concrete projects. In the present study, the cutting-edge optimizers artificial hummingbird algorithm (AHA) and Chaos game optimization algorithm (CHA) were chosen for RF’s hyperparameter tuning. There was a considerable possibility for both RF-AHA and RF-CHA to estimate the V_e. Throughout the training and testing stages, the R² values of the RF-CHA model were 0.9845 and 0.979, respectively, indicating good levels of accuracy. Regarding the comprehensive index (i.e., OBJ), the lower value belonging to RF-CHA equals 0.0055, roughly smaller than the value for RF-AHA equals 0.0072.