Ironmaking & Steelmaking

Processes, Products and Applications

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/yirs20

Computer modelling of burden distribution in the

blast furnace equipped by a bell-less top charging
system

David Fojtik, Jiri Tuma & Petr Faruzel

To cite this article: David Fojtik, Jiri Tuma & Petr Faruzel (2021) Computer modelling of burden
distribution in the blast furnace equipped by a bell-less top charging system, Ironmaking &
Steelmaking, 48:10, 1226-1238, DOI: 10.1080/03019233.2021.1952829

To link to this article: https://doi.org/10.1080/03019233.2021.1952829

© 2021 The Author(s). Published by Informa

UK Limited, trading as Taylor & Francis
Group

Published online: 27 Aug 2021.

Submit your article to this journal

Article views: 1426

View related articles

View Crossmark data

Citing articles: 4 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=yirs20
IRONMAKING & STEELMAKING
2021, VOL. 48, NO. 10, 1226–1238
https://doi.org/10.1080/03019233.2021.1952829

Computer modelling of burden distribution in the blast furnace equipped by a

bell-less top charging system
a a
David Fojtik , Jiri Tuma and Petr Faruzelb
a
Department of Control Systems and Instrumentation, Faculty of Mechanical Engineering, VSB – Technical University of Ostrava, Ostrava, Czech
Republic; bDepartment of Innovation, TŘINECKÉ ŽELEZÁRNY, a. s, Třinec – Staré Město, Czech Republic

ABSTRACT ARTICLE HISTORY

This article describes the simulation method for layering the components of the charge in the blast Received 21 March 2021
furnace, including the particle size distribution and gas flow distribution of bell-less top charging Accepted 2 July 2021
systems in blast furnaces. The Burden Distribution application, which simulates the charge in blast
KEYWORDS
furnaces operated by the company Třinecké železárny (ironworks) to optimize the production of Blast furnace; bell-less top;
pig iron, applied this simulation method. Based on the parabolic trajectory of the material falling burden distribution; stack
from the tilting chute of the bell-less top charging system, the method is calculating the profile of profile; segregation;
individual charge layers. The material forms after impact according to known angles of repose, algorithm; application;
segregating into individual granulometric size fractions. The data incorporated into the simulation computer simulation
enables the estimation of charge and gas flow distribution along the radius of the blast furnace
shaft. The article presents the used mathematical models and equations, including algorithms of
the simulation.

Introduction larger pieces move further from the place of impact, while
smaller pieces stay near the place of impact. An approximate
The profile of the individual blast furnace charge components
differential equation describes granulometric segregation.
determines the reducing gas flow distribution and thus the
Many experimenters explore the behaviour of gas flow
performance and energetic parameters of the operation of
through charges. There are even parameters available describ-
the blast furnace and consequently the production of pig
ing the dependence of pressure loss on the volume flow
iron. Technologists have therefore been striving to attain
square for materials with known granulometric size spectra.
information about gas flow and charge distribution. A bell-
All this data incorporates into the simulation calculation,
less top charging system is a useful tool for controlling
which makes it possible to estimate the charge flow and gas
both these flows. We can measure the effects of changes in
distribution along the radius of the blast furnace stack.
the charging program, which consist of step-by-step chan-
ging the chute inclination during rotation both directly and
indirectly. Indirect information includes temperature distri- Burden distribution simulation model
bution on the perimeter of the charging system and in the The model simulates the burden distribution on the radius of
gas pipelines, whereas profilometers can obtain direct infor- the blast furnace shaft. During charging, the chute of the bell-
mation. Another option is to conduct mathematical simu- less top charging system rotates and gradually changes the
lations, which makes it possible to gain detailed angle of inclination a, so that the individual placements of
information with minimal costs. iron ore or sinter, coke and additives like for example lime-
This article focuses on a method for calculating a simu- stone could be checked along the shaft radius. In simple
lated model of the burden layer profile, including granulo- terms, we can assume that separate charges flow along the
metric size distribution based on measurements of falling trajectory (Figure 1).
individual parameters and a simulation calculation. The The velocity of the burden into the chute and the chute
well-researched parabolic falling trajectory of the burden angle determine the material falling trajectory. Equation (1)
leaving the chute of the bell-less top charging system is the describes the velocity of the burden at the tip of the chute
basis of this model. The angle of the chute and the velocity calculation [1].
of the burden leaving the tip of the chute to define the 















































2 v0 cos (a) + v L sin (a)(sin a + m cos a)
trajectory. 2 2 2 2

After the material falls, it forms into a shape according to v1 = (1)

+ 2gL(cos a − m sin a)
known angles of repose. The granular material segregates
into the individual granulometric size fractions along the where v0 is the vertical velocity of the particles on impact with
shaft radius of the blast furnace based on experimental esti- the chute [m s−1], L is the chute distance travelled [m], v is the
mation. The segregation of granular material means that chute rotation speed [rad s−1], m is the shear friction factor, g

CONTACT David Fojtik [email protected] Department of Control Systems and Instrumentation, Faculty of Mechanical Engineering, VSB – Technical
University of Ostrava, 17. listopadu 2172/15, 708 00 Ostrava, Czech Republic
© 2021 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.
0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any
way.
 IRONMAKING & STEELMAKING 1227

Figure 1. Schematic diagram of the principle of the burden distribution model.

is the acceleration of gravity [m s−2], and a is the angle of falling trajectory [1,6–8]. A polygonal chain gives the
inclination of the chute compared to the axis of symmetry surface and interface between the burden components,
[rad]. which is composed of line segments passing through a
Equations (2) and (3) describe the falling curve [1]. sequence of points Pi (ri , yi ) called its vertices. Line Equation
(5) describes the surface of the charge.
r(t) = L sin a + (v1 sin a)t (2)
pi (r) = ai r + bi , i = 1, 2, . . . , n (5)
gt2
y(t) = y0 − L cos a − − (v1 cos a)t (3) In solving Equation (6), we find the intersection point of
2
the falling curve with the polygonal chain representing the
After expressing time t from Equation (2) and substituting surface of the charge (including the furnace walls) determines
it into Equation (3) we get an Equation (4) of the falling curve. the points of impact I(rI , yI ).

g gL − v12 cos a gL2 g gL − v12 (cos a − ai sin a)

y = − 2 2 r2 + r + y0 − (4) 0=− r 2
+ rI + y0
2v1 sin a v1 sin a
2 2v12 2v12 sin2 a I v12 sin a
gL2
where r represents the distance of the falling trajectory from − − bi (6)
2v12
the central axis, y is the vertical coordinate of the falling tra-
jectory, v0 is the vertical velocity of the particles on impact The maximum angle formed between the surface line of
with the chute [m s−1]. The L is the chute distance travelled the cone formed by the charge and the horizontal plane rep-
[m], v is the chute rotation speed [rad s−1], m is the shear fric- resents the angle of repose. Generally, the repose angle
tion factor, g is the acceleration of gravity [m s−2], a is the depends on the properties of the material, like its density,
chute inclination angle compared to the axis of symmetry size and grain shape, and the coefficient of friction of the
[rad], y0 is the vertical coordinate of the beginning of the material [9]. When the burden settles, its angle of repose
chute. varies concerning the axis of symmetry win (inner angle)
Randomizing the velocity of the burden leaving the chute from the angle in the direction of the wall wout (outer
tip models the scatter of the charge on the impacting stock- angle). The repose angle in the direction of the wall is signifi-
line surface. The volume of the charge falling into one cantly smaller. Equations (7) and (8) express the nominal
annulus (a region bounded by two concentric circles) repose angles [8].
divides into virtual sub-charges with assigned various vel-
ocities for the scatter parameter. These sub-charges then d0.05
∅in = ∅max C (7)
move along various falling curves. This method brings the 4FS0.05
model closer to realistic conditions.
R − rI1.4
The substitution by two or more lines, a cubic curve, or the ∅out = ∅in (8)
Gaussian curve builds the burden profile mathematical model 5.8
[2–5]. It is most often represented by two lines, which are where ∅max is the maximum angle of repose [deg], C is a
beneath the material angles of repose and intersect on the constant, d is the diameter of the particle [m], FS is the
1228 D. FOJTIK ET AL.

shape dimensionless factor [–], R is the radius of the At the same time points L(rL , yL ), K(rK , yK ), Pj (rj , yj ) and the
throat [m], rI is the distance [m] from the symmetry axis points Pi (ri , yi ), i = 1, 2, . . . , n located below the triangle L,
to the intersection point of the trajectory and the burden K, Pj form the vertices of a polygonal chain m(r).
surface. ⎧
⎪
⎪ ai r + bi , r . min(rL , rj ) < r , max(rL , rj ),
The falling trajectory, i.e. the chute angle, also affects the ⎨
i = 1, 2, . . . , n
repose angle. In practice, angles of repose are determined m(r) = (16)
⎪
⎪ k(r), r ≥ min(rL , rj ) < r , rI
experimentally with the help of scaled-down blast furnace ⎩
l(r), r ≤ max(rL , rj ) < r ≥ rI
models [8]. Then it is possible to connect individual repose
angles with each position of the chute. Another option is to The polygonal chain m(r) with vertices Mi (ri , yi ) represents
correct repose angles according to the deviation of the the cross-section of the rotating body of the dumped charge.
tangent angle tI (r) to the falling trajectory at the point of The area of a polynomial multiplied with the path length of its
impact. Equation (9) describes the tangent to the trajectory centre of gravity (Guldin’s rule) calculates the volume of a
at the point of intersection. body. The algorithm calculates the additional individual tri-
angles, which the polygonal chain is based on. The centre
tI (r) = mI r + nI
  of gravity of the triangles is in a third of the median.
rI − l sin a
=− g + cotn(a ) r n 
y1 + yi−1 yi−1 + yi
(vout sin a)2 V = 2p (r1 − ri−1 ) + (ri−1 − ri )
  2 2
rI − l sin a i=3
(17)
+ g − cotn( a ) rI + yI (9) 
(vout sin a)2 yi + y1 r1 + ri−1 + ri
+ (ri − r1 )
rI − L sin a 2 3
mI = −g − cotn(a) (10)
(v1 sin a)2 For the list of points indexed from zero, we can write the
algorithm of the calculation of the volume in the language
The slope of the mI tangent to the falling trajectory at the
C# in the following manner.
point of impact tI (r) is used to determine the difference
between the angle of incidence and the vertical axis, which
is multiplied by the correction constants cin , cout for a given static double V(List<POINT> M)
{
type of angle. The resulting values are then subtracted from double S, DV = 0;
the nominal repose angles ∅in , ∅out . for (int i=2; i < M.Count; i++)
p {
win = ∅in + + tan−1 mI cin (11) S = (M[0].r - M[i-1].r)*(M[0].y + M[i-1].y);
2 S += (M[i-1].r - M[i].r)*(M[i-1].y + M[i].y);
p S += (M[i].r - M[0].r)*(M[i].y + M[0].y);
wout = + tan−1 mI cout − ∅out (12) DV += S*(M[0].r + M[i-1].r + M[i].r);
2 }
return Math.PI*DV/3;
}
Charge layering algorithm
It is necessary to compare the calculated volume V with the
First, we must find the nearest point of the polygon chain of
to-be-dumped volume Uk along the given trajectory. In the
the charged surface (at the same or higher height) Pj (rj , yj )
first run, the remainder VR is set as the required volume Uk .
to the intersection point I(rI ; yI ). Depending on whether
The algorithm performs iterations with the updated surface
the found point is in the direction of the axis of symmetry
if the ratio of the rest of the required volume is greater
or towards the charging system wall, a line k(r) is run
than the permissible error eV (see Equation (18)), while the
through the point under the inner win or the outer repose
calculated point K(rK , yK ) becomes the new point of impact
angle wout . Subsequently, the algorithm inserts a line
I(rI , yI ) and the charge volume reduces by the amount of
through the intersection point of this line K(rK , yK ) with
the calculated volume VR := VR − V.
the tangent to the trajectory tI (r) under the second repose
angle. VR − V
≤ eV (18)
Uk
tan (win )r + yj − tan (win )rj , rj ≤ rI
k(r) = (13)
tan (win )r + yK − tan (win )rK , rj . rI If the calculated volume V in the last cycle is greater than
the charge volume remainder VR , the algorithm keeps the last
tan (wout )r + yK − tan (wout )rK , rj ≤ rI
l(r) = (14) surface, and measure the straight distance lIK between the
tan (wout )r + yj − tan (wout )rj , rj . rI intersection point I(rI , yI ) and the calculated point K(rK , yK ).
The intersection point of this second line with the polygon The distance lIK has also a negative sign assigned. Then the
chain given by the vertices pi (r) determines the second per- numerical cycle for searching for K(rK , yK ) starts, while it is
ipheral point of the surface of the added increment L(rL , yL ). valid that the calculated volume is equal to the remainder
The points L(rL , yL ), K(rK , yK ), Pj (rj , yj ) together with within the given tolerance. The algorithm within the cycle
points Pi (ri , yi ), i = 1, 2, . . . , n that are not below the tri- first halves the distance lIK , and then the point K(rK , yK )
angle L, K, Pj form a new surface resulting in the polygon moves along the tangent by a distance of lIK below or
chain containing the vertices p(r). above the original position, depending on its sign. This
⎧ point is then interlaced with the lines k(r) and l(r).
⎪
⎪ ai r + bi , r, min(rL , rj ) > r . max(rL , rj ),
⎨ k(r) = tan (win )r + yK − tan (win )rK (19)
i = 1, 2, . . . , n
p(r) = (15)
⎪
⎪ k(r), r ≥ min(rL , rj ) < r , rI l(r) = tan (wout )r + yK − tan (wout )rK
⎩ (20)
l(r), r ≤ max(rL , rj ) < r ≥ rI
 IRONMAKING & STEELMAKING 1229

Then the intersection points L(rL , yL ), P(rP , yP ) of l(r) and furnace by Equation (25).
k(r) are defined by the polygonal chain vertices pi (r),
UC
where pi is a vertical coordinate and r is the distance DyT = (25)
pR2
from the blast furnace axis. By adding points
Pi (ri , yi ), i = 1, 2, . . . , n located below the triangle L, K, P, a where UC is the volume of the component layer, and R is the
polygonal chain is formed by the points m(r), and the radius of the cylindrical part of the blast furnace. The radii of
volume V of the rotating body is calculated. The volume the individual vertices of the polygonal chain remain the
is compared with the remaining to be added UR volume, same. That is, we can calculate coordinates y′i and r′i of all
where the sign of the difference between the remaining points of the vertices of the polygon chain after descending
volume and the calculated volume determines the current in the cylindrical parts of the blast furnace by Equations
sign of the calculated distance lIK . The cycle repeats until (26) and (27).
the ratio of the remainder to the required volume is UC
greater than the allowable tolerance. y′i = yi − (26)
pR2

|VR − V| r′i = ri (27)

≤ eV (21)
Uk If there is a layer over the entire diameter of the furnace,
then we can calculate the descent Dys in the conical parts
After a suitable peak K(rK , yK ) is found, the polygon chain
(Stack, Bosh) of the blast furnace by Equation (28).
vertices of the surface updates to p(r).
The described algorithm shown in Figure 2 repeats for all R
Dys = y − + HT
determined charge trajectories. When all charges process all tan b



































 3

set positions of the chute, the resulting polygon chain ver-
3 R 3UC
tices are stored as a representation of the surface of the + − HT − y + (28)
tan b p tan2 b
dump layer.
After every completed layer, the charge decreases by the where y is the coordinate of the point of the original ver-
height of the last layer at the location of measurement with tices of the polygon chain on the furnace wall before the
a depth gauge. This ensures that the height of the material descent, HT is the height of the throat, R is the radius of
is always the same at the location of measurement (see the throat, b is the wall inclination angle, UC is the
Figure 3). Another option is to calculate the reduction volume of the layer. Therefore, we can calculate the coordi-
based on the volume of the dumped layer, for example, nate y′i of the points of the vertices of the polygon chain
using Equation (25) [10]. after descending in the conical parts of the blast furnace
The entire algorithm repeats for all groups and materials in by Equation (29). In addition, we proportionally adjust the
the group (see Figure 4). Repeating calculations of the profile radii of the individual vertices of the polygon chain by
with the corresponding descent of the burden compensates Equation (30) [10].
for the fact of unknown initial burden profile. 



































 3
The calculation of one completed charging program iter- R R 3UC
y′i =
3

ates to the point when the difference of the initial and final − HT − − HT − yi + (29)
tan b tan b p tan2 b
profile is smaller than the permissible error eSL (see Figure 5
and Equation (22)). The error value eSL is usually set to 0.01 R
− HT − y ′ i
m. The charging program represents the sequence of chute ′ tan b
ri = ri (30)
inclinations for individual charge components, including R
− HT − y i
their repetition. tan b
  We can also solve the descent under the described con-
n  
1  1 n
 ditions numerically. First, we approximately calculate the
 li − lj  ≤ eSL (22)
n i=1  n j=1  descent with Equations (25) and (26), and then we adjust
the descent value yi until the new volume corresponds to
the original volume. In the more advanced modelling
Burden descent in stack methods, we calculate the descent with the burden descent
speed variable [10].
The burden descent in the stack of the blast furnace depends
on the angle of the walls and the dimensions of the throat
(see Figure 6). The basic model calculates the same vertical
Modelling the granulometric size distribution of
descent velocity, where every vertex of the polygonal chain
the burden after impact
moves by the same distance Dy, provided that the volume
of the layer UC before and after the descent U′C is the same Many papers deal with the topic of size segregation of par-
[10]. ticle mixture in conical pile formation [6,11,12]. Exper-
iments have shown that an approximation model can be
Dyi = yi − y′i = Dyi−1 , i = 2, . . . n (23) used to describe segregation, according to which the
change in the concentration of the granulometric size frac-
UC = U′C (24) tion c relative to the distance of the point of impact x is pro-
portional to its concentration and the remainder of the
We can calculate the descent DyT of the burden in the concentration to 1 (meaning 100%) [6]. We designate the
cylindrical parts (Throat, Stack, and Barrel) of the blast proportionality factor by coefficient a in the following
1230 D. FOJTIK ET AL.

Figure 2. Flow chart of the charging algorithm of one chute position.

differential Equation (31). where we calculate the parameter b according to Equation (33).
1
dc b= −1 (33)
= −a c (1 − c) (31) c(0)
dx
The value of the coefficient a was experimentally found,
In solving this equation, we find the relation c(x) and according to the literature, is equal to 0.31 [6].
between the change of the concentration of the granulo-
metric size fraction and the distance of the point of
impact (see Equation (32)).
Model of the gas flow distribution
1 We can assume the same pressure in the space above the
c(x) = (32)
1 + b ea x burden stockline as below the reduction zone above the
 IRONMAKING & STEELMAKING 1231

Figure 3. Burden descend during two interactions of the charging components algorithm.

tuyeres. Let the pressure drop Dp in the charge column be The additional resistance again gas flow at the contacting sur-
constant. Furthermore, let the resistance to the gas flow be faces of layers of different materials is ignored.
dependent on the equivalent grain de and the free space 1
between the grains of the material. Assuming a turbulent
flow of reducing gas through the charge, we can describe Burden distribution application
the relation of the mentioned quantities on the distance r
from the shaft axis in Equation (34). We created the program called Burden Distribution on the
Microsoft platform .NET framework. The computational core
Dp Dh(r)[1 − 1(r)]
k(r) = = (34) of the program is written in the programming language
v2 (r) 13 (r) de (r) C++ and the graphic user interface is created in Visual Basic
The next equation (Equation (35)) gives the gas flow rate. .NET. The environment uses the MDI (Multiple Document





Interface), which makes it possible to create any number of
Dr simulations and compare them with each other. You can
v(r) = (35)
k(r) see the appearance of the main program window and some
forms in Figure 7.
The relative values of gas velocity at individual points
We can enter input data and parameters in two groups:
along the charging system radius are sufficient to analyse
the gas flow distribution along its radius. In few works, the
(1) Global settings,
dependence of the size of free spaces on the granulometric
(2) Simulation parameters.
composition of the charge was experimentally determined
[13,14]. We determine the equivalent grain de (see Equation
The dimensions and physical properties of the bell-less
(36)) of the wide range of average grain sizes di of individual
top charging system, charge properties, and other par-
fractions of the charge of concentration ci using a weighted
ameters of the simulation are set up in the global settings.
harmonic mean calculation formula, using weights based
We set these values only once at the beginning for a
on the concentrations of that fraction.
specific blast furnace or we edit them very rarely. The
1 data in this program is set for all simulations in the Global
de =  c (36)
i Default Settings dialogue (see Figure 8). Here, the general
i
di parameters affecting the simulation calculation and its
The ratio of the length of streamline through the coke and accuracy are also set. The program stores these settings
iron ore, i.e. in sinter Dh(r), also influences gas flow resistance. in an XML configuration file, so it is possible to back up
1232 D. FOJTIK ET AL.

Figure 4. Flow chart of the charging algorithm.

the configuration or create variations of the simulation editing simulation parameters. The input data for each simu-
calculation. lation can be stored in a separate file from which they can be
The second group of data is set individually for each simu- loaded. Therefore, the user can easily create different char-
lation. There is no limit to the number of simultaneous simu- ging options just by modifying data and saving the file
lations; the user only selects an item of New model simulation under a new name. All the values from the global settings
in the menu or clicks the icon and immediately a new window are also stored in the file, so it is not dependent on the exist-
containing a simulation appears. This window has two parts ing settings.
(see Figure 9). In the left part, there is a panel with tabs, in Upon pressing the button called Recalculate and plot, the
which there are individual editing fields for inserting and result of the simulation presented in graphical or numerical
 IRONMAKING & STEELMAKING 1233

. RGD: Relative gas distribution

. RDF: Relative distribution of iron
. RFV: Relative flow velocity
. EPS: Empty spaces
. DHC: Total height of charge in each sector
. DEK: The mean particle size in a given sector over the
entire column height
. VFE: Relative distribution of iron flow velocity
. FWR: Flow resistance
. OOC: Ore/(Ore + Coke) ratio.

Of course, it is also possible to represent data graphically

in various ways (colour of lines, font, etc.), which can be set
in the Properties of the context menu. It is also possible to
base calculations in the application on a template – these
are XML files of the application with the prxt-extension
stored in the application folder. This means that it is possible
to add blast furnaces of different parameters or calculation
variants by merely saving the pre-filled file. We can create a
new entirely independent template by saving the performed
simulation with the prxt-extension in the application folder.
After this step, the template automatically appears in the
computation menu (see Figure 10).
One unique feature is the ability to export all calculated
simulation data into an MS Excel file. We must install MS
Figure 5. Principle of calculating the difference between the initial and final Excel for this feature to work. The function is available from
stockline. the menu offering Computation, or Export calculation to MS
Excel menu. The menu activates after we have calculated
the simulation. After selecting export, all computed data are
form appears on the right side. After performing the calcu- exported into a new sheet and list with a name identical to
lation, we display the cross-section of the blast furnace the name of the simulation (it is possible to define the
throat with the profile of the individual charge layers on name when saving the simulation). The image of the simu-
the right side of the window (see Figure 9). Above these lation appears below the table data. Furthermore, the appli-
layers, we can plot graphs with other calculated data as cation automatically creates and implicitly places into a
desired. The context menu with a menu of these graphs separate sheet a graph of the charge layer based on the cal-
appears upon clicking the right mouse button above the culated coordinates of the breaks in the layer profile. The
graphical representation of the data. The user can then output in MS Excel shows Figure 11.
select any combination of graphs.

Results and discussion

We can control the blast furnace in two standard ways, from
the top and the bottom. Understanding and effectively mana-
ging the blast furnace operation is essential in reducing the
cost per tonne of pig iron produced. Furthermore, it can be
used as a preventive measure to avoid abnormal operating
conditions, which are frequently caused by an incorrectly
chosen distribution matrix. Historically, this problem has
also occurred in the blast furnaces used at Třinecké železárny.
In November 2013, Třinecké železárny finished the invest-
ment project and started PCI (Pulverized Coal Injection) tech-
nology [15], which replaced approximately 1/3 of the
previously used metallurgical coke in the charge with pulver-
ized coal. This pulverized coal is injected directly into the blast
furnace through the tuyeres. The implementation of PCI tech-
nology required a new distribution matrix. The correct choice
of the loading matrix has a significant impact on the
efficiency of the technology.
The managers and technologists of the Třinecké železárny
blast furnace department were also aware of this. Already in
2013, the company ordered the development of a simulation
Figure 6. The illustration of the layer descends in the throat and the stack part system for modelling the blast furnace charge from the Tech-
of the blast furnace. nical University of Ostrava. In 2014, technologists began to
1234 D. FOJTIK ET AL.

Figure 7. Simulation software called Burden Distribution.

Figure 8. Global default settings dialogue.

Figure 9. Graphic representation of burden layer profile with context menu for graphs choice.
 IRONMAKING & STEELMAKING 1235

Figure 10. Menu of new model simulation by select the template.

Figure 11. Export MS Excel document.

use simulation software extensively to find the optimal Using the Burden Distribution Application, the loading
method of loading the charge. matrices were gradually optimized, and in addition to redu-
Before the implementation of Pulverized Coal Injection, cing the amount of coke by about 30%, the permeability of
100% of the fuel was blast furnace coke, which not only the central part of the charge column was increased by
serves as fuel in the blast furnace, but also ensures the per- increasing the proportion of metallurgical coke (fraction
meability of the blast column, acts as a reducing agent, and above 40 mm).
a carburizing agent. Figure 12 shows a comparison of the results of the loading
With the partial replacement of the coke charge by finely matrices simulations before PCI injection (left) and after PCI
ground coal injected through tuyeres, there are, among other optimization. The simulation results show that the resistance
things, fundamental changes in the gas-dynamic conditions to flow in the central part of the charge is significantly
in the blast furnace. Because of these changes, it is necessary reduced. A detailed comparison is illustrated by the graph in
to adapt the loading matrices to ensure the permeability of Figure 13, which shows a reduction of the average resistance
the charge column and to avoid uneven running of the to flow by 15% and even 25% in the centre of the furnace.
blast furnace or even the hanging of the charge and stopping In some publications, simulation methods for layering the
the descent. components of the charge in the blast furnace authors
1236 D. FOJTIK ET AL.

Figure 12. Result of burden distribution simulation before implementation PCI (left) and after PCI optimization (right).

Figure 13. Comparison of significant burden distribution parameters before PCI implementation and after PCI optimization.

verified through scaled-down models of blast furnaces [7,8]. composition of the blast furnace gas using the parameter
However, verifying and evaluating the benefits of the hCO (%) – carbon monoxide efficiency [16,17].
model under operational conditions is not easy. First, it is
CO2
not possible to physically verify the actual state of the hCO = · 100% (37)
CO + CO2
charge distribution. Second, it is not possible to guarantee
that the parameters are invariant before and after the Thanks to the gradual optimization using the Burden Dis-
model is implemented. tribution Application, regular operation of the blast furnaces
In practice, the essential parameter indicating efficient in the Třinecké Železárny Ironwork was achieved, and the
production in a blast furnace is the specific fuel consumption achievement of a gas utilization rate of hCO above 50% (for
(sum of metallurgical coke, minor coke fractions, and substi- details, see Table 1).
tute fuels) per tonne of pig iron produced. As the graph in Figure 14 illustrates, since the end of 2013,
Significant is the efficiency of utilization of the reduction when PCI injection into the blast furnaces started, the optim-
potential of the gas for indirect reduction of metal-bearing ization of the blast furnace loading has also succeeded in
charge, which can be expressed by evaluating the increasing the utilization of the reduction potential of the
 IRONMAKING & STEELMAKING 1237

Table 1. Evolution of Blast Furnace Gas (GBF) composition and gas utilization eta CO.
Year 2012 2013 2014 2015 2016 2017 2018 2019 2020
CO in BFG [%] 22.3 22.5 22.7 22.4 22.5 22.3 21.8 22.2 23.4
CO2 in BFG [%] 20.7 20.8 21.4 22.4 22.8 23 23.4 23.4 23
H2 in BFG [%] 1.7 1.7 3 3.1 3.4 3.5 3.5 3.5 3.5
ηCO [%] 47.8 48 48.5 49.9 50.2 50.7 51.7 51.1 49.6

Figure 14. Evolution of blast furnace gas composition and gas utilization ηCO.

blast furnace gas ηCO from 48.5% to 51.7% with stable metal- the data of the required input parameters of actively operated
bearing charge parameters. Values exceeding 50% indicate blast furnaces from the cooperation of the Research Center of
suitable control of the blast furnace process and serve to Advanced Mechatronic Systems with a pig iron producer in
verify the blast furnace loading model. The decrease in the Czech Republic, the company Třinecké železárny.
2020 is mainly due to production curtailment related to the The result is an application, which helps to find the
COVID 19 pandemic. optimal charging method and can quickly respond to conse-
quences of changes of individual material properties or char-
ging program parameters. The graphical representation of
Conclusions the simulation results gives a clear overview of the burden
The blast furnace charging simulation provides data, which distribution, as well as the data about many of other observed
are otherwise obtainable in very complicated and costly quantities along the radius of the blast furnace stack. The
ways. The computational core of the simulation can utilize application is user-friendly, although the entry of input
analytical calculation methods or other advanced finite data, given their nature, requires a highly qualified user
element methods [18]. The main advantage of analytical who can define them correctly.
methods is the significantly lower processing power require- Another advantage is the ability to model various changes
ment and thus the speed of the performed simulation. Such a and defects of the bell-less top charging system quickly. If
solution is then considerably more accessible to blast furnace there is a defect in the positioning system of the chute, it is
operators who can easily explore more variants of charging possible to find out what the consequences will be or it is
methods immediately. possible to adjust the charging program for minimal negative
We have chosen the analytical methods for calculating the effects. We have created the software on an order from the
layering of the material, including segregation and gas flow dis- company Třinecké železáren for simulating the charging in
tribution based on the research of used charging simulation blast furnaces.
methods. We have programmed and implemented the pro-
cedures in the new charging program named Burden Distri-
bution. We have described the calculations, including the Disclosure statement
proposed algorithms, in detail in the article. We have provided No potential conflict of interest was reported by the author(s).
1238 D. FOJTIK ET AL.

Funding [7] Mitra T, Saxén H. Model for fast evaluation of charging programs in
the blast furnace. Metall Mater Trans B. 2014;45:2382–2394.
This work was supported by the European Regional Development Fund [8] Park J-I, Jung H-J, Jo M-K, et al. Mathematical modeling of the
in the Research Centre of Advanced Mechatronic Systems project, burden distribution in the blast furnace shaft. Met Mater Int.
CZ.02.1.01/0.0/0.0/16_019/0000867 within the Operational Programme 2011;17:485–496.
Research, Development and Education and the project SP2021/27 [9] Mehta A, Barker GC. The dynamics of sand. Rep Prog Phys.
Advanced methods and technologies in the field of machine and 1994;57:383–416.
process control supported by the Ministry of Education, Youth and [10] Fu D, Chen Y, Zhou CQ. Mathematical modeling of blast furnace
Sports, Czech Republic. burden distribution with non-uniform descending speed. Appl
Math Model. 2015;39:7554–7567.
[11] Rahman M, Shinohara K, Zhu HP, et al. Size segregation mechanism
ORCID of binary particle mixture in forming a conical pile. Chem Eng Sci.
2011;66:6089–6098.
David Fojtik http://orcid.org/0000-0003-4563-8131 [12] Zhang J, Qiu J, Guo H, et al. Simulation of particle flow in a bell-less
Jiri Tuma http://orcid.org/0000-0001-7053-9835 type charging system of a blast furnace using the discrete element
method. Particuology. 2014;16:167–177.
[13] Yamada T, Sato M, Miyazaki N, et al. Distribution of burden materials
References and gas permeability in a large volume blast furnace. Kawasaki
[1] Radhakrishnan VR, Ram KM. Mathematical model for predictive Steel Giho. 1974;6:16–37.
control of the bell-less top charging system of a blast furnace. J [14] Iwamoto N, Makino Y. State of sulphur in a synthetic blast furnace
Process Control. 2001;11:565–586. slag and relation between segregation of sulphur and morphology
[2] Saxén H, Hinnelä J. Model for fast evaluation of charging programs of primary phase. Tetsu-to-Hagane. 1983;69:220–227.
in the blast furnace. Miner Process Extr Metall Rev. 2004;25:1–27. [15] Ishii K, Ariyama T, Japan HU. Advanced pulverized coal injection
[3] Zhu Q, Lü C-L, Yin Y-X, et al. Burden distribution calculation of bell- technology and blast furnace operation. Oxford: Pergamon; 2000.
less top of blast furnace based on multi-radar data. J Iron Steel Res [16] Azadi P, Minaabad SA, Bartusch H, et al. Nonlinear prediction model
Int. 2013;20:33–37. of blast furnace operation status. In: Sauro Pierucci, editor.
[4] Nag S, Gupta A, Paul S, et al. Prediction of heap shape in blast Computer aided chemical engineering. Vol. 48. Milano: Elsevier;
furnace burden distribution. ISIJ Int. 2014;54:1517–1520. 2020. p. 217–222.
[5] Matsuzaki S. Estimation of stack profile of burden at peripheral zone [17] Geerdes M, Chaigneau R, Lingiardi O. Modern blast furnace iron-
of blast furnace top. ISIJ Int. 2003;43:620–629. making: an introduction. 4th ed. Amsterdam: IOS Press; 2020.
[6] Kondoh M, Konishi Y, Okabe K, et al. Interprétation cinématique d [18] Zhou C, Tang G, Wang J, et al. Comprehensive numerical
essais ́
de chargement à l échelle 1 dans un haut fourneau à gueu- modeling of the blast furnace ironmaking process. JOM.
́ 2016;68:1353–1362.
lard sans cloche. Rev Met Paris. 1981;78:733–744.