Desenvolvimento de vidro cristal de nióbio guiado por inteligência artificial

Abstract

Utensils containing lead have been used for centuries, but the need to reduce exposure to this metal has encouraged the search for lead-free materials. Artificial neural networks (ANN) were used in this work as a tool in the development of new crystal glasses containing niobium instead of lead. Fixing certain parameters, such as refractive index, nd (≥ 1.55), glass transition temperature, Tg (≤ 570 °C), Abbe number, d (≤ 47), and the amount of the network-forming oxide SiO2 (≥ 60 mol%), tens of possible compositions were obtained. The most viable (eight, in total) were selected for synthesis and characterization. The glasses were prepared by melting the reagent mixtures (~ 1500 °C), and the refractive indices, glass transition temperature, chemical durability, and microhardness Vickers were estimated and compared to the values of a commercial lead-free crystal glass. The d values were only predicted established by the ANN. The values of nd and Tg predicted by the ANN were satisfactorily consistent with the experimental values measured. Among the prepared glasses, five presented values to nd similar or superior to the commercial sample and four of these presented Tg’s values just slightly higher than the commercial sample. All these glasses had a high degree of transparency and absence of color, with refractive indices between 1.558 and 1.634, glass transition temperature between 522 and 650 °C, and predicted Abbe number values between 40.9 and 54.4, low chemical solubility, and microhardness equivalent to the commercial sample.