556
3.3 Geometallurgical model reconciliation
The best way to test the Geometallurgical Model created should be to perform a comparison between actual plant performance for each ore-type. Once the plant considered for the
methodology was commissioned during this study, the reconciliation could be affected. By
that, the Geometallurgical check was applied on a group of new samples tested on a lab scale,
through the same flowsheet. They were classified into the 4 ore-types, based on the clustering algorithm, and had their performance predicted, considering the regression equation for
their classification. The results between real metallurgical recovery and the predicted one
shows high correlation for the model created as presented in the figure.
The prediction quality was assessed with a positive correlation of 95% compared to the
virtual truth, with 91.54% of the results explained by the linear model created, that is, for
each expected recovery value, the equation (Y = 4.913 + 0.8662x) should be applied to correct
the predicted value. Therefore, it can be concluded that the model is valid to estimate samples
not used in its construction.
4 CONCLUSION
The methodology created to identify domains applying to metallurgical recovery shows effectiveness since it was able to establish the boundaries of the ore-types and had good correlation to the metallurgical recovery. This metallurgical forecast could act as an assumption
for the production planning, since it was able to identify the domain of worse metallurgical
recovery that should be restricted from the feeding of the beneficiation plant.
The combination of clustering analysis and Indicator Kriging can be applied to create and
model ore types for different process parameters like the work index, Soda consumption,%
fines or density moisture linking performance at the plant to its domain locations at the mine
and having an integrated quality control through the production chain.
REFERENCES
Arroyo Ortiz, C.E. Caracterização geometalúrgica e modelagem geoestatística da Mina de Brucutu—
Quadrilátero Ferrífero (MG). 2014. 206 f. Tese (Doutorado em Evolução Crustal e Recursos Naturais) – Escola de Minas, Universidade Federal de Ouro Preto, Ouro Preto, 2014.
Biondi, J.C. Processos Metalogenéticos e os depósitos brasileiros. São Paulo, Oficina de Textos, 2003.
528p.
Born, H., Kahn, H. Caracterização geológica e mineralógica voltada ao aproveitamento de jazimento
fosfáticos. In: Encontro Nacional De Rochas Fosfáticas, 5, São Paulo, 1990. Anais... São Paulo:
IBRAFOS, 1990.
Dunham, S., Vann, J. Geometallurgy, Geostatistics and Project Value—Does your block model tell you
what you need to know? In: Project Evaluation Conference, Melbourne, 2007. Proceedings… Melbourne: Australasian Institute of Mining and Metallurgy, 2007. p. 189–196.
Goovaerts, P. Geostatistics for natural resources evaluation. New York: Oxford University Press, 1997.
483 p.
Lishchuk, V. Geometallurgical Programs: critical evaluation of applied methods and techniques. 110
p. 2016. Thesis [Minerals and Metallurgical Engineering] – Luleå University of Technology, Lulea,
2016.
Metz, J. Interpretação de clusters gerados por algortimos de clustering hierárquico. Dissertação [Ciência da computação e matmática computacional] – Universidade de São Paulo, 2006.
Mitchell, T.M. (1997). Machine Learning. McGraw-Hill Series in Computer Science.
Schouwstra, R.P.; De Vaux, D.V.; Snyman, Q. Further development of a chemistry proxy for geometallurgical modelling at the Mogalakwena mine. J.S. Afr. Inst. Min. Metall., Johannesburg, v. 117, n. 7,
p. 719–726, July 2017.
3.3 Geometallurgical model reconciliation
The best way to test the Geometallurgical Model created should be to perform a comparison between actual plant performance for each ore-type. Once the plant considered for the
methodology was commissioned during this study, the reconciliation could be affected. By
that, the Geometallurgical check was applied on a group of new samples tested on a lab scale,
through the same flowsheet. They were classified into the 4 ore-types, based on the clustering algorithm, and had their performance predicted, considering the regression equation for
their classification. The results between real metallurgical recovery and the predicted one
shows high correlation for the model created as presented in the figure.
The prediction quality was assessed with a positive correlation of 95% compared to the
virtual truth, with 91.54% of the results explained by the linear model created, that is, for
each expected recovery value, the equation (Y = 4.913 + 0.8662x) should be applied to correct
the predicted value. Therefore, it can be concluded that the model is valid to estimate samples
not used in its construction.
4 CONCLUSION
The methodology created to identify domains applying to metallurgical recovery shows effectiveness since it was able to establish the boundaries of the ore-types and had good correlation to the metallurgical recovery. This metallurgical forecast could act as an assumption
for the production planning, since it was able to identify the domain of worse metallurgical
recovery that should be restricted from the feeding of the beneficiation plant.
The combination of clustering analysis and Indicator Kriging can be applied to create and
model ore types for different process parameters like the work index, Soda consumption,%
fines or density moisture linking performance at the plant to its domain locations at the mine
and having an integrated quality control through the production chain.
REFERENCES
Arroyo Ortiz, C.E. Caracterização geometalúrgica e modelagem geoestatística da Mina de Brucutu—
Quadrilátero Ferrífero (MG). 2014. 206 f. Tese (Doutorado em Evolução Crustal e Recursos Naturais) – Escola de Minas, Universidade Federal de Ouro Preto, Ouro Preto, 2014.
Biondi, J.C. Processos Metalogenéticos e os depósitos brasileiros. São Paulo, Oficina de Textos, 2003.
528p.
Born, H., Kahn, H. Caracterização geológica e mineralógica voltada ao aproveitamento de jazimento
fosfáticos. In: Encontro Nacional De Rochas Fosfáticas, 5, São Paulo, 1990. Anais... São Paulo:
IBRAFOS, 1990.
Dunham, S., Vann, J. Geometallurgy, Geostatistics and Project Value—Does your block model tell you
what you need to know? In: Project Evaluation Conference, Melbourne, 2007. Proceedings… Melbourne: Australasian Institute of Mining and Metallurgy, 2007. p. 189–196.
Goovaerts, P. Geostatistics for natural resources evaluation. New York: Oxford University Press, 1997.
483 p.
Lishchuk, V. Geometallurgical Programs: critical evaluation of applied methods and techniques. 110
p. 2016. Thesis [Minerals and Metallurgical Engineering] – Luleå University of Technology, Lulea,
2016.
Metz, J. Interpretação de clusters gerados por algortimos de clustering hierárquico. Dissertação [Ciência da computação e matmática computacional] – Universidade de São Paulo, 2006.
Mitchell, T.M. (1997). Machine Learning. McGraw-Hill Series in Computer Science.
Schouwstra, R.P.; De Vaux, D.V.; Snyman, Q. Further development of a chemistry proxy for geometallurgical modelling at the Mogalakwena mine. J.S. Afr. Inst. Min. Metall., Johannesburg, v. 117, n. 7,
p. 719–726, July 2017.