APLICAÇÃO DE REDES AUTOENCODERS VARIACIONAIS CONDICIONAIS PARA O AUMENTO DA BASE DE ESPECTROS DE RAMAN DE AMOSTRAS DE SÍNTER

Abstract

Recently, driven by the emergence of Smart Factories, transformations in the industrial landscape have been significant, with advancements in precise material characterization through advanced analytical tools such as Raman Spectroscopy. This study focused on the application of conditional variational autoencoder networks to expand the dataset of Raman spectra from the sinter produced in the steel manufacturing process. The results demonstrated the methodology’s ability to reproduce essential characteristics of the original spectra, although there was no improvement in basicity prediction using a kNN algorithm, suggesting the need for further investigation to optimize these methods in the steel industry.