Preliminary evaluation of Design of Experiments (DoE) approach for the performance prediction of basalt sawing sludge- metakaolin-based Alkali Activated Materials (AAMs)

In the recent decades, Alkali-Activated Materials (AAMs) have gained success as an alternative to the traditionally used Portland cement, due to their sustainability and good properties. However, the optimization of the formulations is challenging and requires much time and resources. In this study, Design of Experiments (DoE) approach was used to predict and optimize the compressive strength and ionic conductivity of AAMs made up of basalt sawing sludge (FB) and metakaolin (MK) activated with NaOH and Na 2 SiO 3 . D-optimal design was employed to understand the effects of four different factors (i.e., abundances of FB, MK, NaOH and Na 2 SiO 3 in the formulations) on the selected responses. Ten formulations were given by the software and tested by XRD, FT-IR, compressive strength, and ionic conductivity measurements of the eluate after the immersion of the consolidated materials. These latter were carried out in water as an index of the chemical stability salt release. The results demonstrated that higher MK and Na 2 SiO 3 contents improved the mechanical performance, while NaOH led to the reduction in strength. Moreover, ionic conductivity was mainly influenced by NaOH, with increased content leading to higher ion releases responsible for the unwanted efflorescence. Within the investigated domain, compressive strength ranged from 5.4 to 32.4 MPa, while ionic conductivity ranged from 128 to 1396 μS/cm. The model predicted both mechanical strength and ionic conductivity values with good accuracy of fit of the statistical model to the data (compressive strength: R² = 0.99, RMSE = 1.23 MPa; ionic conductivity: R² = 0.99, RMSE = 44.43 μS/cm), demonstrating reliability for further optimize AAMs. Five additional formulations, predicted by the software, were synthesized for validation of the model, reaching up to 27.4 ± 1.7 MPa and 112 μS/cm for the best prediction. This statistical model allowed to reduce experimental trials while giving the possibility to control the final desired performance of brand-new formulations.