Optimization of Lime Injection in EAF Systems by Using Pneumatic Sidewall Injectors

Master of Science Thesis

Hourakhsh Bayat

Åbo Akademi University

Faculty of Science and Engineering

Supervisors:

Professor Henrik Saxen

Professor Ron Zevenhoven

Turku, Finland, 2022

Abstract

This master’s thesis work is a project for Nordkalk corporation as a part of a Business Finland
co-funded project on Fossil-Free Steelmaking (FFS). The experiments of this study were
performed in the Laboratory of Process and Systems Engineering at Åbo Akademi University.

The metal and steel industry play a key role in Finland’s pursuit of carbon neutrality by 2035.
The transition to a low-carbon society offers Finland significant business potential but requires
investment in research. In the FFS project, industrial and research partners are studying
different solutions and options for producing fossil-free steel. The project investigates
productive solutions for using green energy, such as hydrogen, biochar and biogas, in the steel
industry. In addition, the project studies the production of fossil-free lime and new solutions for
utilizing the by-products of steelmaking processes [1].

In this master’s thesis work, cold model experiments on lime particle injection systems were
performed to find an optimum particle size and operational conditions for the lime injection
systems in an Electric Arc Furnace (EAF). The aims of the study include gaining key knowledge
on the injection process, minimizing the dust formation through optimal particle size,
investigating the dust formation patterns in comparison to particle size, understanding the
multiple phase behaviors, including solid, liquid, and gas interactions, in the injection process.

The experiments performed in the study were divided into two different parts. In a first set of
experiments, the behavior of a 3-phase system containing solid, gas, and liquids was
investigated. The use of paraffin oil on top of water to simulate the slag and molten iron,
respectively, in the EAF was a very useful tool for studying the effect of the slag on the possible
interactions between particles and slag in the systems. In this set of experiments, by using a
cold model observed by a high-speed camera, calcium oxide (CaO) particles with different size
ranges were injected into the liquids. Operational conditions, such as the air flow rate and the
nozzle size, were changed to study their influence on the injection process. The particle velocity
and the penetration depth of the particles and jet were measured. In addition, gas-particle
behavior, particle rebounding, dust formation patterns, and liquid phase behavior were observed
and studied.

The second part of the experiments was performed by the same rig. In this series of experiments,
the main objective was to measure the amount of dust produced during the injection process in
order to find the particle size that resulted in a minimum amount of dust. Moreover, the
influence of operational conditions on the dust formation, such as the air flow rate and the
nozzle size, were investigated.

Due to the importance of the effect of the conveying on the solid particles behavior and
flowability of the solids, the methods for conveying solids as well as the design of a conveying
system were also reviewed and discussed, and related equations were presented.

In conclusion, based on the results of the experiments the penetration depth was found to
increase with the decrease in particle size and nozzle size, and increase with the air flow rate.
The distribution of particles through the liquid phase in which all particles distribute in different
depths and widths of the liquid occurs with particle sizes in the range of 710-1000 µm and the
air flow rate equal to 30 l/min. Moreover, by increasing the particle size and air flow rate the
dust formation and particle rebounding decreased. These findings can be used for designing
proper operational conditions in the true process by using relations derived based on
dimensionless numbers.
Table of Content

1 Introduction 1
2 Background 5
2.1 Lime quality for the steelmaking processes 5
2.2 Lime injection in EAF 7
3 Pneumatic transport and conveying of solids 10
3.1 Design of pneumatic conveying system 10
3.2 Saltation velocity 11
3.3 Pickup velocity 12
3.4 Solid velocity using slip velocity 12
3.5 Conveying gas and solid particle frictional expressions 13
3.6 Terms for calculating the pressure drop for horizontal conveying 14
3.7 Drag coefficient 14
4 Physical modeling 16
4.1 Principles of physical modeling 16
4.1.1 Geometric similarity 16
4.1.2 Mechanical similarity 17
4.2 Choice of working fluid 20
4.2.1 Molten phase 20
4.2.2 Slag 20
4.3 EAF and prototype dimensions 22
4.4 Real conveying system, injection nozzle, and lime particle 23
characteristics
4.5 Characteristics of a real dust collecting system in the EAF 25
5 Apparatus and experimental methods 26
6 Results and discussion 31
6.1 Effect of operational factors on penetration depth 31
6.1.1 Effect of nozzle size on penetration depth 31
6.1.2 Effect of particle size 34
6.1.3 Effect of air flow rate 36
6.2 Measurement of particle velocity 36
6.3 Effect of operational factors on dust produced by the injection process 37
6.4 Visual observations 41
6.4.1 Particle rebounding and dust formation pattern 41
6.4.2 Cavity formation and penetration depth 45
6.4.3 Gas/particle behavior while penetrating fluids 45
6.4.3.1 Comparison between gas/particle behavior in models with and without 49
oil layer
6.4.4 Comparison between cavity shape formed by particle injection and air 52
blowing
6.4.5 Particle distribution in molten phase 54
6.4.6 Behavior of liquids 57
6.4.6.1 Mixing behavior 57
6.4.6.2 Splashing behavior and droplet generation 60
7 Conclusion 62
References 64
Appendix A Penetration depth as a function of the rate of change of momentum 66
Appendix B Dimensional analysis 68
Appendix C Physical properties of slag 76
Appendix D Link to the YouTube channel and attachments 81
 1 Introduction

In steelmaking, lime (CaO) serves as a flux for removing impurities (silicon, phosphorus, and
sulfur) in refining steel. Furthermore, lime plays a vital role at different stages of steelmaking,
particularly in making a slag which removes impurities and provides a safer platform to
withstand high intensity arc plasma in Electric Arc Furnaces (EAF) and violent reactions in
Basic Oxygen Furnaces (BOF). Steel industry is today facing several challenges regarding the
growing demand of producing cleaner steel, while they are being pushed to reduce their
environmental impacts by managing by-products and finding methods to mitigate carbon
dioxide emissions and optimize energy usage. Even if the cost of lime has a relatively small
impact on the cost of steel, lime quality and methods of lime injection can have a vital impact
on steel quality, metallurgical properties, productivity, total cost of production, and
sustainability issues [4].

Steelmaking, from converter to cater, is categorized into three distinct stages known as primary
steelmaking, secondary steel making, and casting. Steelmaking process routes involve two
dominating technologies: oxygen steelmaking (BOS) and electric steelmaking (EAF
steelmaking). Figure 1 illustrates the main steelmaking process routes [2].

Figure 1. The two dominant steelmaking and processing of steel [2].

1
In the BOS process, molten iron (hot metal) from a blast furnace is refined under an oxidizing
and basic environment. Refining is carried out in a pear-shaped vessel, traditionally termed as
a “converter” (i.e., a BOF). The vessel is lined with basic refractories made from magnesite,

dolomite, etc. that provide a relatively inert ambient to the otherwise corrosive, basic slag,
prepared by dissolving lime. Oxygen is injected at supersonic speed into the molten iron
through a water-cooled, multi-hole lance with de Laval nozzles. The oxygen readily dissolves
in liquid steel and starts oxidizing and eliminating impurities dissolved in the hot metal [2]. A
major part of the reaction occurs in the foaming slag layer, where small iron droplets are in
close contact with the slag. Figure 2 shows a schematic of a top blown converter.

Figure 2. A top blown converter. (a) Characteristics of the vessel and (b) slag-melt-gas
interaction and molten steel flow during blow [2].

The electric arc furnace (EAF) is a central element and the highest energy consumer in the
recycled steel processing industry. The EAF consists of a shell (walls with water cooled panels
and lower vessel), a hearth (refractory material that covers lower vessel), and a roof with the
electrodes [3]. This is a solid charge-based process and uses steel scrap and direct reduced iron
(as opposed to molten iron in the basic oxygen process) as the chief iron-bearing material. The
extent of refining required in an EAF is generally less than that in a BOF, since less amount of
impurity has to be eliminated. Energy required for melting the solid charge, dissolution, and
subsequent refining of the bath is provided by electrical energy. Graphite electrodes connected
to ultrahigh power transformers are used to strike an arc that supplies heat to the solid charge
and subsequently to the bath. Iron ore (as oxidizing agent) and lime (as flux) are both added to
the EAF charge material to eliminate the impurities present in DRI and other charge materials

2
through oxidation and fixing the impurity oxides with a suitable fluxing agent [2]. Figure 3
shows a schematic of an EAF.

Figure 3. Schematic of an Electric Arc Furnace [3].

Research on the injection of lime particles into liquid by using a cold model, in which both
molten phase and slag exist, is scarce. However, a few studies exist regarding physical modeling
of penetration depth of other types of particles into only water as molten phase.

Takahiro et al. [5] worked on the effect of particle penetration depth on solid/liquid mass
transfer rate by particle blowing technique. They obtained a non-dimensional equation of
particle penetration depth based on the experimental results, by injecting pearlite and
polystyrene, by using a top blowing method.

Kimura [6] studied the penetration depth of different solid particles in a water model by using
a top blowing lance to mathematically model a jet of gas-solid mixture penetration into liquid
and compare the results with a hot model.

3
Farias et al. [7] studied the penetration depths of submerged jets in water and described their
behavior by using equations balancing the momentum of the jet and buoyancy.

In all previous studies, the penetration depth and behavior of the particle penetration in liquid
were investigated by using a top blowing and submerged physical model, and the focus of
previous works was the modeling of BOF. Due to the growing interest in pneumatic injection
of lime as a viable technology with benefits for the steelmakers[8], [9], the present work studied
injecting lime particles by using a side wall pneumatic flux conveying system to observe the
behavior of a three-phase model including solid, gas, and liquid. The slag layer in the physical
model was also simulated by using paraffin oil in order to gain a better insight into the effect of
slag on the penetration depth.

Regarding the importance of studying the penetration depth, it can be mentioned that the main
disadvantages of powder or particle injection is that the residence time of the particle is quite
short. Therefore, the reagent efficiencies are often poor and, consequently, the trajectories of
the particles are of fundamental importance to these processes. For a reagent such as lime, the
particle must be in contact with the metal long enough for a reaction to occur. As a result, it is
important that particles penetrate enough and distribute widely in the molten steel phase to have
enough time for reaction [7].

The main topics of the current study are to study the effect of particle size on dust formation
and interaction with liquid phases. Moreover, the aims are to gain a key knowledge on the
injection process, minimizing dust formation by optimizing particle size, and understanding the
behavior of the liquids (slag and steel) during the injection process. This study mainly focused
on the physical modeling of the injection process in an EAF by using similarity criteria.

Two different sets of experiments were performed to pursue the aims of the study. The first set
of experiments investigated the behavior of the three phases (gas-solid-liquid) and the effect of
particle size as well as other operational conditions, such as the air flow rate, and the nozzle
size, on the penetration depth. The second set of experiments studied the influence of the
particle size and the operational conditions on the amount of dust produced during injection
processes. By dust is here meant the part of the particles that “bounce” off the surface of the
molten phases and is diverted upwards, with a potential to be trapped by the outflowing gas.
Chapter 5 gives more detailed information about the apparatus and experimental methods.

4
 2 Background

After the Second World War, world steel production increased significantly. It was estimated
that it has risen from 200 to 1900 million tons from 1950 to 2020. The latest statistics suggest
that global steelmaking capacity could increase to 2485 Mt at the end of 2021, resulting in a
1.3% increase from the level at the end of 2020. Approximately 60% of this steel is produced
via the Basic Oxygen Furnace (BOF) and 34% from the electric arc furnace (EAF) process
routes. The steel production involves high-temperature physical and chemical processes and
complex three-phase flow [10], [11].

Despite the development of many industries using lime (CaO), the steel industries are still
major consumers of lime, consuming almost half of the annual global lime production. Lime is
a product of the calcination of limestone (CaCO3) extracted from quarries. In the steelmaking
processes, lime can be used in the form of high calcium and dolomitic, and in both cases it must
meet specifications regarding physical and chemical properties to form a high-quality slag [4],
[12].

In modern steelmaking, slag quality is the key to metallurgical processes. Steel melting is faster
and more heat intensive than ever before. With power-on times as short as 30 min, dosing slag
builders in the proper quantity and at the right time is crucial to provide slag foaming during
meltdown promptly. Foamy slag shields the panels from the intense arc radiation, therefore,
redirecting the heat toward the melt. Moreover, foamy slag mitigates the current swing,
improves the operative reactance, and protects the expensive refractories from overheating and
chemical attacks. Hence, all these can be highly beneficial for improving the productivity of
EAF systems and obtaining the lowest costs of operation [8].

2.1 Lime quality for the steelmaking processes

Limestone and lime are used in different forms at different steps of the steelmaking processes,
however, the type of lime used in steelmaking processes depends on the application. Lime is
used during agglomeration before sintering for balancing the acidity and as a component in
desulphurization. Most of the volume, however, is used as a fluxing agent to modify the slag
properties and chemistry in the BOF convertor, EAF, and during the secondary metallurgy/ladle
refining (LF) process. Therefore, when lime is qualified for steelmaking processes

5
characteristics including chemical composition and degree of calcination, reactivity, grain size,
and uniformity are important [4].

The chemical characteristics of lime are beyond the scope of this study. Therefore, only the
effect of grain size on the processes is briefly discussed. The grain size is considered as an
important parameter that can help to achieve better yield and allow for an easier and cleaner
working environment. The grain size distribution can be determined, for instance, by the means
of a sieve test and the results can be used to tune the dosing and conveying system. Moreover,
saltation and choking velocity rise exponentially for larger stones, increasing the demand of
carrier gas flow to avoid settlement [8].

The risk of fines being sucked into the de-dusting system is high. Therefore, lime losses in the
waste gases and further problems in dust treatment as well as hindering the injected material
yield are caused by a high percentage of fines [4]. Finer particles cause particle segregation, no
flow due to arching or ratholing and increase the wall friction, and unsteady flow phenomena
[13]. Moreover, powdery material may coat the pipes’ inner surface and the coating may keep
growing until it blocks the line [8]. Thus, steel consumers usually ask for products in the size
range of 5 to 60 mm with a requirement of less than 10% of fines below 5 mm [4].

This study will also describe the effect of the particle size on the injection process by using
different ranges of particle size. Figure 4 shows the different particle size uses in this study,
which is produced by the means of sieve shaker in the lab.

Figure 4. Different lime particle sizes used in the present study.

6
2.2 Lime injection in EAF

Electric arc furnaces (EAF) are used to produce carbon steels and alloy steels primarily by
recycling ferrous scrap by means of electricity to melt the steel at the graphite electrode [14].
Improvement in reducing electrical energy in the EAF via injection technology is well
documented using efficient designed modules for oxygen, gas, and carbon injection. The
technological efforts concentrating on the use of oxygen and carbon for injection have resulted
in attention being shifted to optimizing the use of solid materials for further improving slag
performance and recycling of by products in the EAF [9].

The injection of lime in EAF has gained interest among steelmakers due to recent improvements
by companies working on injection systems, like More, in the methods and injection equipment,
the development of optimized lime products, and the refinement of EAF slag practices.
Improvements in the environmental aspects of the plant, operational costs, and flexibility in the
design of slag chemistry have been realized by steelmakers who are using lime injection
systems [4].

Methods that are used for the addition of slag builders in the EAF can be categorized into
different types [8], [9]:

1. scrap bucket addition

2. using conveyor belt and charging through a bin placed on the EAF roof
3. pneumatical conveying and injection system through the roof of EAF
4. pneumatical conveying system and injection through EAF sidewall injectors.

Each of the above-mentioned methods has its advantages and disadvantages; therefore, it is
important to identify the most suitable method to meet all the requirements for the most
sustainable and efficient process. Although different methods have been developed to reduce
lime emission during injection processes, lime fines are still generated [12].

One of the recommended methods for charging fluxes in the EAF is the pneumatic conveying
system with sidewall panels that can overcome the drawbacks of the other methods. Figure 5
shows the sidewall pneumatic injectors in an EAF [4], [8], [9]. Figure 6 illustrates a schematic
of the apparatus used as the model for experimenting the process of injection in the lab-scale
model of the present work.

7
 Figure 5. Sidewall pneumatic flux conveying injectors [8].

Pressure
Solid feeder
gage EAF physical
Air flowmeter
model

Slag
Air inlet (paraffin oil)
Injection
nozzle Molten phase
(water)

Air-Particle stream
Vp Slag (paraffin oil)
dp
σoil
H0 ρp
μoil
𝑚̇𝑠
ρoil

Lp: particle penetration depth

Molten phase(water)
μw
ρw
σw

Figure 6. A) Schematic diagram of the experimental apparatus for measuring the penetration
depth. B) Furnace model and physical properties of different phases.

8
The reasons for the interest of steelmaking industries in lime injection through the sidewall
lance systems are [8]:
1. Improved slag properties and desire to control them
2. Injection technology improvements over the last few years
3. Efforts for cleaner shop environments for workers
4. Reduced cost of waste disposal and maintenance of material handling systems
5. Flexibility in additions of dolomitic and high calcium lime in EAF
6. Further improvements in process performance

Sidewall injectors can inject the slag formers very close to the slag surface, which causes an
increase in the yield of the injected material. By precisely controlling amounts and feed rates,
in this method, a good performance can be ensured. Therefore, this can facilitate immediate
controls on keeping slag in balance in terms of saturation, viscosity, rate of reaction, and related
parameters to foaminess during the entire process. Lime injection and control of foaming save
power-on time, energy requirements, refractories, lime consumption, and give a safer and
cleaner environment [8], [12].

9
 3 Pneumatic transport and conveying of solids

Several interrelated factors must be considered to effectively perform a powder or particulate

injection process. The dynamics of gas-solid injection involve a series of processes including
the upstream gas-solid flow, the interaction of the liquid and gas-solid stream, and the behavior
of solids and gas as they leave the injection point [15].

The behavior of a gas-particle jet on injection into a liquid is determined to a significant degree
by transport conditions of the solid and gas in the incoming stream [15]. Thus, this section is
dedicated to the definitions and concepts of the design of a conveying system.

3.1 Design of pneumatic conveying system

A pneumatic conveying system moves solids through a pipeline by the means of a gaseous
medium, usually air. The flow regime depends on the ratio of solid material to gas, and type of
material being conveyed. There are two primary modes of pneumatic transport: dense phase
and dilute phase. The dense phase conveying system occurs at low velocity (below the saltation
velocity) in plug flow, dune flow, or sliding flow. The dilute phase conveying system works
above the saltation velocity in suspended flow. A comparison of typical operating conditions
for dilute phase and dense phase pneumatic transport is shown in Table 1 [8], [16].

Table 1. Dilute phase versus dense phase pneumatic transport [16].

A key parameter for a successful operation of a pneumatic conveying system is the minimum
velocity required for particle entrainment, also known as pickup velocity. The minimum
velocity to maintain the particulate flow is called the saltation velocity. The knowledge of these
velocities is critical for an accurate design of a pneumatic conveying system. If the fluid velocity
is higher than necessary, the system is subject to unnecessary energy losses, particle attrition,
and excessive pipe erosion. A flow velocity below these critical velocities can result in clogged
dusts [8], [16], [18].

10
Both these velocities are functions of the particle properties, such as size, density and sphericity,
fluid properties, such as density and viscosity, and the main variable, the solid mass flow rate
[17].

For the specific application of flux injection into EAF, both dense phase and dilute phase are
recommended. According to Morsut et al. [8], the dense phase transportation is not a viable
option because the flux material has a saltation velocity as high as 20 m/s, and therefore, tends
to pack along the pipeline whenever it decelerates to lower speeds. This study is based on the
design criteria of the conveying system, where the saltation velocity for different particle sizes
is calculated to ensure that the required air volume flowrate is indeed available for the processes.

3.2 Saltation velocity

The saltation velocity is the gas velocity in a horizontal pipeline in which the particles begin to
fall from their state of suspension and are deposited at the bottom of the pipeline. Several
correlations have been derived regarding the saltation velocity. However, one of the most
popular is the one developed by Rizk (1973) [17]

𝑚̇s 1
𝜁= = ( 𝑑 ) Frs𝑥 (1)
𝜌f 𝑈s 𝐴 10

where the Froude number at the saltation velocity is

𝑈s
Frs = (2)
√𝑔 𝐷T
with

d = 1.44 (dp/mm) + 1.69

x = 1.1(dp/mm)+ 2.5

In the equations, ζ is the solid loading ratio (mass of solid/mass of gas), 𝑚̇s is the solid mass
flow rate, dp is the particle diameter, DT is the pipe diameter, Us is the saltation velocity, ρf is
the fluid density, g is the acceleration of gravity (m/s2) and A is the cross-sectional area of the
pipe.

11
3.3 Pickup velocity

The pickup velocity is defined as the gas velocity required to suspend the particles initially at
rest in the bottom of the pipeline or it may be defined as the fluid velocity necessary to initiate
a sliding motion, rolling, and suspension of the particles. The prediction of pickup velocity can
be estimated by using the correlation of Klinzing [17]

0.25
Vp 0.175
𝐷𝑇 𝜌p 0.75
= 0.0428 Rep ( ) ( ) (3)
√𝑔𝑑p 𝑑p 𝜌f

valid for 25 < Rep < 5000, 8 < (DT/dp) < 1340 and 700 < (ρp/ρg) < 4240. Here, 𝑉𝑝 is the particle
velocity and 𝜌p is the particle density.

3.4 Solid velocity using slip velocity

The saltation velocity calculated by Eq. (1) is defined as a superficial gas velocity based on the
tube diameter. Therefore, the actual gas velocity between particles by using voidage (or
porosity) 𝜀 can be written as [18]

𝑈s
𝑉g = (4)
𝜀
An excess gas velocity of 50% may be employed as the required gas velocity for use in the
dilute phase region. Thus, with 50% excess gas velocity the gas velocity is calculated by [19]

𝑈s
𝑉g = 1.5 (5)
𝜀
The porosity ε is, in general, close to unity [19].

The slip velocity is the resultant velocity between the fluid and solid caused by the particle-
particle and particle-wall interactions. Determination of this approximates exact modelling of
pneumatic transport systems. By using slip ratio, the solid velocity can be estimated by using
an empirical correlation called IGT [18]

0.92
𝑑p 𝜌p −0.2 𝐷T −0.54
Vr = Vg (1 − 0.68 ( ) ( ) ( ) ) m/s (6)
m kg/m3 m

where 𝑉r is the relative(slip) velocity (m/s)

12
The significance of this equation lies in the fact that the particle velocity is only a function of
the system parameters [18].

3.5 Conveying gas and solid particle frictional expressions

The expression for the pressure drop has two frictional representations which come from the
energy loss due to the conveying gas and the solid particles interacting with the wall and with
other particles. In most pneumatic transport systems, the gas flows found in a turbulent regime,
thus, the friction factor for single-phase flow in pipes is employed to represent the energy loss
for transport gas [18].

The Blasius equation (Eq. (7)) represents the friction factor (fs) for Newtonian fluids in smooth
pipes quite well over a range of Reynolds numbers from 5,000 to 100,000 [16].

f𝑠 = 0.079 Re−0.25 (7)

In this study, this equation is used to calculate of the friction factor for the gas in the pressure
drop calculations. The Reynolds number is calculated using the gas properties in the transport
line [18].

The solids’ contribution to the pressure drop includes contributions from both the particle–wall
and the particle–particle interactions. The latter is reflected in the dependence of the friction
factor fs on the particle diameter, along with the drag coefficient, density, and the relative (slip)
velocity [16]:

2
3 𝜌g 𝐷T 𝑉g − 𝑉p
𝑓s = ( ) ( ) 𝐶d ( ) (8)
8 𝜌p 𝑑p 𝑉p

This equation is used to calculate the friction factor of the solid particles in the pressure drop
calculations.

13
3.6 Terms for calculating the pressure drop for horizontal conveying

It is assumed that the overall pneumatic conveying pressure drop is the sum of all pressure
drops due to the following equations which also list the mechanisms [19].

From the Bernoulli equation, we have

1
Acceleration of the gas: 𝜌 𝜀𝑉 (9)
2 𝑔 g
1
Acceleration of the solids: 𝜌 (1 − 𝜀)𝑈s (10)
2 s
𝐿
Friction of the gas on the pipe wall: 2𝑓g 𝜌g 𝑉g 2 𝐷 𝜀 (11)
𝐿 (12)
Friction of the solids on the pipe wall: 2𝑓s 𝜌𝑠 𝑉s 2 (1 − ε)
𝐷

By using the continuity equation, Eq. (12) can be written as

𝐿
2𝑓s 𝐺s 𝑉𝑠
𝐷
(13)
where 𝐺s is the mass flux of solids.

3.7 Drag coefficient

The basic dynamic equation for a particle in pneumatic transport is [18]

d𝑉
𝑚𝑝 = external forces − drag force − deformation forces − weaker forces (14)
d𝑡

The drag term, in which the drag coefficient appears, is one of the important parameters. The
drag coefficient and relative velocity are intimately associated such that various flow regimes
need to be explored [18].

3 𝐶D 𝜌f
𝐹D = 𝑚p (𝑉 − 𝑉s ) 2 (15)
4 𝑑p 𝜌p g

The Reynolds number can be used to compare the inertial to viscous forces of the particle. Table
2 reports these coefficients in various ranges of the Reynolds number [18].

14
 Table 2. Drag coefficients CD [18].

However, according to the calculations for the pneumatic conveying system regarding different
particle sizes, based on the pipe diameter used in the experiment, and the solid mass flow rate
in the range of 2 -7 g/s, for a horizontal pipeline, the minimum required air volume flow rate is
9 - 12.9 l/min. As a result, the minimum air flow rate for the experiment reported here is 10
l/min.

15
4 Physical modeling

4.1 Principles of physical modeling

To simulate an industrial process reliably, the dimensions, the working fluid, and the
operational parameters of the physical model need to be defined properly. For this purpose,
similarity criteria are used [10]. The industrial unit, for instance, furnace, ladle, and other
equipment, is known as the prototype or the full-scale system and its physical replica as the
“physical model” or simply “model” [2].

Physical models in which water is used to simulate molten steel are known as “water models”
or “aqueous models” and have been extremely popular in steelmaking research. The physical
model is constructed based on the similarity principle. The four main similarity criteria are [2]:

1. Geometrical similarity

2. Mechanical similarity

3. Thermal similarity

4. Chemical similarity

In physical modeling, dimensions and operating conditions of a full-scale system are scaled
down (prototype to model) or those of a laboratory-scale model scaled up (model to prototype).
Geometrical similarity and mechanical similarity are usually considered during the water
modeling process [20]. A possible subdivision of physical models commonly employed in
steelmaking is shown in Figure 7 [2]. The focus of this study is on the geometrical and
mechanical similarities. Consequently, thermal conditions and chemical reactions are not
considered. Therefore, the role of geometrical and mechanical similarities is discussed further.

4.1.1 Geometric similarity

The geometrical similarity is based on the similarity of shapes. The geometrical scale factor λ,
is used to scale the industrial equipment. For a physical model with reduced dimensions clearly
λ < 1 and obviously for a full-scale physical model λ = 1 [10]. The scale factor for establishing
a physical model, in the current study, is defined as λ = 0.05.

16
 Figure 7. Modeling approaches in steelmaking [2].

4.1.2 Mechanical similarity

The similarity of forces, in the case of physical modeling, deals with the mechanical similarity
criteria. Mechanical similarity can be categorized into three classes, namely, static similarity,
dynamic similarity, and kinematic similarity. The static similarity in the EAF system is
unimportant [2].

In real steelmaking furnaces, the size of reactors is often large and the intensity of stirring, or
motion is appreciable. Furthermore, since the kinematic viscosity (= μ/ρ) of steel is extremely
small (~10−6 m2/s), the Reynolds number associated with such flows is often significant (> 104
or so). In contrast, the Froude number is two to three orders of magnitude smaller. These,
therefore, suggest that relative to the inertial forces, the contribution of viscous forces to the
flow is small and hence, can be ignored. In reduced-scale aqueous model studies of steelmaking,
flows are therefore frequently taken to be dominated by inertial and gravitational forces [2].
However, the dead zones might possibly occur in some low-velocity regions in the molten
phase. In these regions, the viscous force and buoyancy force could be important. For modeling
these effects, the equality of the Reynolds number and or the Richardson number (ratio of
thermal buoyancy to inertial forces, Tu) could be imposed on the physical model [10].

The dynamic similarity in the EAF is based on various forces acting on the liquid(s). As a result,
dynamic similarity is concerned with the similarity between pressure, inertial, viscose, surface
tension, and gravitational forces [2]. Dynamic similarity between the model and the prototype
17
in any flow system can be derived based on relevant dimensionless groups which can represent
different forces acting on the system. In order to determine the corresponding dimensionless
groups, the equations of motion or momentum conservation can be used [16].

Equation 16 represents the force balance of a particle motion in the fluid [5]

d𝑉
𝑚 = 𝐹drag + 𝐹pressure + 𝐹virtual mass + 𝐹gravitation + 𝐹contact (16)
d𝑡

where m is the mass of the particle and V is particle velocity. Moreover, the equations of
continuity and motion which control the fluid flow can be expressed by Eqs. (17-19). The terms
in Eq. (18) are numbered with Roman numbers described in Table 3 [2], [5].

∂𝜌
+ (∇𝜌𝐮) = 0 (17)
∂𝑡
∂(𝜌𝐮) (18)
+ [∇𝜌𝐮𝐮] = −∇𝑝 + 𝜇turb ∇2 𝐮 + 𝜌𝐠
∂𝑡
I II III IV
U
𝜌 = 𝛼l 𝜌𝑙 + 𝛼g 𝜌𝑔 (19)

where ρ, ρl and ρg are the mean, liquid and gas densities, respectively, u is the velocity vector,
p is the static pressure, μturb is the turbulent viscosity, g is the gravity acceleration vector, while
𝛼𝑙 and 𝛼𝑔 are the volume fraction of liquid (water) and gas (air), respectively.

Table 3. Explanation of terms in Eq. (18).

Terms Interpretation Characteristic expression of the force

I Inertial force ρv2L2
II Pressure force pL2
III Viscous force μvL
IV Gravitational force ρgL3

18
Based on the momentum conservation equation, Eq. (18), dynamic similarity between model
and the prototype requires that, at the corresponding location and time, the following
relationships hold [16]

𝐹I,mod 𝐹P,mod 𝐹𝑉,𝑚𝑜𝑑 𝐹G,mod

= = = = 𝐶F (20)
𝐹I,pro 𝐹P,pro 𝐹V,pro 𝐹G,pro

As a result, the dynamic similarity is ensured when the following equalities are satisfied at the
corresponding time and location [16]:

𝐹I 𝐹I (21)
( ) =( )
𝐹V mod 𝐹V pro

𝐹I 𝐹I (22)
( ) =( )
𝐹G mod 𝐹G pro

𝐹I 𝐹I (23)
( ) =( )
𝐹P mod 𝐹P pro

Based on the above-mentioned equations, the dynamic similarity is the ratio of different forces
that can express various dimensionless groups. If an isothermal model is considered, the
following equalities between model and prototype are important [2]:

(Re)mod = (Re)pro (24)

(Fr)mod = (Fr)pro (25)

(We)mod = (We)pro (26)

(Eu)mod = (Eu)pro (27)

where Eu is Euler number.

It is worth mentioning that dimensionless groups are generally expressed in terms of a

characteristic velocity and a length scale, and these must be appropriately defined. In many
cases, as several velocities and length scales exist, it would be difficult to define these
parameters. In such cases, dimensionless groups can be defined based on a priori knowledge of
velocity, length and other related parameters at some given points and locations in the system
[16].

19
4.2 Choice of working fluid
4.2.1 Molten phase

As shown in Table 4, water at 20 °C and molten steel at 1600 °C have nearly the same kinematic
viscosity (m2/s). This makes the model an excellent tool for investigating various transport
phenomena in a steelmaking reactor. The key objective of the physical modeling is to measure
and visualize the characteristics of an actual converter inexpensively and conveniently [20].

Table 4. Physical properties of water at 20 °C and steel at 1600 °C [20].

4.2.2 Slag

Slag plays an important role in metallurgical systems, such as protecting the metal and
removing undesirable impurities. Therefore, a liquid slag layer performs different functions,
e.g., covering the metal and preventing oxidation of metal, removing impurities like sulphur
and phosphorus from metal, reducing the heat losses from the metal surface and preventing
skull formation. In continuous casting of steel slag (created by mould powder) infiltrates
continuously between the metal and mould and provides both lubrication and control of heat
extraction. As a result, slag plays an important role in steelmaking to produce clean metals [21].

Slag can be of different types with a wide range of compositions. In metallurgical systems, slag
composition is usually based on the CaO-Al2O3-SiO2 system called metallurgical slags [21]. In
different parts of the EAF, during the operation process, there is a considerable variation in
temperature and composition that can cause the slag to become partially solid at some points.
Therefore, a change in slag viscosity due to the presence of solid material, such as undissolved
lime, is inevitable and can also affect the foamability. The term apparent viscosity is defined
for the slag as [14]

20
 μ = μ0 (1 − αf)−n (28)

where μ0 is the viscosity of the solid-free melt, α and n are constants and f is volume fraction of
solid particles in the melts. A typical viscosity diagram of Al2O3-CaO-SiO2 slag composition
based on mole fraction at 1550 °C is shown in Figure 8 [14].

Figure 8 – Typical viscosity diagram for slag [14].

However, in physical models in which water is utilized to simulate the molten phase, it is
difficult to find a material to simulate the slag to satisfy the equality of the dimensionless
numbers simultaneously. In physical models, mainly the slag is simulated by using any liquid
that is immiscible in water with a lower density. This liquid can be chosen among organic
liquids [10]. In this experiment for better visibility and environmentally friendly issues, food
grade of paraffin oil is used to simulate the slag layer. Typical physical properties of paraffin
oil are shown in Table 5.

Table 5. Physical properties of paraffin oil.

Property Paraffin oil Reference

Density (ρ), kg/m3 833 Measured as part of this work
Surface tension (σ), N/m at 10 °C 0.046 [22]
Kinematic Viscosity mm2/s at 40 °C 20.5 [23]

Appendix C presents more information regarding physical properties of the slag and methods
used for prediction of them.

21
4.3 EAF and prototype dimensions

In order to scale down the EAF by using the scale factor, the real dimensions of a prototype are
required. Based on different articles, the shape and dimensions of an EAF prototype were
estimated and, using this information, the dimension of the model was defined. The general
internal configuration of the furnace, illustrated in Figures 9 A and B, show a schematic of the
outside of an EAF [24], [25].

A B

Figure 9. A) Schematic of an internal part of an electric arc furnace. B) Schematic of the outside of an
electric arc furnace [24], [25].

Dimensions for Figure 9 A are listed in Table 6, while dimensions for Figure 9 B are listed in
Table 7 [24], [25].

Table 6. Dimensions of EAF in Figure 9A [24].

22
 Table 7. Geometry of EAF Figure 9B [25].

Based on the above-mentioned data, the selected dimensions of the prototype are reported in
Table 8. The slag height is assumed to extend to 0.2 m above the molten phase [26].

Table 8. Comparison between dimensions of prototype and the model.

Item Prototype dimensions (mm) Model dimensions (mm) λ=0.05
EAF diameter 6000 300
Slag height 200 10
Molten phase height 1000 50

4.4 Real conveying system, injection nozzle, and lime particle characteristics

The design of lime injectors was centered around the design, location in the furnace, flow rate
requirements, and reliability. It is known that the injector(s) should be located in the vicinity of
an oxygen source or burner to assist in the delivery of the product to the slag. Additionally, the
injector should have an orifice sizing of 2” to 3” angled downward at 42° to 45° and positioned
18” to 22” above the slag line. This prevents any clogging of the injector related to steel or slag
splashing [8], [9].

23
Morst et al. [8] suggest optimum conditions for the lime injection system. Recommended values
for such a system are shown in Table 9.
Table 9. Recommended conditions for the injection system [4], [8], [9].

Item value unit

Flow speed 25-35 m/s
Response time 2 sec
Distance between silo and EAF 50-70 m
Termination speed 40-50 m/s
Load factor (max) 20 kg solid/kg air
Load factor (design) 14 kg solid/kg air
Feed rate (max) 160-250 kg/min
Conveying air 12.5-13.3 Nm3/min
Pipe size for conveying 6-8 inch
Particle terminal velocity 60-80 m/s

Table 10 shows the lime characteristics for iron and steelmaking [4]. The size of the lime
particles used in the experiments was defined based on the data from Table 10 and the size of
the particles scaled down by using a scale factor equal to λ = 0.05.

Table 10. Lime characteristics for iron and steelmaking [4].

24
4.5 Characteristics of a real dust collecting system in the EAF

In the second set of experiments, the amount of dust produced during the injection process was
measured to determine the effect of the operational conditions on the dust formation. Therefore,
in this section, some information about the dust collecting system is given to gain a better insight
into the operational conditions.

In the EAF systems, the dust-laden exhaust gas is extracted through the off-gas elbow into the
post-combustion chamber and routed to the secondary dedusting equipment. The elbow
diameter, on the side of the furnace, is D = 2.2 m. The primary gas flow rate is about
56000 Nm3h-1 and the temperature is about 1900 °C [26]. Figure 12 shows the model and the
dimensions of the model which is used for the second set of experiments.

25
5 Apparatus and experimental methods

The schematic diagram of the apparatus for to measuring the particle penetration depth, for the
first set of experiments in the cold model, is shown in Figure 10. A photo of the vessel
containing water and paraffin oil, with the dimensions used for the experiment, is given in
Figure 11. This set up includes a conveying system of a transparent plastic pipe with 6 mm
inner diameter. For the solid feeder, a separation funnel with a PTFE valve was used.
Compressed air, from the laboratory air network, was used as a carrier gas for the lime particles.
A flowmeter and a pressure gage and valve were utilized to monitor and control the volume
flow rate and pressure of the air. Plastic nozzles with different diameter were used for injection
of the lime particles into the model. The angle of the nozzle and the distance between nozzle
and surface of the oil were set to 45° and 30 mm, respectively.

Pressure Solid feeder

gage EAF physical
Air flowmeter
model

Slag
Air inlet (paraffin oil)
Injection
nozzle Molten phase
(water)

Figure 10. Schematic diagram of the experimental setup for measuring the penetration depth.

A plastic container shown in Figure 11 was used as a model of EAF. A video camera (Photron
FASTCAM SA3) was applied to monitor different phase behavior and measuring penetration
depth as well as the velocities of particles. Sampling speed was set to 250 frames/s and the
shutter speed was set to 1/1500.

26
 Injection nozzle

3 mm Paraffin oil
45°
oil

Water
50 mm
300 mm

Figure 11. Photo of the plastic vessel containing paraffin oil and water with dimensions.

Figure 12 shows the model used to measure the amount of dust produced during the injection
process in the second set of experiments, while Figure 13 reports the dimensions of the model.
In this model, water, as the only liquid, was used in experiments. The operational conditions
and other equipment utilized in this set of experiments were the same as in the first set of
experiments. For collecting the dust in the exhaust, as illustrated in Figure 12, a HEPA filter
used in vacuum cleaners was utilized.

Pressure gage

Exhaust

Hepa filter Particle feeder

Flowmeter

Nozzle
Water

Figure 12. Apparatus used for the measuring of the amount of the dust.

27
 120 mm

260 mm

30mm
60 mm

300 mm

Figure 13. Dimensions of the model used in the second set of experiments.

The air velocity in the exhaust entrance was measured by an air velocity meter (COMPUFLOW
GGA-65P) to calculate the air volume flow rate through the exhaust. The measurement was
done under two different conditions: in the presence of the filter and without the filter. The
measured velocities were 1.06 m/s and 1.50 m/s, respectively. As a result, the air flow rate, in
the presence of the filter and at the entrance of the exhaust, was 46.8 Nm3/hr. These
measurements can be used for the scale-up purposes.

In the second set of experiments, the procedure used for performing the experiments is as
follows: For every experiment, 20 g of lime was weighed and then added to the feeder. Then
the HEPA filter was weighed and put in its place in the exhaust. The airflow rate was set to the
required amount and the valve was opened to start the injection. After the injection process was
completed, the air flow was stopped and after 1 minute the HEPA filter was removed from the
exhaust and was weighed. The difference between the first and the second mass of the HEPA
filter is the mass of dust. After every experiment, the filter was cleaned using pressurized air
under the fume hood.

Lime (CaO) particles used in this study were produced in the Nordkalk corporation (CAS No.
of 1305-78-8). Basic physical and chemical properties of the calcium oxide used in the

28
experiment are reported in Table 11. In order to gain a better understanding of particle behavior
based on their size, a sieve shaker device was used to classify particles in different size ranges.
The size ranges of particles are 250–500 μm, 500-710 μm, 710-1000 μm, and 1000-1400 μm.
The experimental conditions of the particle injection are given in Table 12. After producing
different particle sizes by using a sieve shaker tool, the samples were analyzed to check the
particle size distribution to ensure that they are in the appropriate size ranges. The results were
found to indicate a quite narrow particle size distribution. Figure 14 shows the results of the
particle distribution test. For the range 1000-1400 μm, due to the large size of the particles the
device could not determine the size distribution.

Table 11. Basic physical and chemical properties of the calcium oxide.
Item Unit Value
Density kg/m3 3310
Bulk density kg/m3 750–1300
Solubility in water mg/l 1337.6
Color White, light brown

Table 12. Experimental conditions of particle injection experiments.

Item Unit Value
Air volume flowrate l/min 10, 20, 30
Nozzle dimeter mm 3.5, 4, 4.4
Distance of nozzle to oil surface mm 30
Nozzle angle 45°
Vessel diameter mm 300
Water Hight mm 50
Oil Hight mm 10

29
 A

Figure 14. Results of the particle size distribution test. A) Particle size range 250-500
μm, average particle size 435 μm. B) Particle size range 500-710 μm, average particle
size 689 μm. C) Particle size range 710-1000 μm, average particle size 861 μm.

30
6 Results and discussion

In the first set of experiments, the focus was on providing new experimental data to achieve an
efficient injection process by changing the operational conditions which affect the penetration
depth of the particles as well as observation of the behavior of a three-phase system of gas, solid
and liquids (oil and water). The slag was simulated by using a paraffin oil layer to study its
influence on the efficiency of the injection process, the penetration depth, and distribution of
the particles in the liquid phases.

In the second set of experiments, the most significant objective was to optimize the particle size
of lime to reduce the amount of dust formation. Therefore, by comparing different data from
the experiments, the optimum operational conditions, such as particle size, the nozzle size, and
gas flow rate, were determined.

The results were shown in the form of graphs, which indicate the influence of the air flow rate,
the nozzle size, and the particle size on the depth of penetration and the amount of dust. In
addition, regarding the first set of experiments, recorded videos of every experiment were
investigated frame by frame and the results of observations were illustrated by different pictures
given in the following sections.

6.1 Effect of operational factors on penetration depth

6.1.1 Effect of nozzle size on penetration depth
In the following graphs, the penetration depth is shown as a function of the air flow rate, particle
size, and nozzle diameter. In Figure 15, experimental data are given for three different nozzle
sizes as well as four different ranges of particle size. It is observed that, as expected, the smaller
the nozzle diameter, the larger the penetration depth at a given gas flow rate and fixed nozzle
size. As can also be seen in the graphs, by increasing the particle size and air flow rate, the
difference between penetration depths decreased.

31
 105 A
95
85

penetration depth (mm) 75 particle size 250-500

65 μm 3.5mm
55 particle size 250-500
μm 4mm
45
particle size 250-500
35
μm 4.4 mm
25
15
5 15 25 35
air volume flow rate (l/min)

75
B
65 particle size 500-710
penetration depth (mm)

μm nozzle size
55 3.5mm
particle size 500-710
45 μm nozzle size 4mm

35 particle size 500-710

μm nozzle size
25 4.4mm

15
5 15 25 35
air volume flow rate (l/min)

40
38 C
36
penetration depth (mm)

34 particle size 710-1000 μm

32 nozzle size 3.5 mm
30 particle size 710- 1000 μm
nozzle size 4mm
28
particle size 710-1000 μm
26
nozzle size 4.4mm
24
22
20
5 15 25 35
air volume flow rate (l/min)

(Figure caption on following page)

32
 55
D
50

45

penetration depth (mm)

40 particle size 1000-1400
μm nozzle size 3.5mm
35
particle size 1000-1400
μm nozzle size 4mm
30
particle size 1000-1400
25 μm nozzle size 4.4mm

20

15
5 10 15 20 25 30 35
air volume flow rate (l/min)

105
E
95

85
penetration depth (mm)

75

65

55

45

35

25

15
5 10 15 20 25 30 35
air volume flow rate (l/min)
Linear (particle size 250-500 μm 3.5mm ) Linear (particle size 250-500 μm 4mm)
Linear (particle size 250-500 μm 4.4 mm) Linear (particle size 500-710 μm nozzle size 3.5mm )
Linear (particle size 500-710 μm nozzle size 4mm ) Linear (particle size 710-1000 μm nozzle size 3.5 mm )
Linear (particle size 710- 1000 μm nozzle size 4mm) Linear (particle size 710-1000 μm nozzle size 4.4mm)
Linear (particle size 1000-1400 μm nozzle size 3.5mm) Linear (particle size 1000-1400 μm nozzle size 4mm)

Figure 15. Effect of nozzle size on penetration depth. Panels A, B, C, and D show the penetration
depth vs air flow rate for the same particle size and different nozzle sizes. Panel E shows a combination
of the charts in the previous panels.

33
6.1.2 Effect of particle size

Figure 16 depicts experimental data for four different ranges of particle size with the same
nozzle diameter and three different air flow rates. It is seen that the smaller the particle size, the
larger the penetration depth at a given gas flow rate and nozzle size.

nozzle size = 3.5mm

A
100
90
penetration depth (mm)

80 10 l/min 20 l/min 30 l/min

70
60
50
40
30
20
200 300 400 500 600 700 800 900 1000

particle size (μm)

nozzle size = 4mm

B
80
10 l/min 20 l/min
70
penetration depth (mm)

30 l/min
60

50

40

30

20
200 300 400 500 600 700 800 900 1000
particle size (μm)

(Figure caption on following page)

34
 nozzle size = 4.4 mm
C
10 l/min
55
20 l/min
50

penetration depth (mm) 45 30 l/min

40
35
30
25
20
15
200 400 600 800 1000
particle size (μm)

95

85

75
penetration depth (mm)

65

55

45

35

25

15
200 300 400 500 600 700 800 900 1000
particle size (μm)

Linear (10 l/min 3.5mm) Linear (20 l/min 3.5mm) Linear (30 l/min 3.5)
Linear (10 l/min 4mm) Linear (20 l/min 4mm) Linear (30 l/min 4mm)
Linear (10 l/min 4.4mm) Linear (20 l/min 4.4mm) Linear (30 l/min 4.4mm)

Figure 16. Effect of particle size on the penetration depth. Panels A, B, and C show the
penetration depth vs. particle size for the same nozzle size and different air flow rates. Panel D
shows a combination of the charts in previous panels.

35
6.1.3 Effect of air flow rate

Figures 15 and 16 present the obvious influence of air flow rate in different particle size ranges
and nozzle sizes. Based on these figures, it can be stated that the penetration depth increases
with an increase in the air flow rate. For the particle size range 710-1000 μm, it appears that the
air flow rate has the lowest impact on the penetration depth as the results are very similar.

6.2 Measurement of particle velocity

An important issue regarding the study of the penetration is the particle velocity. It is easy to
measure the velocity of an isolated particle, but for a stream of particles, it is quite difficult to
quantify the velocity. For this part of the study, particle velocity was measured by a high-speed
camera in different particle size ranges, air flow rates, and nozzle sizes by conducting thirty-six
experiments. The sampling speed of the camera was set to 250 fps and the shutter speed was
set to 1/10,000. Videos from the high-speed camera were analyzed by the Tracker software [27]
in order to follow the particle movement when leaving the nozzle. For each experiment, the
speed of more than 10 particles was measured, while the average of these was taken as the mean
velocity of particles. The ratio of particle velocity to air velocity at the nozzle is shown in Figure
17. The velocity ratio is seen to be between 20 % and 30%, and the average ratio, for the particle
size and size of the nozzle studied, is 28.7%. It must be mentioned that the air velocity reported
is the superficial velocity of the air calculated at the nozzle outlet diameter and based on the air
volume flow rate, so the volume occupied by the particles is excluded.

70
(particle velocity/gas velocity )*100

60

50

40

30

20

10

0
0 10 20 30 40 50 60
gas velocity at nozzle exit (m/s)

Figure 17. Ratio of mean particle velocity to carrier gas velocity at the nozzle exit.

36
6.3 Effect of operational factors on dust formation

The second set of experiments was performed to measure the amount of dust “produced” during
the lime injection process. By dust is here meant those of the injected particles that will bounce
off from the surface without being trapped by the liquids. In this section, the influence of
operational conditions, including the air flow rate, the particle size, and the nozzle size, was
investigated. The results are given as graphs which indicate the influence of the conditions on
the dust formation.

Figure 18 shows the results from the second set of experiments. The amount of dust is depicted
on the ordinate and the particle size is on the abscissa. Each graph shows the effect of nozzle
size with the same air flow rate and different particle sizes. Based on these graphs, an increase
in particle size is, quite logically, seen to cause a decrease in the amount of the dust. It is difficult
to determine the effect of nozzle size on the dust formation but from Figure 18 A and B it can
be concluded that the smaller the nozzle size, the smaller the amount of dust. Based on Figure
18 D, under three different operational conditions, the lowest amount of dust was produced for
1) particle size range 710–1000 μm, the air flow rate 20 l/min, nozzle size 3.5 mm, 2) particle
range 1000–1400 μm, air flow rate 10 l/min, nozzle size 3.5 mm, and 3) particle range 1000–
1400 μm, air flow rate 20 l/min, nozzle size 4 mm.

37
 0.14
A
0.12

0.1
air flowrate 10 l/m,
dust (gr)

0.08 nozzle size 3.5 mm

0.06 air flowrate 10 l/m,
nozzle size 4 mm
0.04
air flowrate 10 l/m,
0.02 nozzle size 4.4 mm

0
0 200 400 600 800 1000 1200
particle size (μm)

0.14
B
0.12

0.1
air flowrate 20 l/m,
dust (gr)

0.08
nozzle size 3.5 mm
0.06 air flowrate 20 l/m,
nozzle size 4 mm
0.04 air flowrate 20 l/m,
nozzle size 4.4 mm
0.02

0
0 200 400 600 800 1000 1200
particle size (μm)

(Figure caption on next page)

38
 0.14
C
0.12

0.1

air flowrate 30 l/m,

dust (gr)
0.08
nozzle size 3.5 mm
0.06 air flowrate 30 l/m,
nozzle size 4
0.04 air flowrate 30 l/m,
nozzle size 4.4 mm
0.02

0
0 200 400 600 800 1000 1200
particle size (μm)

Chart Title
D
0.14
10 l/m, 3.5 mm
0.12
10 l/m, 4 mm
0.1
10 l/m, 4.4 mm
dust (gr)

0.08 20 l/m, 3.5 mm

0.06 20 l/m, 4 mm
0.04 20 l/m, 4.4 mm

0.02 30 l/m, 3.5 mm

0 30 l/m, 4 mm
0 200 400 600 800 1000 1200
30 l/m, 4.4 mm
particle size (μm)

Figure 18. Results from the second set of experiments showing the amount of dust produced
vs. particle size. Each graph shows results for different nozzle sizes, while the air flow rate
is the same for each experiment.

Figure 19 shows the influence of different air flow rates with the same nozzle size but different
particle sizes. From these graphs, it is difficult to determine the operational condition with
minimum dust formation. In general, it can be stated that by increasing the air flow rate the
amount of dust decreased, obviously because the penetration of the jet, including the particles,
becomes more substantial. An increase in the particle size also yields a decrease in the amount
of dust.

39
 0.14

0.12

0.1
dust (gr)

0.08 air flowrate 10 l/m,

nozzle size 3.5 mm
0.06
air flowrate 20 l/m,
0.04 nozzle size 3.5 mm
air flowrate 30 l/m,
0.02 nozzle size 3.5 mm

0
0 200 400 600 800 1000 1200
particle size (μm)

0.14

0.12

0.1
air flowrate 10 l/m,
dust (gr)

0.08 nozzle size 4 mm

air flowrate 20 l/m,
0.06
nozzle size 4 mm
0.04 air flowrate 30 l/m,
nozzle size 4
0.02

0
0 200 400 600 800 1000 1200
particle size (μm)

0.14

0.12

0.1
air flowrate 10 l/m,
dust (gr)

0.08
nozzle size 4.4 mm
0.06 air flowrate 20 l/m,
nozzle size 4.4 mm
0.04 air flowrate 30 l/m,
nozzle size 4.4 mm
0.02

0
0 200 400 600 800 1000 1200
particle size (μm)

Figure 19. Influence on dust formation of different air flow rates with the same nozzle size
and for different particles sizes.

40
6.4 Visual observations
6.4.1 Particle rebounding and dust formation pattern

In the process of particle injection, when a particle hits the gas-liquid interface which is of much
larger size, different scenarios might be considered based on the physical characteristics of the
system, such as retained at, penetrating through, or rebounding off the interface. Such
interaction behavior also depends on the contact angle forming on the particle surface, where
the gas-liquid interface attaches as well as the movement of the gas-liquid interface relative to
the particle velocity [28]. Particle rebounding is considered to be a major problem with respect
to loss of lime during injection process that must be mitigated. The other important issue that
causes an unhealthy working environment and emissions is the dust formation. Therefore, one
of the aims of this study is to determine the pattern of dust formation during the injection
process. It is difficult to analyze the results from high-speed camera using image analysis
programs since the boundary layers of the dust was difficult to detect by the software. Therefore,
in this manuscript, it was attempted to compare the images and videos by visual observations
to depict the possible pattern of the dust which is formed during the injection process.

Figure 20 shows several pictures of particles rebounding off the interface. For the particle size
range 250-500 μm and the air flow rate 10 l/min, the number of rebounded particles is
considerable. By increasing the air flow rate and the size of the particles, the number of
rebounded particles decreased.

Figure 21 shows the dust formation pattern in different particle size ranges, air flow rates, and
nozzle sizes. The red lines in these pictures indicate the boundary layer of the dust. According
to the visual observations, by increasing the size of the particles and the air flow rate, the amount
of the dust decreased. For size ranges bigger than the range 710-1000 μm, it was difficult to see
the dust formed by the injection process.

41
 A

Figure 20. Particles rebounding off the interface for air flow rate 10 l/min and particle size 250-
500 μm. A) Nozzle size 3.5 mm, B) Nozzle size 4 mm. C) Nozzle size 4.4 mm.

42
(Figure caption on next page)

43
Figure 21. Dust formation patterns: red lines indicate the boundary layers formed by the dust
produced during the injection processes. Description is given in Table 13.

Table 13. Operational conditions for Figure 21.

air flow rate particle size range nozzle size

A 10 l/min 250-500 μm 3.5 mm
B 20 l/min 250-500 μm 3.5 mm
C 30 l/min 250-500 μm 3.5 mm
D 30 l/min 250-500 μm 4.4 mm
E 30 l/min 500-710 μm 3.5 mm
F 30 l/min 500-710 μm 4 mm
G 30 l/min 500-710 μm 4.4 mm

44
6.4.2 Cavity formation and penetration depth

This study considers two different definitions for the cavity characteristics: particle penetration
depth and cavity depth. The definitions of the cavity depth and the penetration depth of the
particles are shown in Figure 22. As can be seen, the cavity depth is the distance between the
bottom of the cavity and the undisturbed upper oil surface (see vertical arrow). The penetration
depth, in turn, is the distance between the point where the particle hits the surface and the point
in the liquid where it has decelerated; this distance is along the particle path when it leaves the
nozzle (tilted arrow). According to the measurements and graphs which are based on the cavity
depth and penetration depth versus operational conditions, the two parameters follow the same
pattern.

A B

Figure 22. Definition of A) cavity depth and B) penetration depth.

6.4.3 Gas-particle behavior while penetrating fluids

Based on the visual observations of the high-speed camera videos, gas-particle behavior for
different operational conditions was investigated. In this study, these behaviors are explained
based on the observations and the works of others.

Different studies using water models have shown that the most important variables affecting
the gas-particle behavior are the particle size and loading (defined as the mass flow rate ratio
of particles to gas). Farias et al. [7] developed a criterion to predict the transition between
bubbling, when gas separates from particles and form bubbles, and jetting, when gas and
powder flow together. Their study was performed for the submerged powder injection. In
another work by Kimura [6], fluxes were blown through lances together with the air from the
top of the liquid surface. The behavior of the flux injection was like the results obtained by

45
submerged injection. Cavity formation was similar to cavities reported for gas-only injection
and penetrating jet formation, where gas and powder follow together and no cavity was
observed. It was also found that particle size and loading influence the transition between jetting
and bubbling.

In the current study, owing to the presence of oil layer mimicking slag, the gas-particle behavior
was different depending on the particle size, air flow rate, and nozzle size. Figure 23 illustrates
the behavior of particles in which gas and powder flow together. In these pictures, different
steps during the injection process and at its end are shown. The particle size range was 250–
500 μm, the air flow rate was 10 l/min, and the nozzle size was 3.5 mm; the gas and particles
in the liquid seem to behave like jetting and separation of gas and particles does not occur.
Particles appear to become trapped in the boundaries formed by an oil layer. After penetrating
the molten steel (water), they move back to the slag layer. The same behavior was observed
with other nozzle sizes, the only difference being that the penetration of particles decreased by
increasing the size of the nozzle, and the particles mostly remain in the oil layer.

1 2

4
3

5
clean and clear both
slag and molten
phase

Figure 23. Behavior of trapped particles in the oil layer.

46
However, by increasing the air flow rate and with the same particle size range, observations
showed that bubbles in which particles were trapped had formed, and then they carried the
particles to the surface. Figure 24 illustrates the behavior of particles in the fluid where the air
flow rates are 20 l/min and 30 l/min. In these pictures, as can be seen, although the separation
between particles and air occurred, bubbles that contain many particles still formed and carried
particles up to the oil layer. By increasing the air flow rate (Figure 24 B) the number of bubbles
increased and their sizes also decreased.

Figure 24. Gas/particle penetrating the liquid for a nozzle size of 3.5 mm. A) Particle size
250-500 μm, air flow rate 20 l/min. B) Particle size 250-500 μm, air flow rate 30 l/min.

The same behavior was observed in the particle size range 500-710 μm. By increasing the air
flow rate, bubbling also occurred. For this particle size range, a separation between particles
and air occurred, while the penetration depth increased and the dispersion of particles in the
liquid improved. Figure 25 shows the behavior of particles range 500-710 μm, nozzle size 3.5
mm and different air flow rates.

47
 A

Figure 25. Gas-particle penetrating the liquid for a nozzle size of 3.5 mm. A) Particle size
500-710 μm, air flow rate 10 l/min. B) Particle size 500-710 μm, air flow rate 20 l/min, C)
Particle size 500-710 μm, air flow rate 30 l/min.

In the particle size range 710-1000 μm, due to the poor penetration when the air flow rate is 10
l/min all particles remain in the oil layer, and then they start settling because of the gravity. By
increasing the air flow rate, a normal cavity similar to cavities reported for gas-only injection,
formed and separation between air and particles occurred while bubbles still formed. According
to Figure 26 D, for an air flow rate of 30 l/min, particle penetration improved and bubbling
increased at the same time, and the distribution of particles in the liquid phase also enhanced.

48
The best possible form of distribution in the liquid phase occurred for this particle size range
and when the air flow rate was 30 l/min.

Figure 26. Gas-particle penetrating the liquid for a nozzle size of 3.5 mm and particle size
710-1000 μm. Air flow rate of A) 10 l/min, B) 20 l/min, and C) 30 l/min.

6.4.3.1 Comparison between gas-particle behavior with and without oil layer

In order to ensure that the observed gas-particle jet behavior is because of the oil layer, a
separate set of experiments was performed in which a model containing only water was used.
For this set of experiments, the particle size range was 250–500 μm, and the air flow rates were
10 l/min, 20 l/min, and 30 l/min, and the nozzle size was 3.5mm. As shown in Figure 27, the
gas-particle behavior completely differs from that of the previous experiments: the gas and
particles start to follow each other, and the separation of particles and air occurs after they

49
penetrate the liquid phase. The shape of the cavity was difficult to observe. Moreover, the
distribution of particles was quite good and the penetration of particles in the model containing
only water was smaller. Thus, the effect of the oil layer on the formation of jetting and gas-
particle separation is obvious.

Without oil With oil

A B

Separation of gas and particles happened

C D

E
F

Figure 27. Gas-particle behavior in models with and without oil. Right panels show a model
without oil with different air flow rates 10 l/min, 20 l/min, and 30 l/min (top to bottom). Left
panels show corresponding results for the model with an oil layer.

In another experiment for a comparison between the model with and without oil layer, particles
injection in the range of 710-1000 μm was examined. The results were interesting. In this
particle size range, the patterns of distribution of the particles were similar. The penetration
depth was 21.9 mm, and 23.77 mm for the model with and without oil, respectively. Figure 28
shows the particle behaviors for both cases.

50
 A B

Figure 28. Comparison of results for models A) with oil layer and B) without oil. Air flow
rate 30 l/min, nozzle size 3.5 mm, particle size 710–1000 μm.

Based on the results of the above experiments, it can be stated that by increasing the size of
particles at the same air flow rate, the boundary layers of the particles no longer overlap and
the gas is uncoupled and can form bubbles. For fine particles which are traveling at near the
same velocity as gas, the boundary layer becomes relatively large. As a result, most of the gas
travels with the particles in the coupled state, which can be seen in Figure 29 A [7].
It is easy to measure the velocity of one particle and follow its trajectory, but it is difficult to
measure the velocity of particles in a stream of gas and particles. In this study, the average
speed of particles of different sizes was estimated by using photos from a high-speed camera
with e a sampling speed of 250 fps and the shutter speed of 1/10000 s-1. Figure 29 shows the
behavior of particles of different size ranges captured by the high-speed camera, where the
boundary layer overlap of particles can be seen.

51
 A B

C D

Figure 29. Behavior of gas-particle streams. Nozzle size 3.5 mm and air flow rate 30 l/min.
Particle size A) 250–500 μm, B) 500–710 μm, C) 710–1000 μm and D) 1000–1400 μm.

6.4.4 Comparison between cavity shape formed by particle injection and air blowing

The effect of particles on the shape of the cavity was examined as well. In these series of
experiments, only air was injected into the model to observe the effect of the presence of the
particles on the shape and the depth of the cavity. The same operational conditions as the
particle injection experiments were applied using the model containing both water and oil.
Figure 30 illustrates the effect of the size of the particles on the shape and depth of the cavity

52
with the same air flow rate and the same nozzle size. Figure 30 A shows the cavity formed by
injecting only air, the depth of the cavity was 35.6 mm. Figure 30 B shows the cavity formed
by the particle size range 250–500 μm: the cavity depth (i.e., the deepest cavity formed) was 60
mm. Figure 30 C shows the cavity shaped by the particle size range 500-710 μm, with a depth
of 50.8 mm. In both cases, the depth of the cavity was higher than for the case with only air
injection. Figure 30 D shows the results for the particle size range 710–100 μm, yielding a
cavity depth of 26.4 mm. Finally, Figure 30 E shows that the particle size range 1000–1400 μm
gave rise to a cavity depth of 37.5 mm, which was also larger than for the case with only air.

A B

C D

E
Q

Figure 30. Comparison between arising cavity shape for air-only injection and co-injection
of air and particles for a nozzle size of 3.5 mm and an air flow rate of 30 l/min. A) Air
injection without particles. Injection of particles B) 250–500 μm, C) 500–710 μm, D) 710–
1000 μm, and E) 1000–1400 μm.

53
According to Figure 30, different forms of cavities with different depth were observed
depending on the size of the particles. For Figure 30 B and C, due to the small size of particles
and formation of the jet, a gas-liquid emulation is formed at the injection area. Therefore, the
average density of the emulsion phase is much smaller than that of a pure liquid. As a result, it
can improve the penetration depth of particles. Kimura [6] explained the role of solid particles
by using a simplified model, as shown in Figure 31, where an isolated particle that collides on
the surface and penetrates liquid leaving a vacant channel, as shown in panel (a). Then, the
channel will be filled up quickly (b). If the feeding rate is large enough, the second particle
closely follows the first one and collides with the surface of the liquid before the vacant channel,
formed by the first one disappears (c and d). Under such conditions, fine bubbles may form and
the life of the bubble will become longer and may temporary give rise to a gas-solid emulsion.

Figure 31. Penetration mechanism of solid particles at the liquid surface [6].

In the gas-solid injection process, where the size of particles increases, the overlap of the
boundary layer between particles disappears and they behave independently. Under such
conditions, the cavity in Figure 30 D and E occurred. As a result, the penetration depth is
smaller, and particles penetrate mostly by gravity. In this case, the role of surface tension in
determining the position of the particles concerning the gas-liquid or oil-water interface, as they
float on the surface, is important. In the case of smaller particle size, where the behavior of the
gas-solid streams is like a jet, the gas-solid mixture acts as a continuum fluid [7].

6.4.5 Particle distribution in the molten phase

For the particles injected into the molten phase, from the viewpoint of the reaction between
particles and steel melt to occur in different parts of the phase, it is necessary to distribute them
uniformly in different zones of the liquid. Different patterns of particle distribution in the liquid

54
phase were observed in the experiments that are affected by the size of the particles and the air
flow rate. Figure 32 shows pictures of the distribution pattern of the particles in the liquid phase
for a nozzle size of 3.5 mm, air flow rates of 20 l/min and 30 l/min, and different particle size
ranges in order to find the optimum distribution of the particles in the molten phase by
comparing different operational conditions. As shown in Figure 32, the optimum distribution
of particles occurs for the particle size range 710–1000 μm, and for both air flow rates 20 l/min
and 30 l/min. The distribution patterns were practically equal in these cases.

However, for air flow rates of 10 l/min, the particles first remained in the oil layer, and then
started to settle down by gravity. Figure 33 shows the particle distribution in the oil layer for
the particle size range 710-1000 μm and the air flowrate 10 l/min. For nozzle sizes of 4.0 mm
and 4.4 mm, due to the poor penetration depth of the particles the same issue was observed, and
particles mostly remained in the oil layer. Thus, it can be stated that the optimum particle
distribution in the system occurs for the particle size range 710-1000 μm, nozzle size 3.5 mm,
and air flow rates higher than 10 l/min.

55
 Air flowrate 20 l/min Air flowrate 30 l/min

A20 lm - 250 - 3.5mm B

C D

E F

G H

Figure 32. Observed distribution patterns of particles in the liquid phase. Particle size A, B)
250-500 μm, C, D) 500-710 μm, E, F) 710-1000 μm, and G, H) 1000-1400 μm.

56
Figure 33. Particles remain in the oil layer due to poor penetration in the liquid. Particle size
710-1000 μm, air flowrate 10 l/min, and nozzle size 4.0 mm.

6.4.6 Behavior of liquids

6.4.6.1 Mixing behavior

A significant phenomena in the steelmaking processes is the mixing and behavior of molten
phases that is influenced by the air injection. Therefore, intensive work has been undertaken to
clarify the interaction between the injected gas and molten liquid as well as the mixing behavior.
The interaction between gas and liquids is a complicated phenomenon and in this study the
formation of cavity as a typical characteristic of this interaction was investigated. The cavity
dimension, which is closely related to the interfacial area in actual production, is influenced by
the nozzle diameter, the nozzle angle, the distance between nozzle and the slag surface, the air
flow rate, and the slag properties [20]. Moreover, according to the observations from the current
study, a good cavity profile can help penetration and mixing effects in the furnace. As a result,
in order to determine the effect of the oil layer on the cavity formation, experiments were
conducted by injecting air into the model under different operational conditions with and
without an oil layer.
Figure 34 shows the effect of air flow rate and nozzle size on the diameter of the cavity. An
increase in air flow rate as well as a decrease in nozzle diameter is seen to result in an increase
in the length of the cavity.

57
 100
90
80

diameter of caviy mm
70
60
50
40
30 nozzle size 3.5 mm nozzle size 4 mm
20
10 nozzle size 4.4
0
0 5 10 15 20 25 30 35
air flow rate (l/min)

70

60
diameter of caviy mm

50

40

30
nozzle size 3.5 mm, with oil
20
nozzle size 4 mm, with oil
10
nozzle size 4.4 mm, with oil
0
0 5 10 15 20 25 30 35
air flow rate (l/min)

Figure 34. Effect of air flow rate and nozzle size on the diameter of the cavity A) without oil
layer, and B) with oil layer.

Figure 35 illustrates the effect of the oil layer on the diameter of the cavity. Because of the
presence of the oil layer, the diameter of the cavity decreased. Figure 36 also shows the
difference between the cavity depth and the diameter in the models with the oil layer and
without the oil layer.

58
 100

90

diameter of caviy (mm) 80

70

60

50

40 nozzle size 3.5 mm

nozzle size 4 mm
30 nozzle size 4.4
nozzle size 3.5 mm, with oil
20 nozzle size 4 mm, with oil
nozzle size 4.4 mm, with oil
10
0 5 10 15 20 25 30 35
air flow rate (l/min)

Figure 35. Effect of oil layer on the diameter of the cavity

A B

Figure 36. Difference between the cavity depth and the diameter for experiments with a
nozzle size of 4 mm and an air flow rate of 30 l/min in models A) with oil layer, and B)
without oil layer.

In videos captured by the high-speed camera from the top view, the behavior of the liquid phase
was observed. By blowing the air into the liquid the cavity forms and the slag layer, moves in
a circular motion as shown in Figure 37.

59
 cavity

Motion
of slag

Figure 37. Top view of cavity formation and movement of slag influenced by the air blowing.

6.4.6.2 Splashing behavior and droplet generation

Splashing refers to the liquid projected from the bath including the material ejected from the
vessel, whereas droplet generation refers to the part of splash that ends up in the emulsion as
distinct droplets. Droplet generation has advantages and disadvantages. Although droplet
generation can increase the interfacial area, which, in turn, increases the refining rate, splashing
and droplet generation may cause clogging of the injector nozzle, which can result in a loss of
quality [20] and splashing on the walls results in loss of yield. Observations in the experiments
showed that the droplets and the splashes generated during air blowing and particle injection
occurred far enough from the nozzle outlet to avoid clogging effects. The distance between
nozzle and the slag surface clearly prevents clogging on the nozzle outlet. Furthermore, the
amount of splashes and the droplets generated during the injection process of the particles
decreased significantly compared to the air-only blowing case (without the presence of the
particles). Figure 38 shows the difference between the splashes and the droplets generated under
the conditions where A) particles are injected and B) only air is injected.

60
 A B

Figure 38. Difference between splash and droplet generation during A) particle injection and
B) air-only blowing.

Finally, by sampling splashes and droplets, the ratio of oil and water contained in the splashed
phases was measured. As a result, the volume ratio of oil to water was quantified and found to
be between 10% and 15%. Figure 39 shows the process of the sampling and the way to
determine the ratio of oil to water by gravity separation.

Oil layer

Water layer

Figure 39. Process of splash sampling and for determining ratio of oil to water in samples.

61
7 Conclusions

The current study was undertaken as project for Nordkalk corporation as part of a Business
Finland co-funded project on fossil-free steelmaking focusing on lime injection in the electric
arc furnace (EAF). The aims of the study were defined as follows:
• Gaining a key knowledge of the lime injection process.
• Minimizing the dust formation through optimal particle size
• Investigating the dust formation patterns in comparison to particle size
• Understanding the multiple phase behaviors, including solid, liquid, and gas interactions,
in the injection process.

To pursue the aims of the study, a small apparatus was designed to perform experiments to
study the main parameters in regard to the particle penetration depth, the amount of dust
produced, the dust formation pattern, and the multiphase flow behavior influencing the injection
process. The test rig used water to mimic steel (the heavier liquid phase) and paraffin oil to
mimic slag (the lighter liquid phase) in a stratified liquid system, where air with or without lime
particles was injected through a small nozzle arranged above the liquids in a certain angle. The
system was observed visually and recorded by a high-speed camera. The results were illustrated
in the form of graphs and images from the high-speed camera. These were analyzed in detail
and the results were compared and discussed. Based on the observations and the analysis of the
results, the most important outcomes of the study can be concluded as follows:
1) The particle penetration depth increased with an increase in the air flow rate, a decrease
in the particle size, and a decrease in the nozzle size.
2) For the particle size range 710-1000 μm and air flow rates 20 l/min and 30 l/min, the
optimum distribution of the particles in the liquid phase occurred. For air flow rate10
l/min, particles tend to remain in the oil layer.
3) Dust formation and particle rebounding tend to decrease by an increase in particle size,
air flow rate and a decrease in nozzle size.
4) For the smaller particle size ranges 250-500 μm and 500-710 μm and for every nozzle
size and air flow rate, particles move with the air stream in the form of a jet when
penetrating in the liquid phase. Consequently, separation between air and particles was
not possible. Additionally, the particles penetrating in the liquid phase were trapped by
a boundary layer formed by oil, after which they were carried up to the oil layer. For the

62
 same operational condition performed in a model without oil, the separation between
particles and air occurred and the particle distribution in liquid was quite good.
5) The average ratio of mean particle velocity to superficial air velocity at the nozzle outlet
was 29%.
6) Samples of splashes showed a high proportion of water in the splashes. For the nozzle
setup studied in the experiments, the risk of clogging of the nozzle with splashes and
droplets seems very low.

Despite these findings, many challenges still have to be addressed in future works. For instance,
in the EAF systems the process of stirring the molten phase is performed by a bottom injection
of a gas and this effect could be studied in the cold model as well. In addition, conducting
experiments by using different liquids with different density and viscosity could also provide
more generic results to derive a correlation regarding prediction of penetration depth.
In this study, the density of the particles was higher than the density of the fluids, but in real
systems the density of molten phase is higher than the density of the particles. By using liquids
with higher density or different types of particles, the buoyancy effect can be simulated under
conditions that better resemble those in the real EAF.

63
References

[1] “Accelerating the green transition – Research is fostering the sustainable reform of the
metal and steel industry.” https://www.businessfinland.fi/en/whats-new/news/cision-
releases/2021/research-is-fostering-the-sustainable-reform-of-the-metal-and-steel-industry
(accessed Dec. 22, 2022).
[2] D. Mazumdar and J. W. Evans, Modeling of steelmaking processes. CrC Press, 2009.
[3] M. Kovačič, K. Stopar, R. Vertnik, and B. Šarler, “Comprehensive electric arc furnace
electric energy consumption modeling: A pilot study,” Energies, vol. 12, no. 11, p. 2142,
2019.
[4] S. Manocha and F. Ponchon, “Management of lime in steel,” Metals, vol. 8, no. 9, p. 686,
2018.
[5] T. Okuno, M. A. Uddin, Y. Kato, S. B. Lee, and Y. H. Kim, “Effect of particle penetration
depth on solid/liquid mass transfer rate by particle blowing technique,” ISIJ Int., vol. 57,
no. 11, pp. 1902–1910, 2017.
[6] E. Kimura, “Fundamental research of lancing mechanism in Mitsubishi continuous
smelting furnace,” Trans. Iron Steel Inst. Jpn., vol. 23, no. 6, pp. 522–529, 1983.
[7] L. R. Farias and G. A. Irons, “A unified approach to bubbling-jetting phenomena in powder
injection into iron and steel,” Metall. Trans. B, vol. 16, no. 2, pp. 211–225, 1985.
[8] S. Morsut, S. Marcuzzi, and M. Medeot, “Process optimization by pneumatic injection of
slag builders in the electric arc furnace”. technical paper: https://more-oxy.com/wp-
content/uploads/2020/11/Process-optimization-by-pneumatic-injection-of-slag-builders-
in-the-EAF-convertito.pdf.
[9] Wolfe, Larry, Jean Paul Massin, Tuna Hunturk, and Walter Ripamonti. "Lime Injection
Technology–A Viable Tool for The Electric Arc Furnace." technical paper:
http://www.carmeusena.com/sites/default/files/brochures/steel/tp-lime-inj-eaf-2008.
pdf (2008).
[10] F. Gonçalves et al., “Measurements methodologies for basic oxygen furnace cold
modeling,” J. Mater. Res. Technol., vol. 13, pp. 834–856, 2021.
[11] F. Mercier, “Steel Market Developments 4 2021,” Organ. Econ. Co-Oper. Dev., Feb.
2022, [Online]. Available: https://one.oecd.org/document/DSTI/SC(2021)9/FINAL/en/pdf
[12] “lime-essential-chemical.pdf.” Accessed: Dec. 21, 2022. [Online]. Available:
https://www.lime.org/documents/publications/free_downloads/lime-essential-
chemical.pdf
[13] T. A. Royal and J. W. Carson, “Fine powder flow phenomena in bins, hoppers and
processing vessels,” in Conference: Bulk. Jenike And Johanson Science Engineering
Design, 2000, pp. 1–10.
[14] “steeluniversity - Learning for the steel industry,” Jul. 04, 2016.
https://steeluniversity.org/ (accessed Dec. 22, 2022).
[15] G. G. Richards, “Fundamental aspects of gas-particle injection,” in Materials Handling
in Pyrometallurgy, Elsevier, 1990, pp. 3–13.
[16] R. Darby and R. P. Chhabra, Chemical engineering fluid mechanics. CRC Press, 2016.
[17] L. M. Gomes and A. L. Mesquita, “On the prediction of pickup and saltation velocities
in pneumatic conveying,” Braz. J. Chem. Eng., vol. 31, pp. 35–46, 2014.
[18] G. E. Klinzing, F. Rizk, R. Marcus, and L. S. Leung, Pneumatic conveying of solids: a
theoretical and practical approach, vol. 8. Springer Science & Business Media, 2011.
[19] R. Holdich, Fundamentals of particle technology. MidlandIT, 2020.
[20] Cao, L., Wang, Y., Liu, Q. and Feng, X., 2018. Physical and mathematical modeling of
multiphase flows in a converter. ISIJ International, 58(4), pp.573-584.

64
[21] K. C. Mills and S. Sridhar, “Viscosities of ironmaking and steelmaking slags,” Ironmak.
Steelmak., vol. 26, no. 4, pp. 262–268, 1999.
[22] M. SATO, “Studies on the Wetting Effect and the Surface Tension of Sotids The Surface
Tensions of Paraffin, Beeswax and Japan Wax,” Proc. Jpn. Acad., vol. 30, no. 5, pp. 369–
373, 1954.
[23] “Haku | Rakennuskemia Oy.” https://docs.rakennuskemia.com/msds/6418091102574
(accessed Dec. 22, 2022).
[24] J. C. Gruber, T. Echterhof, and H. Pfeifer, “Investigation on the influence of the arc
region on heat and mass transport in an EAF freeboard using numerical modeling,” Steel
Res. Int., vol. 87, no. 1, pp. 15–28, 2016.
[25] E. Khodabandeh, M. Ghaderi, A. Afzalabadi, A. Rouboa, and A. Salarifard, “Parametric
study of heat transfer in an electric arc furnace and cooling system,” Appl. Therm. Eng.,
vol. 123, pp. 1190–1200, 2017.
[26] H.-J. Odenthal, A. Kemminger, F. Krause, L. Sankowski, N. Uebber, and N. Vogl,
“Review on modeling and simulation of the electric arc furnace (EAF),” Steel Res. Int., vol.
89, no. 1, p. 1700098, 2018.
[27] “Tracker Video Analysis and Modeling Tool for Physics Education.”
https://physlets.org/tracker/ (accessed Jan. 30, 2023).
[28] S. Mitra, E. Doroodchi, V. Pareek, J. B. Joshi, and G. M. Evans, “Collision behaviour
of a smaller particle into a larger stationary droplet,” Adv. Powder Technol., vol. 26, no. 1,
pp. 280–295, 2015.
[29] K. Mills, “The estimation of slag properties,” South. Afr. Pyrometallurgy, vol. 7, no. 3,
pp. 35–42, 2011.

65
Appendix A

Penetration depth as a function of the rate of the change of the momentum

The influence of air flow rate, nozzle size, and particle size on the penetration depth was studied
in the thesis. It was found that the smaller the nozzle diameter and the higher the air flow rate,
the larger the penetration depth. To explain these results, an axial rate of the change of the
momentum of the particle stream immediately after exiting the nozzle was formulated as

𝑚̇s (𝑉p2 − 𝑉p1 ) = mass flow rate × change of velocity (A1)

Figure A1 shows the control volume for the change of the momentum for the gas-particle flow.
The rate of the change of the momentum across the control volume is given by Eq. (A1), where
𝑉p1 is the velocity at which a particle enters the control volume (cf. Section 6.2) and 𝑉p2 equals
zero when the particle has penetrated the liquid and decelerated to fully stop. Newton’s second
law of motion expresses that this change of momentum per unit time will be caused by a force

𝐹 = 𝑚̇(𝑉p2 − 𝑉p1 ) (A2)

The force exerted by the solid stream on the liquid body touching the control volume is equal
and opposite to this force, and is here called the reaction force, so

𝑅 = −𝐹 (A3)

Vp
DN dp σoil
ρp
μoil
Control volume H0 𝑚̇𝑠 ρoil
Lp

μw
ρw
σw

Figure A1. Control volume for the calculation of the momentum change of the particle stream

66
Based on the above definition of the control volume and the reaction force, the penetration
depth as a function of the rate of the momentum change for the experiments can be studied, as
illustrated in Figure A2. Despite the scatter, it is seen that the penetration depth increases almost
linearly with the rate of momentum change, slightly levelling out at higher values of the
momentum.

100
y = -6332.6x2 + 1403.8x + 10.378
90 R² = 0.6945
80
particle pentration (mm)

70
60
50
40
30
20
10
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
rate of change of momentum (kg.m/s)

Figure A2. Penetration depth as a function of the rate of momentum change

67
Appendix B

Dimensional Analysis

The concept of conservation of dimensions will be applied to the dimensional analysis and
scale-up of engineering systems. In this section principles of dimensional analysis, as described
in reference [16], were used to interpret the laboratory experiments on the cold model of the
EAF system to dimensionless groups. This section is an introduction to a mathematical model
of the current physical model. Moreover, the main aim of this section is to define the model and
understand what the requirements of a good mathematical model are.
An appropriate set of dimensionless groups that can be used to represent the relationship
between the physical properties of the particles and the liquids, the dimensions of the nozzle,
and the driving forces for particle and fluid movements were defined. Figure B1 illustrates the
required parameters that can be considered in performing dimensional analysis.

Vp
DN dp σoil
ρp μoil
H0 𝑚̇𝑠
ρoil
Lp

μw
ρw
σw

Figure B1. Required parameters regarding performing dimensional analysis.

In order to perform the dimensional analysis, the problem should be defined, and the most
important fundamental variables should be determined. As is shown in Figure B1, the process
involves the injection of lime particles in the container containing water and oil as a simulation
of the molten phase and the slag in the EAF. The most significant variables are shown in the
figure. Table B1 shows the variables and their dimensions expressed in terms of dimensions for
length (L), mass (M) and time (t).

68
 Table B1. Fundamental variables of the process and their dimensions.

Variable Dimensions
dp L
ρp M L-3
Vp L t-1
Lp L
μoil M L-1 t-1
ρoil M L-3
σoil M t-2
μw M L-1t-1
ρw M L-3
σw M t-2
Dn L

According to the table, there are 11 - 3 = 8 dimensionless groups in this problem. In the next
step, choosing a set of reference variables could be helpful to derive the dimensionless numbers.
For choosing the reference values, some criteria must be satisfied including the number of
reference variables that must be equal to the minimum number of fundamental dimensions,
choosing reference values that do not have the same dimensions, and the dimensions that appear
in the problem variables must also be represented among the dimensions of the reference
variables. In addition, the reference variables should have the simplest combination of
dimensions [16].

Based on the above-mentioned criteria, Lp, Vp, and ρl related to the particles and the fluids were
selected as reference variables. As a result, dimensions and variables are defined as follows:

[Lp] = L (B1)
[ρl] = M L-3 (B2)
[Vp] = L t-1 (B3)

69
The above-mentioned dimensional equations were solved for the dimensions in terms of
reference variables.

L = [Lp] (B4)
t = [ Lp / Vp] (B5)
M = [ ρl Lp 3] (B6)

If the dimensional equations for other variables are written, the following equations were
obtained:

[dp] = L = [Lp] (B7)

[Dn] = L = [Lp] (B8)

[ρp] = M L-3 = [ρl] (B9)

𝜌𝑝 𝐿𝑝 3
[μoil] = M L-1 t-1 = [ 𝐿𝑝 ] = [ρl Lp 3 ]
𝐿p ( ⁄𝑉 )
p (B10)

𝜌p 𝐿p 3
[μw] = M L-1 t-1 = [ 𝐿p ] = [ρl 𝐿𝑝 3 ] (B11)
𝐿p ( ⁄𝑉 )
p

[σoil] = M t-2 = [ρl Vp 2 Lp] (B12)

[σw] = M t-2 = [ρl Vp 2 Lp] (B13)

In the final step the dimensionless numbers can be obtained as follows:

𝐿p
𝑁1 = (B14)
𝐷n
𝐿p
𝑁2 = (B15)
𝑑p

70
 𝜌oil
𝑁3 = (B16)
𝜌p
𝜌w (B17)
𝑁4 =
𝜌p

𝐿p 𝑉p 𝜌oil (B18)
𝑁5 =
𝜇oil

𝜌oil 𝑉p 2 𝐿p
𝑁6 = (B19)
𝜎go
𝐿p 𝑉p 𝜌𝑤
𝑁7 = (B20)
𝜇w
𝜌w 𝑉p 2 𝐿p
𝑁8 = (B21)
𝜎wo

where σgo is the surface tension in oil-air interface, and σwo is the surface tension in oil-water
interface.

The surface tension in the oil-water interface was considered as the average of the surface
tension of water-air and oil-air interface.
These dimensionless numbers can be used for defining the behavior of the system instead of
using the original variables. By using experimental data, some parameters involved in the
dimensionless groups have been measured and they will be utilized to analyze the behavior of
the system.

At this point, we can conjecture whether additional variables, such as gravity (g) or surface
tension (σ), should be included [16]. If gravity is going to be consider for particles, as a result,
𝑉p2
the Froude number can be added to these dimensionless groups: Fr = .
𝑔𝐿p

In this study, the penetration depth plays an important role in characterizing the efficiency of
the lime injection process. Therefore, the dimensionless penetration depth can be expressed as
a combination of N1 and N2 as

𝐿p 𝐿p 𝐿2p (B22)
𝐿∗p = ( ) ( ) =
𝐷n 𝑑p 𝐷n 𝑑p

71
The velocity of the particles, as discussed in Chapter 6, was measured before the particles hit
the air-liquid interface. The level of the oil in the model was set to 10 mm. As a result, the
penetration depth in oil related dimensionless numbers was considered to be 10 mm. Thus, by
using these assumptions and the dimensionless numbers, and linear multiple regression model
was developed. The dimensionless penetration depth was considered as the dependent variable
and the other dimensionless numbers as independent variables. Results of linear regression is
shown in Table B2.

Table B2. Results of linear regression

SUMMARY
OUTPUT
Regression Statistics
Multiple R 0.987
R Square 0.975
Adjusted R Square 0.937
Standard Error 225.4
Observations 36

ANOVA
Df SS MS F Significance F
Regression 6 59175995 9862666 232.8 3.610-23
Residual 30 1524607 50820
Total 36 60700602

Standard
Coefficients Error t Stat P-value Lower 95%
Intercept 308.0 661.8 0.4654 0.6449 -1044
N3N4 0 0 65535 #NUM! 0
N7 0.01723 0.002612 6.597 #NUM! 0.01190
N9 15.18 2.500 6.071 1.14410-6 10.07
N8 0.008765 0.011120 0.7881 0.4368 -0.01394
N6 -0.4342 0.1123 -3.866 0.0005509 -0.6636
N5 -12.66 5.5508 -2.281 0.02979 -24-00

According to the summary output, the coefficient of parameter N3×N4 is zero, since it has no
effect on the Lp*. As a result, it was omitted from the independent variables. In addition,
according to the p value which indicates the null hypothesis that can be rejected, p > 0.05:
consequently, it can also be omitted. Thus, by using a new set of independent variables, the
regression was performed that the results shown in Table B3 were obtained.

72
 Table B3. Results of second round of multiple regression.
Regression Statistics
Multiple R 0.9871
R Square 0.9744
Adjusted R Square 0.9711
Standard Error 224.0
Observations 36

ANOVA
Df SS MS F Significance F
Regression 4 59144425 14786106 294.5 3.510-24
Residual 31 1556176 50199
Total 35 60700602

Coefficients Standard Error t Stat P-value Lower 95%

Intercept 769.5 306.6 2.510 0.01752 144.1
N7 0.01875 0.001757 10.67 6.6710-12 0.01516
N9 15.12 2.484 6.088 9.56810-7 10.06
N6 -0.3597 0.06019 -5.976 1.31610-6 -0.4824
N5 -16.60 2.412 -6.882 1.02510-7 -21.52

In the next stage, after defining the normalized error as

measured value – calculated value 2

Normalized error = ( ) (B23)
measured value

a solver was applied to minimize the sum of the normalized error for all the experimental
results, yielding the linear relation

𝐿∗𝑝 = 273 − 4.72𝑁5 + 0.0073𝑁7 + 3.19𝑁9 − 0.098𝑁6 (B24)

It is worth mentioning that to obtain an accurate correlation more experiments must be

performed, for instance, with different liquids of different viscosities and densities, particles
with different physical properties, and wider range of operational conditions.

73
However, in different papers a variety of methods have been used for determining the
dimensionless numbers and the velocity of particles and deriving a correlation regarding
prediction of penetration depth. For instance, Okuno et al. [5] used the dimensionless numbers

𝜌𝑙 𝑁𝐿𝑝 𝑑𝑝′
Re𝑝 = (B25)
𝜇
𝜌𝑙 𝐿3𝑝 𝑁 2
𝑤𝑒 = (B26)
𝛾𝑆𝐿

𝐿𝑝
Fr = 𝑁√ (B27)
𝑔

where Rep is particle Reynolds number, We is the Weber number, Fr is the Froude number, 𝛾
is the interfacial tension, N is the number of particles per second, Lp is the penetration depth, 𝑑𝑝′
is the particle diameter, g is the gravitational acceleration and µ is the dynamic viscosity.

The final correlation, derived by multiple regression by using Eqs. (25-27) is

𝜌𝑙 − 𝜌𝑝 0.447
Re𝑝 = 187Fr 0.673 We0.199 ( ) (B28)
𝜌𝑙

Kimura [6]obtained an empirical equation of particle penetration depth as

1.266 0.135 −0.609 0.12 0.494 0.182

𝐻𝑝 = 0.136 𝑑𝑃𝐾 𝑈𝑆𝐾 𝜌𝐼𝐾 𝜌𝑆𝐾 𝑁𝑃𝐾 𝑈𝐶𝐾 (𝑑𝑂𝐾 + 2𝐻𝑂𝐾 tan 𝜃)0.475 (B29)

𝑁𝑃𝐾 = 𝜋
𝑊𝑆𝐾 (B30)
3
𝑑 𝜌 𝑆 10(𝐴−2)
6 𝑃𝐾 𝑆𝐾 𝑂𝐾

𝐻𝑂𝐾 2 𝐻𝑂𝐾 𝑈𝐶𝐾 𝑊𝑆𝐾 (B31)

𝐴 = 1.091 − 0.999 (log ( )) + 1.46231 log ( ) − 0.0843 log ( ) + 0.253( )
𝑑𝑂𝐾 𝑑𝑂𝐾 100 cm/s 𝑆𝑂𝐾

where Hp is the particle penetration depth including cavity depth, dPK is the particle diameter,
USK is the particle velocity at the nozzle exit, NPK is the number of particles per unit of jet
volume (cf. Eq. (B29)), UCK is the axial gas velocity, dOK is the inner diameter of the lance,

74
HOK is the lance height, Θ is the gas jet cone angle (= 4.6°), WSK is the particle feeding rate
(expressed in g/(cm3s)), SOK is the nozzle cross-sectional area, and A is defined by Eq. (B31).

The above-mentioned correlations are based on the top bellowing injection systems in BOF and
agreed well with the experimental results. This can be a good introduction for future works and
would clarify the type of measurements, experiments, and methods that can be used to develop
an appropriate correlation for predicting particle behaviors.

75
Appendix C

Physical properties of slag

This appendix is a review on estimation of different physical properties of slag. Due to the
importance of producing a good slag in the steelmaking processes, and its effects on the
injection process, this information is provided here. Several models have been developed based
on the type and composition of the slag, and only some of these are shown in this appendix.

1 Different types of slags with a wide range of composition

Based on the variety of processes, in the first place, different types of slag are introduced as
follows [29]:

1. Metallurgical slags are usually based on the CaO-Al2O3- SiO2 system.

2. Non- ferrous slags are frequently based on fayalite, for example, FeO.SiO2.
3. Coal slags/ fly ashes are basically CaO-Al2O3-SiO2 with relatively low CaO levels
(<10%) and tend to be very viscous.
4. ESR slags are composed of oxides and fluorides (in significant amounts), for example,
CaF2-CaO-Al2O3.
5. Continuous casting fluxes and welding fluxes are made up of oxides with < 10% CaF2.
Some types of slags have similarities to glasses and slags.
6. Glasses which tend to have high SiO2 contents.
7. Enamels which contain appreciable amounts of B2O3.

Liquid slags mostly contain ions. It is known that ions are involved in reactions between metal
and slag. Equations (C1) and (C2) represent the reactions between ions [29].

2 Al +1.5 SiO2 (slag) → 1.5 Si + Al2O3 (slag) (C1)

2 Al +1.5 Si4+ → 1.5 Si + 2 Al3+ (C2)

2 Prediction of Slag properties

Slag properties are important since they have a decisive effect on both process control and on
the quality of the product. Moreover, good quality property data are needed to develop reliable
mathematical models of the process and to optimize the efficiency of the process and quality

76
control [29]. For predicting the properties of slags, several types of mathematical models exist
as follows [29]:

1. Thermodynamic models for establishing the viability of the process and the optimum
conditions.
2. Kinetic models for establishing the productivity rate.
3. Heat and fluid flow models for improving process design and control.
4. Models regarding environmental and financial impact of the process.

It is worth mentioning that reliable property values are principally needed as input data for heat
and fluid flow models. These models often involve simulation of the fluid flow and heat
transfer in the process, and it has been shown that reliable predictions require accurate data for
properties such as viscosity, density, and thermal conductivity of the slag. In addition, such
measurements are often tricky due to the high temperatures of the process and the reactions
between slag and container [29].

3 Determining physical properties by modeling:

Several types of models have been reported for calculating the properties of slags; these are
classified into five groups as follows [29]:
1. Conventional regression models
2. Neural network models
3. Partial molar models
4. Models based on structural parameters
5. Thermodynamic models (which take structure into account)
6. Molecular Dynamics (MD) models.

3.1 Density
The density derived directly in thermodynamic models, and it can be determined to a reasonable
accuracy (± 2 – 3%). The available methods for modelling the density of molten and solid slag
are summarized in Table C1 [29].

77
 Table C1. Details of models to calculate densities of alloys and slags [29].

3.2 Viscosity
The fluid flow of the molten slag is important since it can affect the kinetics of the refining
reactions, where mass transfer in the slag phase is the rate controlling factor, or where it
provides lubrications of the metal, for instance, in continuous casting of steel. It naturally also
affects the transportation and removal of the slag in the refining process, and furthermore also
how the slag deposits on the walls of metallurgical vessels. Therefore, from the viewpoint of
process control it is advantageous to have a universally applicable method to deduce the
changes in slag viscosity due to changes in, for instance, temperature and composition. The
measurement of viscosity is time consuming, difficult and require considerable expertise.
Consequently, a model for the rapid calculation of viscosity based on the chemical composition
and temperature is an industrial requirement. Therefore, several models have been produced in
response to this need. These models were developed based on a numerical fit of viscosity to
chemical composition. However, the estimated viscosities differ from measured values by > ±
30%, which shows the large range of slag viscosities (<10-1 – 10+2 Pa·s) [21].
The model referred to as the NPL model, relates the viscosity (η) of slags to the structure
through the optical basicity corrected for the cations used for charge balancing (Ʌcorr). The
viscosity of the slags depends on temperature, so minimizing the effect of temperature is
important when modeling the compositional effect. The assumptions for modeling accounted
for the temperature dependance are as follows [21]:

78
 1. Arrhenius behavior: ln η = ln A + B/T, where T is the thermodynamic temperature and
A and B are constants with respect to temperature.
2. A and B are functions of the composition through Ʌcorr.
Table C2 shows the composition of the slag, and Table C3 represents the viscosity-temperature
relations of the industrial slags.

Table C2. Composition of slags [21].

79
 Table C3. Viscosity-temperature data for industrial slags shown in Table C2 [21].

3.3 Surface tension

Surface tension (σ) is not a bulk property. Surface-active components such as sulfur or oxygen
in molten metals have a low surface tension compared with the metal, and consequently tend to
occupy the surface layer in much higher concentrations than the exist bulk. Therefore, surface
tension and its temperature dependance (d𝜎⁄d𝑇) are extremely sensitive to sulfur or oxygen
concentrations: for instance, 50 ppm of sulfur causes a decrease of 25% in surface tension
yielding a local minimum [29].
However, compared to the oxygen and sulfur in metals, in the slag the difference between the
surface tensions of the surface-active components, such as B2O3, CaF2, and SiO2, and those of
the other components, for instance, CaO, is minor. Surface tension of slags does not tend to be
very structure dependent. The surface tension of the slag depends on the concentration of the
surface-active components, and the chemical activity of the surface-active components, which
determines the surface-active concentration in the surface layer. As a result, models of surface
tension tend to be based on the thermodynamics of the slag system [29].

80
Appendix D

Link to the YouTube channel and attachment

In order to give the readers better insight into the experiments, high-speed camera videos related
to the figures used in this thesis have been uploaded on a YouTube channel. The link to the
channels is https://www.youtube.com/@hoorakhshbayat8159/videos

Required information regarding operational conditions has been also provided on the channel
and below each video.

An excel file which includes the calculations regarding the design of a solid pneumatic
conveying system, multiple regression, and the results of the experiments is also available under
the title of Lime Injection Master’s Thesis.

81