Aplicações de técnicas de machine learning na metalurgia de aços: avanços e perspectivas

Abstract

Steel is one of the most widely used materials globally, with essential applications in the automotive, aerospace, and construction sectors due to its versatility, strength, and ability to adapt to different usage conditions. However, the complexity of metallurgical processes and the growing demand for increasingly specific performance have driven the search for tools that assist, especially, in the optimization of its mechanical and microstructural properties. In this context, machine learning (ML) techniques have emerged as an innovative approach to tackle such challenges, enabling the prediction and optimization of properties based on large volumes of experimental and computational data. The use of algorithms such as artificial neural networks, support vector machines, reinforcement learning, and genetic algorithms has made it possible to identify ideal chemical compositions and finely control processing parameters, aiming to maximize strength and ductility while reducing brittleness. This work aims to critically analyze the main applications of ML techniques in steel metallurgy, focusing on the prediction of mechanical properties, the development of new alloys, and the improvement of microstructures and metallurgical processes. To this end, a detailed literature review was conducted, mapping the techniques employed in recent studies and correlating each approach with its respective applications and outcomes. Based on the literature, it is evident that ML has already been widely and successfully used for performance prediction, composition optimization, and definition of operational parameters, promoting greater efficiency, resource savings, and technological innovation in the metallurgical sector. It is concluded that the integrated use of these techniques represents a promising path for the advancement of materials engineering, contributing significantly to the digitalization and modernization of the steel industry.