Utilization of a novel artificial intelligence technique (ANFIS) to predict the compressive strength of fly ash-based geopolymer

Fly ash (FA) is widely used to synthesize geopolymers, but it is heterogeneous as it consists of reactive, partially reactive, and inert parts, which may influence the behavior of the resultant geopolymer. Therefore, in this experimental and analytical work, at first, the reactivity of the FA was assessed by modified Chapelle test, which was further investigated by conducting a dissolution test to study the influence of temperature (20, 60, and 100 degrees C) and time (6 - 24 h). Afterward, geopolymer paste was synthesized by varying: (i) alkaline to precursor ratio (0.3 - 0.5), (ii) sodium silicate to sodium hydroxide ratio (2 - 3), and (iii) curing temperature and age. Based on the additional parameter of the molarity of NaOH including the above-mentioned parameters, an adaptive neuro-fuzzy inference system (ANFIS) to predict the compressive strength was optimized. A prominent increase in reactivity was observed at 60 degrees C as compared to 20 degrees C. The compressive strength improved significantly at A/P ratio of 0.4 and 0.5 resulted in improved compressive strength in particular for the 2.5 SS/SH ratio which was verified by FTIR. Analytical results by ANFIS were compared with the multivariate adaptive regression spline (MARS) model in terms of R-2, RMSE, and MAE it was concluded that the ANFIS model showed better correlation and significantly fewer errors as compared to the MARS model. Finally, the developed model was checked and validated by employing the real experiment test results based on parametric values obtained from the ANFIS model. The developed model of this study can provide a novel approach for the design of geopolymers based on artificial intelligence technique.