Prediction of Sulfur in the Hot Metal based on Data Mining and
Artificial Neural Networks
Wandercleiton Cardoso
a
and Rendo di Felice
b
Dipartimento di Ingegneria Civile, Chimica e Ambientale (DICCA), Università degli Studi di Genova,
Via All'Opera Pia, 15, CAP 16145, Genova (GE), Italy
Keywords: Big Data, Machine Learning, Blast Furnace, Sulfur, Industry 4.0.
Abstract: In recent years, interest in artificial intelligence and the integration of Industry 4.0 technologies to improve
and monitor steel production conditions has increased. In the current scenario of the world economy, where
the prices of energy and inputs used in industrial processes are increasingly volatile, strict control of all stages
of the production process is of paramount importance. For the steel production process, the temperature of
the metal in the liquid state is one of the most important parameters to be evaluated, since its lack of control
negatively affects the final quality of the product. Every day, several models are proposed to simulate
industrial processes. In this sense, data mining and the use of artificial neural networks are competitive
alternatives to solve this task. In this context, the objective of this work was to perform data mining in a Big
Data with more than 300,000 pieces of information, processing them using an artificial neural network and
probabilistic reasoning. It is concluded that data mining and neural networks can be used in practice as a tool
for predicting and controlling impurities during the production of hot metal in a blast furnace.
1 INTRODUCTION
The blast furnace is a chemical-metallurgical reactor
used to produce molten iron, which is the product
formed by the reduction of metallic oxides that
chemically react with reducing elements such as
carbon monoxide (CO) and hydrogen gas (H
2
)
(Chizhikova and Best, 2020).
Blast furnaces are chemical metallurgical reactors
for the production of pig iron and slag. Pig iron is
obtained in a liquid state and consists of iron (92 to
95%), carbon (3 to 4.5%) and impurities such as
sulphur, phosphorus and silica (Arif and Ahmad,
2021).
The raw materials used (metallic feedstock) are
sinter, granulated ore and pellets. The main fuel is
metallurgical coke. All these materials are loaded
through the upper part of the reactor, with hot air
blown into the lower section (Zhao et al., 2020).
The injected hot air gasifiers the coke and
produces CO reducing gas and a large amount of heat
that rises upwards in counter current to the descent of
the charge, providing heating, reduction and melting
a
https://orcid.org/0000-0001-8531-4049
b
https://orcid.org/0000-0002-8169-3325
of the metallic charge. Pulverized coal is used as an
additional fuel, which is blown in together with hot
air (Blotevogel, 2021).
The preheated air with a temperature of about
1200°C is blown through the blast tuyeres of the blast
furnace and comes into contact with the coke in the
raceway area. The contact of the oxygen in the air
with the carbon of the coke heated to 1500°C first
leads to a reaction that produces carbon dioxide (CO
2
)
(Zhang et al., 2019a).
This highly exothermic reaction generates a large
amount of heat for the process. The carbon dioxide
immediately reacts with the carbon in the coke to
form carbon monoxide (CO), according to the loss-
of-solution or Boudouard reaction (C + CO
2
→ 2CO),
which is very endothermic (Cardoso et al., 2021b).
The moisture contained in the injected air reacts
with the carbon in the coke to produce the reducing
gases CO and H
2
. Although these reactions are
endothermic, i.e. proceed under heat absorption, the
exit of the reducing gases from the duct effectively
results in a high heat input into the process, producing
400
Cardoso, W. and di Felice, R.
Prediction of Sulfur in the Hot Metal based on Data Mining and Artificial Neural Networks.
DOI: 10.5220/0011276700003269
In Proceedings of the 11th International Conference on Data Science, Technology and Applications (DATA 2022), pages 400-407
ISBN: 978-989-758-583-8; ISSN: 2184-285X
Copyright
c
 2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

flame temperatures in excess of 2000°C (Fontes et al.,
2020; Cardoso et al., 2021a; Cardoso et al., 2022).
On the rest of the way through the furnace, the
rising gas gives off heat to the descending metal
layers and leaves the furnace with temperatures in the
order of 100 to 150°C (Kurunov, 2019).
Due to the different heat requirements for a
number of chemical reactions taking place at different
levels in the furnace, the temperature profile takes on
a characteristic shape: an upper preheating zone (0-
800ºC) separated from a lower melting zone (900-
1500ºC) and a vertical thermal reserve zone whose
temperature is in the range of 800-1000ºC (Ibragimov
et al., 2019; Kong et al., 2021).
The thermal reserve zone, where there is little heat
exchange between gas and solids, occupies 40-50%
of the total height of the furnace (Kurunov, 2018).
The nature of the countercurrent process allows a
highly reducing gas (high content of CO) to contact
the metallic mineral wustite, which has the lowest
oxygen potential of the iron oxides, and then hematite
and magnetite in the upper zone to be reduced by a
gas with a lower reduction potential (Li et al., 2021a).
Since CO
2
is the end product of carbon
combustion, the more oxygen that is removed, the
more complete the utilisation of the thermal and
chemical energy of the carbon (Li et al., 2021b).
These reactions are called indirect reduction, and
the overall reaction is slightly exothermic. If some of
the wustite remains unreduced, it is further reduced
by direct reduction in the range where temperatures
exceed 1000°C (Matino et al., 2019a).
The high temperature ramp gas generated in the
combustion zone (the tuyeres region) causes heating
of the charge, decomposition reactions and reduction
of oxides during its ascent in the blast furnace. As a
result, the temperature of the gas gradually decreases
while its chemical composition changes (Muraveva et
al., 2021; Pavlov et al., 2019).
First, near the charge level, the charge undergoes
moisture evaporation and preheating. When the
charge decreases, the reduction of iron oxides takes
place. In the softening and melting zone, in the area
of the lower vat and the belly, begins the softening
and melting of the charge, which develops to the
crucible (Cardoso et al., 2022).
The pig iron (hot metal) and slag that are in the
crucible are removed at controlled intervals through
the running holes. In the area of the tuyeres, the coke
gradually decreases in size as it burns (Rasul et al.,
2007; Saxén and Pettersson, 2007).
Together with the fusion of the materials that
make up the charge, this causes the level in the blast
furnace to drop, so that a new charge has to be
conveyed at the top (Semenov et al., 2020).
Coke is considered the permeabilizer of the blast
furnace charge. This role cannot be assumed by any
other fuel, as coke is the only material capable of
maintaining the permeability of the bed to the
ascending gas, as well as that of the descending liquid
slag and hot metal (Tan et al., 2020).
Coke remains solid under the high-temperature
conditions prevailing in the oven and maintains levels
of resistance to the different stresses it undergoes
inside the oven. This allows it to maintain a suitable
size and size distribution for good permeability,
without which the manufacture of pig iron in a blast
furnace would be impossible (Cardoso el al., 2022).
However, the thermal and chemical roles can be
played, in part, by other liquid fuels (petroleum fuel
oil and coal tar), gaseous with high calorific value
(reducing gas, natural gas, and coke oven gas) or
solids (mainly, mineral coal), injected through the
tuyeres of the kiln. Thus, these auxiliary fuels also
participate as sources of heat and reducing gas for the
process. Figure 1 illustrates the working principle of
a blast furnace:
Figure 1: Blast furnace working principle.
Blast furnace monitoring is of paramount
importance in the production of a quality product.
Sulfur in steel is an undesirable residue that
Prediction of Sulfur in the Hot Metal based on Data Mining and Artificial Neural Networks
401

negatively affects properties such as ductility,
toughness, weldability and corrosion resistance.
In recent years, the demand for steels with higher
toughness and ductility has increased, and low sulfur
levels are important to achieve these properties.
Furthermore, sulfur may play an important role in
some corrosion processes in steel.
Therefore, in the production of steel for the pipe
and automotive industries, for example, the control of
the sulfur content is fundamental. The production of
low sulfur steel is of utmost importance for
shipbuilding and pipelines for the oil industry. This
requires high production control in the blast furnace
and an efficient desulfurization process at the lowest
possible cost.
In the field of technology and modelling, in
addition to predicting the effects of changes in
production parameters, several blast furnace
simulation models have been developed with the aim
of improving production conditions, including two-
and three-dimensional models that allow progress and
detailed information on fluid flow and mass and heat
balances within the blast furnace.
Considering the existing difficulties in the field of
simulation of complex processes, the application of
solutions based on neural networks has gained space
due to its diversity of application and increase in the
reliability of responses, since the neural network
receives new data in the operating process/training
without necessarily drawing conclusions about values
or types of interaction between raw materials for the
use of neural models.
In computer science and related fields, artificial
neural networks are computational models inspired
by an animal's central nervous system (in particular
the brain) that are capable of performing machine
learning as well as pattern recognition. Artificial
neural networks are generally presented as systems of
"interconnected neurons, which can compute input
values", simulating the behaviour of biological neural
networks. Figure 2 illustrates an artificial neural
network.
Figure 2: Artificial neural network.
The objective of this work is to mine a database
and numerically simulate an artificial neural network
with 25 neurons in the hidden layer.
2 RESEARCH METHOD
The database used for numerical simulation
corresponds to 11 years of reactor operation. Big Data
contains 301,125 pieces of information divided into
75 variables. The neural network input is composed
of 74 input variables and 1 output variable.
The artificial neural network has a structure
similar to Figure 3 with a simple layer and 25 neurons
in the hidden layer, using the Levenberg-Marquardt
training algorithm, and a sigmoid activation function.
Figure 3: Artificial neural network architecture.
According to the literature, 85% of the database
should be used to train and validate the neural
network and the remaining 15% will be used to test
the model's predictive capacity during the test step.
Table 1 illustrates the database division. Table 2 to 8
illustrates the input variable and Table 9 illustrates the
output variable
Table 1: Division of samples.
Step Samples
Training 210.789
Validation 45.168
Test 45.168
Table 2: Blast Furnace Gas.
Variable Mean Std_dev
CO (%) 23.8 0.74
CO
2
(%) 24.3 0.66
N
2
(%) 47.2 1.39
H
2
(%) 4.50 0.43
CO + CO
2
47.9 0.6
CO efficiency (%) 49.5 0.85
H
2
efficiency (%) 40.7 3.25
DATA 2022 - 11th International Conference on Data Science, Technology and Applications
402

Table 3: Hot metal.
Variable Mean Std_dev
Estim. Production
(
ton
)
7789.5 314.5
Real production (ton) 7787.2 324.5
Carbon (%) 4.635 0.169
Chrome (%) 0.025 0.002
Co
pp
e
r
(
%
)
0.007 0.001
Man
g
anese
(
%
)
0.29 0.03
Mn ratio
(
-
)
0.13 0.22
Table 4: Slag.
Variable Mean Std_dev
Sla
g
rate
(
k
g
/ton
)
246.99 13.74
B2 basicit
y
(
-
)
1.2 0.04
B4 basicit
y
(
-
)
1.07 0.04
Al
2
O
3
(%) 10.71 0.62
CaO (%) 43.06 1.55
Sulfu
r
(
%
)
1.15 0.14
FeO
(
%
)
0.42 0.04
M
g
O
(
%
)
6.83 0.86
MnO (%) 0.31 0.1
SIO
2
(%) 36.05 1.36
TIO
2
(%) 0.58 0.05
Production
(
ton
)
1980.6 190.8
Mn ratio
(
-
)
0.87 0.22
Table 5: Fuel.
Variable Mean Std_dev
Injection PCI (kg/ton) 58.99 6.16
Gas rate (kg/ton) - -
Coal/O
2
tax (-) 755.27 75.57
Coal/air tax (-) 170.03 74.12
PCI rate 175.98 15.61
Direct reduction (%) 23.38 12.41
PCI tax (kg/ton) 1078.3 540.9
Coke total (kg/ton) 1932.2 911.7
Small coke (kg/ton) 294.63 134.86
Coke 1 (kg/ton) 210.7 259.8
Coke 2 (kg/ton) 742 716
Coke 3 (kg/ton) 946 956
Coke 4 (kg/ton) 1878 143
Coke 5 (kg/ton) 1327.5 847.6
Moisture (kg/ton) 6.4 1.41
Coke/load (kg/ton) 11.89 9.56
Small coke total (kg/ton) 4.28 0.03
PCI/load (kg/ton) 174.74 14.32
Fuel rate/load (kg/ton) 484.08 18.14
Coke total/load (kg/ton) 24.52 0.89
PCI/day (-) 1214.4 44.9
Coke rate (kg/ton) 319.68 25.98
Table 6: Thermal control.
Variable Mean Std_dev
Hot metal
(
°C
)
1508.3 12.2
Blowing ai
r
(°C) 1243.3 13.9
Top gas (°C) 121,35 10,34
Flame temperature (°C) 2177.6 2108
Sla
g
1508.3 12.2
Thermal index
(
-
)
504.7 54.03
Table 7: Minerals.
Variable Mean Std_dev
Ore/Coque (-) 5.1 0.31
Sinte
r
1
(
ton
)
4536.3 884.2
Sinte
r
2
(
ton
)
1697.2 1326.2
Pellet 1
(
ton
)
5132 1898.3
Pellet 2 (ton) 4813.7 2183.1
Total metal loa
d
(ton) 12312 670
Raw material rate 1578.8 15.1
Ore
(
%
)
8.9 4.5
Sinter
(
%
)
39.6 2.8
Pellet (%) 51.5 5.1
Ore (day) 12747 703
Table 8: Blow air.
Variable Mean Std_dev
Volume
(
Nm
3
/min
)
4852.9 148.6
Pressure (Kgf/cm
2
) 3.87 0.1
Moisture (kg/m
3
) 19.81 3.73
O2 enrichment
(
%
)
5.27 0.95
Stea
m
(
%
)
1.51 1.01
Comsum
p
tion
(
Nm
3
/min
)
7030.3 213.6
Table 9: Sulfur output (%).
Mean 0.023
Standard deviation 0.008
Minimu
m
0.008
Median 0.021
Maximu
m
0.083
Skewness 1.5
Kurtosis 4.8
The method used to evaluate the quality of the
neural network model was the RMSE (root mean
square error). Small values close to zero indicate
better predictive capacity of the model. Pearson's
mathematical correlation coefficient (R) was also
used to validate the mathematical models.
RMSE=



∑(
C

−C

)



(1)
Prediction of Sulfur in the Hot Metal based on Data Mining and Artificial Neural Networks
403

R=


(
C

−C

)




(
C

−C

)




(2)
The RMSE mathematical equation is presented in
Eq(1) and Pearson's mathematical correlation
coefficient is presented in Eq(2).
3 RESULTS AND DISCUSSION
The sulfur prediction model presented greater
difficulty in the prediction, but even so, the result was
excellent compared to the literature. There was no
evidence of overfitting or underfitting in the hot metal
sulfur prediction model.
There was no discrepancy between the RMSE
values for the training, validation and testing phases.
The artificial neural network required a maximum of
687 epochs to converge the model, indicating greater
complexity to stabilize the error values.
The mathematical correlation (R) and the RMSE
of the artificial neural network is shown in Table 10 e
Table 11 and confirms an excellent correlation value.
Figure 4 illustrates the dispersion between the values
calculated by the neural network and the values of Big
Data.
Table 10: Root Mean Square Error.
Overall 0.0027
Training 0.0031
Validation 0.0030
Testin
g
0.0031
Table 11: Pearson’s correlation coefficient.
Overall 0.9632
Training 0.9682
Validation 0.9660
Testin
g
0.9373
Figure 4: Scatterplot sulfur.
From a metallurgical point of view Most of the
sulfur contained in hot metal (about 80%) is
introduced into the blast furnace by the metallurgical
coke in the form of iron sulfide (FeS) and calcium
sulfide (CaS) contained in the coke ash, and as
organic sulfur. The rest comes via the other materials
in the metallic charge and the fluxes.
In blast furnaces fed with metallurgical coke,
about 90 to 80% of the sulfur is part of the chemical
composition of the slag, while 10% to 15% is
precipitated with the blast furnace gas and values
between 2% and 5% dissolve in the hot metal. Recent
studies have shown that the main mechanism of sulfur
reactions is very similar to that of silicon.
Small amounts of sulfur are also absorbed by the
slag in the area of the blast furnace channel. It should
also be mentioned that sulfur forms other compounds
such as sulfur dioxide (SO
2
) and carbon disulfide
(CS), which are also transported by the gas stream
and undergo chemical reactions. However, it must be
emphasized that among sulfur gasses, silicon sulfide
(SiS) is the dominant species.
In the blast furnace raceway, the sulfur produced
reacts with calcium (Ca), silicon (Si), and iron (Fe)
according to Eq.3, Eq. 4, and Eq. 5. The chemical
reaction that removes sulfur from the hot metal is
often represented by Eq. 6.
CaS
( )
+ SiO
(
g
)
→ SiS
(
g
)
+ CaO (3)
FeS+SiO+C→ SiS+CO+Fe (4)
SiS(g) → [Si] + [S] (5)
S + (CaO) + C → (CaS) + CO(g) (6)
The transfer of sulfur from the gasses to the hot
metal takes place in an area of the blast furnace
known as the dripping zone. Inside the blast furnace,
in the softening and melting zone, when silicon and
sulfur-containing hot metal droplets percolate
through the slag, in the absence of MnO, the silicon
content of the hot metal increases and no transfer of
sulfur occurs.
However, in the presence of MnO, silicon is
removed from the hot metal and the transfer of
manganese from the slag to the metal occurs together
with the transfer of sulfur from the metal to the slag.
Literature states that gasses such as SiO and SiS are
formed in the beneficiation zone, while silicon and
sulfur are transferred to the hot metal in the
combustion zone.
Silicon is reduced by FeO and MnO and dissolved
in the slag while the hot metal droplets penetrate the
slag layer. It should be mentioned that CaO has a
much greater desulfurization potential than MgO,
DATA 2022 - 11th International Conference on Data Science, Technology and Applications
404

about 100 times stronger. It should also be noted that
for proper desulfurization, the oxygen content of the
metal must be very low.
This is possible due to the reaction of oxygen
dissolved in the hot metal with strong oxide forming
elements. Thus, if considering the oxide-forming
elements, may write the following reactions given in
Eq. 7, Eq. 8, and Eq. 9.
(CaO) + [S] + [C] →(CaS) + CO
(
g
)
(7)
(CaO) + [S] + [Mn] →(CaS) + (MnO) (8)
(CaO) + [S] + 0,5[Si] →(CaS) + 0,5(SiO

) (9)
The increase of the basicity of the slag increases
the CaO activity and favors desulfurization. Low FeO
concentrations in the slag favor the incorporation of
sulfur into the hot metal. Carbon and silicon dissolved
in the hot metal favor desulfurization by increasing
the thermodynamic activity of sulfur in the slag.
Considering all this information and the fact that
sulfur is an extremely harmful chemical element for
steel, the use of a ANN to predict sulfur in metal
production is justified.
4 CONCLUSIONS
Regarding simulation methods for predicting process
variables, the increasing development of
computational capacity, leading to cheaper and more
powerful equipment, is driving the development of
more complex algorithms with better results, such as
neural networks.
Thus, the progress in computational capacity
enabled the development of different types of
simulation models.
This is one of the factors that enabled the use of
the ANN model in this study, as well as the
identification of the main variables that affect the
model.
The processing of the data to be used for the
development of the model is highlighted as an
important part of the process, which is sometimes a
slow process since the information must be evaluated
to find the best way to identify outliers for the
development of models.
However, it is important to emphasize that the
model in isolation may predict good results for each
of the target variables. It should be noted that the use
of the above modelling technique has enabled the
construction of higher fidelity models that may be
used as tools for decision making and operational
planning related to fuel economy, operational
stability, and delivery of a product for steelmaking
and help improve process monitoring.
In short, neural networks may be used in practice
because the model is both a predictive tool and a
guide for operation due to the excellent correlations
between the real values and the values computed by
the neural network.
REFERENCES
Arif, M. S.; Ahmad, I. Artificial Intelligence Based
Prediction of Exergetic Efficiency of a Blast Furnace.
Computer Aided Chemical Engineering, v.50, pp.
1047-1052, 2021.
Bai, Y.; Tan, M. Dynamic committee machine with fuzzy-
C-means clustering for total organic carbon content
prediction from wireline logs. Computers &
Geosciences, v.146, pp. 104626, 2021.
Blotevogel, S. Glass structure of industrial ground
granulated blast furnace slags (GGBS) investigated by
time-resolved Raman and NMR spectroscopies. Journal
of Materials Science, v.56, pp. 17490-17504, 2021
Cardoso, W.; Baptista, R.; di Felice, R. Artificial Neural
Networks for Modelling and Controlling the Variables
of a Blast Furnace. In IEEE 6th International Forum on
Research and Technology for Society and Industry
(RTSI), pp. 148-152, 2021b.
Cardoso, W.; Baptista, R.; di Felice, R. Artificial neural
network for predicting silicon content in the hot metal
produced in a blast furnace fueled by metallurgical
coke. Materials Research, v.25, pp. 20210439, 2022,
Cardoso, W.; Baptista, R.; di Felice, R. Mathematical
modeling of a solid oxide fuel cell operating on biogás.
Bulletin of Electrical Engineering and Informatics,
v.10, n.06, pp. 2929-2942, 2021c.
Cardoso, W.; Barros, D.; Baptista, R.; di Felice, R.
Mathematical Modelling to Control the Chemical
Composition of Blast Furnace Slag Using Artificial
Neural Networks and Empirical Correlation. IOP
Conference Series: Materials Science and Engineering,
v.1203, pp. 032096, 2021a.
Cardoso, W.; di Felice, R. Prediction of silicon content in
the hot metal using Bayesian networks and probabilistic
reasoning. International Journal of Advances in
Intelligent Informatics, v.07, n.03, pp. 268-281, 2021d.
Carro, K. B.; Leite, G. R.; Oliveira, A. G.; Santos, C. B.;
Pinto, I. S.; Fux, B.; Falqueto, A. Assessing geographic
and climatic variables to predict the potential
distribution of the visceral leishmaniasis vector
Lutzomyia longipalpis in the state of Espı
́
rito Santo,
Brazil. Plos One, v.15, n.9, e0238198, 2020.
Chen, M.; Wan, X.; Shi, J.; Taskinen, P.; Jokilaakso, A.
Experimental Study on the Phase Relations of the SiO
2
-
MgO-TiO
2
System in Air at 1500°C. JOM, v.74, pp.
676–688, 2022
Chizhikova, V. M. Best Available Techniques in the Blast-
Furnace Production. Metallurgist, v.64, pp. 13-35,
2020.
Prediction of Sulfur in the Hot Metal based on Data Mining and Artificial Neural Networks
405

Ducic, N.; Jovicic, A.; Manasijevic, S.; Radisa, R.;
Cojbasic, Z.; Savković, B. Application of Machine
Learning in the Control of Metal Melting Production
Process. Applied Sciences, v.10, n.17, pp. 6048-6063,
2020.
Fontes, D. O. L.; Vasconcelos, L. G.; Brito, R. P. Blast
furnace hot metal temperature and silicon content
prediction using soft sensor based on fuzzy C-means
and exogenous nonlinear autoregressive models.
Computers & Chemical Engineering, 141, 107028,
2020.
He, F.; Zhang, L. Prediction model of end-point phosphorus
content in BOF steelmaking process based on PCA and
BP neural network. Journal of Process Control, v.66,
pp. 51-58, 2018.
Hou, Y.; Wu, Y.; Liu, Z.; Han, H.; Wang, P. Dynamic
multi-objective differential evolution algorithm based
on the information of evolution progress. Science China
Technological Sciences, v.64, n.08, pp. 1676–1689,
2021.
Ibragimov, A. F.; Iskhakov, I. I.; Skopov, G. B.;
Kirichenko, A. N. Using Oxygen-Enriched Blast
During the Operation of Shaft Furnaces of the
Mednogorsk Copper–Sulfur Combine LLC.
Metallurgist, v.63, pp. 62-69, 2019.
Jantre, S. R.; Bhattacharya, S.; Maiti, T. Quantile
Regression Neural Networks: A Bayesian Approach.
Journal of Statistical Theory and Practice, v.15, n.03,
pp. 01-31, 2021.
Jiang, Y.; Zhou, P.; Yu, G. Multivariate molten iron quality
based on improved incremental Randon vector
functional-link networks. IFAC PapersOnLine, pp.
290-294, 2018.
Kang, Y. B. Progress of Thermodynamic Modeling for
Sulfide Dissolution in Molten Oxide Slags: Sulfide
Capacity and Phase Diagram. Metallurgical and
Materials Transactions B, v.52, n.05, pp. 2859-2882,
2021.
Kina, C.; Turk, K.; Atalay, E.; Donmez, I.; Tanyildizi, H.
Comparison of extreme learning machine and deep
learning model in the estimation of the fresh properties
of hybrid fiber-reinforced SCC. Neural Computing and
Applications, v.33, n.18, pp. 11641-11659, 2021.
Kong, W.; Liu, J.; Yu, Y.; Hou, X.; He, Z. Effect of
w(MgO)/w(Al
2
O
3
) ratio and basicity on microstructure
and metallurgical properties of blast furnace slag.
Journal of Iron and Steel Research International, v.28,
n.10, pp. 1223-1232, 2021.
Kurunov, I. F. Ways of Improving Blast Furnace Smelting
Efficiency with Injection of Coal-Dust Fuel and Natural
Gas. Metallurgist, v.61, n.09, pp. 736–744, 2018.
Li, J.; Hua, C.; Qian, J.; Guan, X. Low-rank based Multi-
Input Multi-Output Takagi-Sugeno fuzzy modeling for
prediction of molten iron quality in blast furnace. Fuzzy
Sets and Systems, v. 421, pp. 178-192, 2021.
Li, W.; Zhuo, Y.; Bao, J.; Shen, Y. A data-based soft-sensor
approach to estimating raceway depth in ironmaking
blast furnace. Powder Technology, v.390, pp. 529-538,
2021.
Li, Y.; Zhang, J.; Zhang, S.; Xiao, W. Dual ensemble online
modeling for dynamic estimation of hot metal silicon
content in blast furnace system. ISA Transactions,
2022, article in press.
Liang, W.; Wang, G.; Ning, X.; Zhang, J.; Li, Y.; Jiang, C.;
Zhang, N. Application of BP neural network to the
prediction of coal ash melting characteristic
temperature. Fuel, v.260, 116324, 2020.
Liu, Y.; Wang, Y.; Chen, L.; Zhao, J.; Wang, W.; Liu, Q.
Incremental Bayesian broad learning system and its
industrial application. Artificial Intelligence Review,
v.54, n.05, 2021.
Matino, I.; Dettori, S.; Colla, V.; Weber, V.; Salame, S.
Two innovative modelling approaches in order to
forecast consumption of blast furnace gas by hot blast
stoves. Energy Procedia, v.158, pp. 4043-4048, 2019a.
Matino, I; Dettori, S; Colla, V; Weber, V.; Salame, S.
Application of Echo State Neural Networks to forecast
blast furnace gas production: pave the way to off-gas
optimized management, Energy Procedia, v.158, pp.
4037-4042, 2019b.
Matino, I; Dettori, S; Colla, V; Weber, V.; Salame, S.
Forecasting blast furnace gas production and demand
through echo state neural network-based models: Pave
the way to off-gas optimized management, Applied
Energy, v.253, pp. 113578, 2019c.
Mhaya, A. M.; Huseien, G. F.; Faridmehr, I.; Abidin, A. R.;
Alyousef, R.; Ismail, M. Evaluating mechanical
properties and impact resistance of modified concrete
containing ground Blast Furnace slag and discarded
rubber tire crumbs. Construction and Building
Materials, v. 295, pp. 123603, 2021.
Muchnik, D. A.; Trikilo, A. I.; Lyalyuk, V. P.; Kassim, D.
A. Coke Quality and Blast-Furnace Performance. Coke
and Chemistry, v.61, pp. 12-18, 2018.
Muraveva, I. G.; Togobitskaya, D. N.; Ivancha, N. G.;
Bel’kova, A. I.; Nesterov, A. S. Concept Development
of an Expert System for Selecting the Optimal
Composition of a Multicomponent Blast-Furnace
Charge and Functional and Algorithmic Structure. Steel
in Translation, v.51, pp. 33-38, 2021.
North, L.; Blackmore, K.; Nesbitt, K.; Mahoney, M. R.
Methods of coke quality prediction: A review. Fuel,
v.219, pp. 426-445, 2018.
Oliveira, A. G.; Totola, L. B.; Bicalho, K. V.; Hisatugu, W.
H. Prediction of compression index of soft soils from
the Brazilian coast using artificial neural networks and
empirical correlations. Soils and Rocks, v.43, pp. 109-
121, 2020.
Pandey, T. N.; Jagadev, A. K.; Dehuri, S.; Cho, S. B. A
novel committee machine and reviews of neural
network and statistical models for currency exchange
rate prediction: An experimental analysis. Journal of
King Saud University - Computer and Information
Sciences, v.32, n.9, pp. 987-999, 2020.
Pavlov, A. V.; Polinov, A. A.; Spirin, N. A.; Onorin, O. P.;
Lavrov, V. V.; Gurin, I. A. Decision-Making Support
in Blast-Furnace Operation. Steel in Translation, v.49,
n.3, pp. 185-193, 2019.
DATA 2022 - 11th International Conference on Data Science, Technology and Applications
406

Quesada, D.; Valverde, G.; Larrañaga, P.; Bielza, C. Long-
term forecasting of multivariate time series in industrial
furnaces with dynamic Gaussian Bayesian networks,
Engineering Applications of Artificial Intelligence,
v.103, pp. 104301, 2021.
Radhakrishnan, V. R.; Mohamed, A. R. Neural networks
for the identification and control of blast furnace hot
metal quality. Journal of Process Control, v.10, n.6, pp.
509-524, 2000.
Rasul, M. G.; Tanty, B. S.; Mohanty, B. Modelling and
Analysis of Blast Furnace Performance for Efficient
Utilization of Energy. Applied Thermal Engineering,
v.27, n.01, pp. 78-88, 2007.
Reynolds, Q. G.; Rhamdhani, M. A. Computational
Modeling in Pyrometallurgy: Part I. JOM, v.73, n.9, pp.
2658-2659, 2021.
Rhamdhani, M. A.; Reynolds, Q. G. Computational
Modeling in Pyrometallurgy: Part II. JOM, v.73, n.9,
pp. 2885-2887, 2021.
Saxén, H.; Pettersson, F. Nonlinear Prediction of the Hot
Metal Silicon Content in the Blast Furnace. ISIJ
International, 47, n.12, pp. 1732-1737, 2007.
Semenov, Y. S.; Gorupakha, V. V.; Kuznetsov, A. M.;
Semion, I. Y.; Schumel’chik, E. I.; Vashchenko, S. V.;
Khudyakov, A. Y. Experience of Using Manganese-
Containing Materials in Blast-Furnace Charge.
Metallurgist, v.63, n.9, pp. 1013-1023, 2020.
Sohn, S. K. I. Application of complex systems topologies
in artificial neural networks optimization: An overview,
Expert Systems with Applications, v.180, pp. 115073,
2021.
Souza, P. V. C. Fuzzy neural networks and neuro-fuzzy
networks: A review the main techniques and
applications used in the literature, Applied Soft
Computing, v.92, pp. 106275, 2020.
Stein, S.; Leng, C.; Thornton, S.; Michel, R. A guided
analytics tool for feature selection in steel
manufacturing with an application to blast furnace top
gas efficiency. Computational Materials Science,
v.186, pp. 110053, 2021.
Tan, M.; Bai, Y.; Zhang, H.; Li, G.; Wei, X.; Wang, A.
Fluid typing in tight sandstone from wireline logs using
classification committee machine. Fuel, v.271, pp.
117601, 2020.
Tao, J.; Yu, Z.; Zhang, R.; Gao, F. RBF neural network
modeling approach using PCA based LM–GA
optimization for coke furnace system, Applied Soft
Computing, v.111, pp. 107691, 2021.
Völker, C.; Firdous, R.; Stephan, D.; Kruschwitz, S.
Sequential learning to accelerate discovery of alkali-
activated binders. Journal of Materials Science, v.56,
n.28, pp. 15859–15881, 2021.
Wang, Y. H.; Zhang, H.; Jiang, Z. G.; Zhao, G. Research of
Coke Rate Prediction of Blast Furnace Based on
Operative Characteristics of Auxiliary Resources.
Advanced Materials Research, v.605-607, pp. 1792-
1797, 2012.
Wong, P. K.; Zhong, J.; Yang, Z.; Vong, C. M. Sparse
Bayesian extreme learning committee machine for
engine simultaneous fault diagnosis. Neurocomputing,
v.174, Part A, pp. 331-343, 2016.
Xu, Z. J.; Zheng, Z.; Gao, X. Q. Operation optimization of
the steel manufacturing process: A brief review.
International Journal of Minerals, Metallurgy and
Materials, v. 28, n.8, pp. 1274-1287, 2021.
Yi, Z.; Liu, Q.; Shao, H. Effect of MgO on Highly Basic
Sinters with High Al
2
O
3
. Mining, Metallurgy &
Exploration, v.38, n.5, pp. 2175–2183, 2021.
Zablotskii, P. A.; Petrenko, V. A.; Kovshov, V. N.
Procedure for Numerical Optimization of Blast-
Furnace Charging Parameters Using a Mathematical
Three-Factor Model. Metallurgist, v.61, n.3, pp. 175–
178, 2017.
Zhai, X.; Chen, M.; Lu, W. Fuel Ratio Optimization of
Blast Furnace Based on Data Mining. ISIJ
International, v.60, n.11, pp. 2471-2476, 2020.
Zhang, J.; Li, S.; Li, Z. Investigation the synergistic effects
in quaternary binder containing red mud, blast furnace
slag, steel slag and flue gas desulfurization gypsum
based on artificial neural networks. Journal of Cleaner
Production, v.273, pp. 122972, 2020.
Zhang, L.; Xue, Y.; Xie, Q.; Ren, Z. Analysis and neural
network prediction of combustion stability for
industrial gases. Fuel, v.287, pp.119507.
Zhang, X.; Kano, M.; Matsuzaki, S. A comparative study
of deep and shallow predictive techniques for hot metal
temperature prediction in blast furnace ironmaking.
Computers & Chemical Engineering, v.130,
pp.106575, 2019b.
Zhang, X.; Kano, M.; Matsuzaki, S. Ensemble pattern trees
for predicting hot metal temperature in blast furnace.
Computers & Chemical Engineering, v. 121, pp. 442-
449, 2019a.
Zhao, X.; Fang, Y.; Liu, L.; Xu, M.; Zhang, P. Ameliorated
moth-flame algorithm and its application for modeling
of silicon content in liquid iron of blast furnace based
fast learning network. Applied Soft Computing Journal,
v. 94, 106418, 2020.
Prediction of Sulfur in the Hot Metal based on Data Mining and Artificial Neural Networks
407