Per: matheus batista dos anjos (samarco), Anderson Christo Cunha (samarco), Fábio Ricardo Oliveira Bento (ifes), Marcos Ruy Soares Gaudio (edp), Maykcilane Fernandes Miguel (samarco), rafael damasceno xavier de brito (rafaeldamascenobrito@gmail.com)
Abstract:
The production of iron ore pellets is essential for the steel industry, as the performance of these pellets in customers' reactors and furnaces directly influences the quality of the final product. However, the production process faces challenges due to variations in ore quality, inputs, and other interference factors, resulting in inconsistencies in the pellets. This work aims to analyze the use of neural networks in controlling a pelletizing plant. This control is crucial due to variations in various process variables, such as the average furnace temperature, the physical and chemical characteristics of the ore, and the dosage of inputs. Even with rigorous control measures, such as regular testing of variables, errors can still occur due to this variability. Therefore, it is essential to develop a tool capable of analyzing the process and providing quick and accurate insights to prevent failures. This study highlights the potential of artificial neural networks (ANNs) in assessing the quality of continuous production processes, pointing to possible improvements in the quality of iron ore pellets and, consequently, significant benefits for the steel industry.