Título: Development of Soft Sensor for Mass Flow Rate Estimation in an Ore Reclaimer
Autores: Eduardo Filipe Pereira Maciel, Jose Perez, Paulo Marcos de Barros Monteiro & Thomas Vargas Barsante Pinto
Resumo: This paper presents the development and implementation of virtual sensors for estimating the reclaiming rate of a bucketwheel reclaimer in an iron ore processing context. The methodology was structured in two phases. In the first, multiple machine learning algorithms were evaluated for soft sensor modeling based on process variables collected from the field. In the second phase, the linear regression model—selected for its balance between accuracy and implementation simplicity—was integrated into a programmable logic controller (PLC) to enable real-time rate estimation. Data preprocessing techniques, including dead-time compensation and signal alignment, were applied to improve model accuracy. The models were assessed using RMSE, MAE, and correlation coefficient. The results demonstrated that the implemented soft sensor is capable of reliably reproducing belt scale measurements, even under variable operating conditions. These findings reinforce the applicability of data-driven models in enhancing monitoring and control strategies in mining automation systems.
Palavras-chave: Soft sensors; machine learning; mineral processing; mass flow estimation.
Páginas: 7
Código DOI: 10.21528/CBIC2025-1174129
Artigo em PDF: CBIC_2025_paper1174129.pdf
Arquivo BibTeX:
CBIC_2025_1174129.bib