Artificial Intelligence Models for Predicting the Compressive Strength of Geopolymer Cements

The utilization of nanosilica and cellulose nanocrystals (CNCs) in cement geopolymers remains challenged by intricacies and uncertainties regarding their concentration, posing difficulties in the formulation of systematic geopolymer mix designs. This study aims to formulate models based on Artificial Neural Networks (ANN) capable of forecasting the compressive strength of geopolymers through the utilization of experimentally acquired data. Nanosilica was applied at concentrations of 2%–4% and CNCs at 1%–3%. ANN was modeled using MATLAB to predict the compressive strength of the geopolymer. The results indicated an effect of nanosilica and CNCs on the compressive strength of geopolymer at 2%–4% concentration and 1%–3% CNCs. The best ANN was the GDX training function, purelin activation function, LGD and LGDM learning functions, Lr 0.1 and 0.01 at the number of epochs 3812 out of 25000 and 1774 out of 25000, resulting in the best correlation values of 0.994 and 0.959; the lowest RMSE values are 0.022 and 0.110. The results of the ANN model built based on actual data prove that the model is helpful for accurate simulation to predict the compressive strength of geopolymer cement. This study contributes novelty by optimizing the design model for Geopolymer Cements incorporating nanosilica and CNCs. Doi: 10.28991/CEJ-SP2024-010-03 Full Text: PDF