Application of artificial neural networks to high-throughput synthesis of zeolites

It is shown that multi-phase crystalline systems in zeolite synthesis can be modelled by using artificial neural networks (ANNs) and considering as input variables the molar compositions of the starting synthesis gel. Experimental data were obtained using high-throughput tools for synthesis of solid materials under hydrothermal conditions and following a multi-level factorial experimental design of the system TEA:SiO2:Na2O:Al2O3:H2O. The study of several neural networks resulted in a non-linear model able to predict the occurrence and crystallinity of zeolite beta and competing phases, being the predictions much better than those obtained by classical quadratic models.