Effect of ingredients and experimental parameter on the rheological and mechanical properties of self-compacting agro-geopolymer paste: Experimental and autoencoder based deep learning techniques

Cement is one of the best binder that is used in the construction industry, but it is neither a sustainable nor an economical choice. Cement manufacture industry has a major contribution to the anthropogenic carbon dioxide emissions that leads to increase in carbon footprint. So here comes the importance of finding an alternative to cement, which is both sustainable and economical. Geopolymers are such inorganic aluminosilicate polymers which gain its strength by polymerization. This study aims to investigate the influence of ingredients and experimental parameters on the various rheological and mechanical properties of self-compacting geopolymer paste. It was found that the testing temperature is the most influencing factor for the variation of rheology of geopolymer paste and the mechanical properties were adversely affected with the addition of superplasticizer (SP). As duration after mixing increases, the mix losses its plasticity and starts to set, hence the mix becomes more viscous. An autoencoder based deep learning algorithm is used to model and predict the rheological (mini slump, viscosity) and mechanical properties of geopolymer mixes with different SP dosages. In total, there were 330 data which is the database for training, validation and testing. Out of the 330 data sets, 270 were used for training, 30 for validation and 30 for testing the algorithm. The predicted and measured data were compared and accuracy was checked. A good correlation was obtained between the predicted and measured data. • This study aims to investigate the influence of ingredients and experimental parameters on the various rheological and mechanical properties of self-compacting geopolymer paste. • It was found that the testing temperature is the most influencing factor for the variation of rheology of geopolymer paste and the mechanical properties were adversely affected with the addition of superplasticizer. • The viscosity of the mixture was assessed over time, and it was observed that as the duration after mixing increased, the viscosity also increased due to the loss of plasticity as the mixture began to set. • An autoencoder based deep learning algorithm is used to model and predict the rheological (mini slump,viscosity) and mechanical properties of geopolymer mixes with different SP dosages. A good correlation was obtained between the predicted and measured data