Portable X-ray fluorescence sensor for ecofriendly, low-cost, and fast assessment of eucalypt charcoal attributes

Brazilian steel industries require high-quality charcoal to produce pig iron. Desirable charcoal attributes include high elemental carbon content, large mean particle size (MPS), and high density, while producing low contents of ash and volatile matter, and presenting low contents of water and conta...

Descripción completa

Detalles Bibliográficos
Autores: Andrade, R., Benedet, L., Mancini ,M., Silva, S. H. G., Freitas, C. S., Carneiro, M. A. C., Curi, N.
Tipo de recurso: artículo
Estado:Versión publicada
Fecha de publicación:2025
País:Brasil
Institución:Universidade Federal de Lavras (UFLA)
Repositorio:Repositório Institucional da UFLA
Idioma:inglés
OAI Identifier:oai:repositorio.ufla.br:1/60151
Acceso en línea:https://repositorio.ufla.br/handle/1/60151
https://doi.org/10.1590/1413-7054202549026424
Access Level:acceso abierto
Palabra clave:Proximal sensors
machine learning algorithms
metallurgy
pXRF
waste minimization; Sensores proximais; pXRF; aprendizado de máquina; metalurgia; minimização de resíduos
Descripción
Sumario:Brazilian steel industries require high-quality charcoal to produce pig iron. Desirable charcoal attributes include high elemental carbon content, large mean particle size (MPS), and high density, while producing low contents of ash and volatile matter, and presenting low contents of water and contaminants (e.g., phosphorous). These attributes are commonly determined by standardized laboratory analyses, which are time consuming and costly, besides generating chemical effluents. Portable X-ray fluorescence (pXRF) spectrometry can be used to avoid the downsides of laboratory analyses. The objective of this study was to evaluate the use of pXRF data in machine-learning models trained to predict attributes of eucalypt charcoal. pXRF data (elemental contents) from 276 charcoal samples were used to train predictive models using six machine-learning algorithms. Auxiliary explanatory variables (drying time, wood age, fine particle content, and friability) were included in the models. Models were trained to predict the following charcoal attributes: fixed C (%), ash content (%), volatile matter (%), MPS (mm), water content (%), density (kg/m3), and P contents (%). Satisfactory predictions were obtained for volatile matter, MPS, moisture, and density (R2 > 0.6), and very accurate predictions were obtained for ash and P contents (R2 > 0.75). The inclusion of auxiliary explanatory variables increased the prediction accuracy of MPS (R2 increased from 0.61 to 0.82), bulk density (from 0.56 to 0.73), and P contents (from 0.86 to 0.94). These results indicate that pXRF can be used as an ecofriendly alternative to assess the quality of eucalypt charcoal utilized in metallurgy. © 2025, Federal University of Lavras. All rights reserved.