IN DEGREE PROJECT TECHNOLOGY,FIRST CYCLE, 15 CREDITS

, STOCKHOLM SWEDEN 2016

The viability of oxygen gas blowing as a foaming slag suppression system for slopping prevention in BOF-processes

TEODOR HAGLUND

JOAR HUSS

KTH ROYAL INSTITUTE OF TECHNOLOGYSCHOOL OF INDUSTRIAL ENGINEERING AND MANAGEMENT

Abstract Slopping in BOS-processes poses many problems, most significantly to work environment and process effectiveness. Due to the current weaknesses in slopping prediction systems a foaming slag suppression system with immediate effect is needed to prevent slopping. This project aims primarily to be a proof of concept for pressurized oxygen gas blowing as a mean of foam steady-state height suppression to prevent slopping and secondarily to appreciate the viability of this concept economically. Five nozzles were designed and used to blow pressurized air onto foam made of silicon oil, at different air flows. It was then determined which nozzle was the most effective by comparing height difference with airflow. The airflow was compared to a life scale scenario to determine the real flow rate which was used to determine economic viability. Results show that a nozzle with a circular small hole is the most effective nozzle requiring 20 [ln min-1] to reduce the foam height by 69.6%. The real flow to achieve this would be 0.605 [m3 s-1], however due to the cold models limitations this is not the true value. The slopping suppression technique shows promise as a concept both economically and practically.

Sammanfattning

Överkok i BOF-processen skapar många problem, mest i arbetsmiljön och minskar så väl produktionens effektivitet som takt. Eftersom de kontrollsystem som finns att tillgå idag har vissa begränsningar så behövs det ett system för att motverka skumtillväxt med direkt inverkan för att hindra överkokning. Det här projektet ämnar huvudsakligen till att bevisa att ett pålagt flöde av syrgas kan trycka ned det skummande slaggets höjd och på så sätt förhindra överkokning och sekundärt till att bedöma konceptets ekonomiska rimlighet.Fem munstycken designades och användes till att blåsa tryckluft, med olika flöden, på skum bestående av silikonolja. Effektiviteten hos munstyckena utvärderades genom att jämföra höjdskillnaden mot det pålagda luftflödet. Luftflödet jämnfördes sedan mot ett scenario i industriell skala och det verkliga luftflödet kunde därefter beräknas. Med detta som bakgrund gjordes en ekonomisk analys. Resultat visar att munstycket med ett litet cirkulärt håll är mest effektivt då det krävdes ett flöde på 20 [ln min-1] för att reducera skumhöjden med 69,6%. Det verkliga flödet beräknades till 0,605 [m3 s-1], men eftersom den kalla modellen har vissa begränsningar så är detta värde inte sant. Den här tekniken för att förhindra överkokning ser lovande ut både ur ett ekonomiskt men också ur ett praktiskt perspektiv.

Table of Content 1. Introduction .................................................................... 1

1.1 Background .............................................................. 1 1.1.1 The LD-process ................................................. 1

1.2 Foaming .................................................................... 1 1.2.1 Foaming index ................................................... 2 1.2.2 Process gases ...................................................... 2

1.3 The Slopping phenomenon ...................................... 2 1.3.1 Oxygen blowing ................................................. 2 1.3.2 Slopping causes .................................................. 3

1.4 Slopping prediction .................................................. 3 1.4.1 Modelling ........................................................... 3 1.4.2 Vibration measurements .................................. 3 1.4.3 Audiometry ........................................................ 4 1.4.4 Radio wave interferometry............................... 4 1.4.5 Vessel vibration ................................................. 4

1.5 Slopping prevention ................................................. 4 1.5.1 Static slopping control ...................................... 4 1.5.2 Dynamic slopping control ................................. 5 1.5.2.1 Oxygen lance, -flow and bath agitation ........ 5 1.5.2.2 Foam suppressing soundwaves ..................... 5 1.5.2.3 Additives ......................................................... 5

1.6 Turbulent jets .......................................................... 5 2.Aim of work ..................................................................... 5 3. Experimental section ..................................................... 6

3.1 Experimental preparation ....................................... 6 3.2 Experimental Setup ................................................. 6

3.3 Measurement method .............................................. 7 4. Results ............................................................................. 8 5. Discussion ...................................................................... 11

5.1 Impact of nozzle geometry ..................................... 11 5.1.1 Jet area ............................................................. 11 5.1.2 Kinetic energy vs momentum ......................... 11 5.1.3 Converging behavior ....................................... 11

5.2 Oxygen as blowing gas ........................................... 12 5.3 Foam analysis ......................................................... 12 5.4 Kinetic energy vs momentum assessment by argon gas flow variation ......................................... 12 5.5 Jet momentum number and penetration depth ... 13 5.6 Model basis in reality ............................................. 13 5.7 Economic viability .................................................. 13 5.8 Sustainability .......................................................... 14

6. Conclusion .................................................................... 14 7. Future work .................................................................. 15 8. Summary ....................................................................... 159. Acknowledgements ....................................................... 16

10. References .................................................................. 17

To our fellow students

Fear no more the heat o' the sun; Nor the furious winter's rages,

Thou thy worldly task hast done, Home art gone, and ta'en thy wages;

Golden lads and girls all must, As chimney sweepers come to dust.

Fear no more the frown of the great, Thou art past the tyrant's stroke: Care no more to clothe and eat; To thee the reed is as the oak:

The sceptre, learning, physic, must All follow this, and come to dust.

Fear no more the lightning-flash, Nor the all-dread thunder-stone; Fear not slander, censure rash;

Thou hast finished joy and moan; All lovers young, all lovers must

Consign to thee, and come to dust. No exorciser harm thee!

Nor no witchcraft charm thee! Ghost unlaid forbear thee!

Nothing I’ll come near thee! Quiet consummation have;

And renowned be thy grave!

“Fear no more”- William Shakespeare

1

1. Introduction

1.1 Background

1.1.1 The LD-process

One of the world’s most common steelmaking processes is the Linz-Donawitz process (LD). By using 99.5% pure oxygen blown through a lance to decarburize the steel bath, and flux to create the necessary slag for metal-slag interactions that are vital for purification processes, pig iron is converted into crude steel. The basics are outlined in figure 1 [1]. As the decarburization is initiated the redox-reaction of iron and carbon take place releasing CO/CO2 gas which is trapped inside the slag turning the liquid slag into a foamy slag [1]. If too much foam is generated the batch spills over as slag and metal is forced out through the top opening, a phenomenon called slopping, leading to losses in yield, production time, reduced equipment lifespan and to pollution of the environment [2]. This problem is yet to be solved, as slopping behavior is very complex, even unpredictable, and there is a lack of accuracy in measurement methods during blowing [3]. In a world that strives towards sustainability and efficiency, new slopping preventing techniques are needed; one idea to such a technique is slopping suppression by oxygen gas blowing. However as this is an entirely new technique its validity must first be tested.

Figure 1: Basic principles of the Linz-Donawitz process [4].

1.2 Foaming

Foam consists of many polyhedral bubbles that are separated by thin films of liquid. When three films meet it is called a plateau border channel. The liquid is driven from the film to these plateau channels and flows to the bulk of the liquid due to gravity, a phenomenon called film drainage. Film rupture is when the films become too thin and rupture due to mechanical and/or thermal fluctuation [5]. In the LD process multi-phased foam is formed consisting of liquid slag, metal droplets, solid particles, and process gases e.g. CO. There are four main reasons why this is the case [4].

• As the oxygen jet hits the iron melt itpushes metal droplets into the upper partsof the vessel.

• Added flux is not dissolved directly andexists as suspended solid particles in thefoam.

• As process gas forms it is trapped in theliquid slag inflating it and forming aviscous emulsion.

• Some process gas form within the foamitself as the dissolved carbon reacts withthe iron oxide inflating it further.

An equilibrium between foam growth and rupture, by escaping gas, is often reached although a number of factors displace this equilibrium and so slopping occur. The mathematical description is shown in eq. (1).

𝑑𝑑ℎ𝑑𝑑𝑑𝑑

= 𝑘𝑘1𝑄𝑄𝑔𝑔 − 𝑘𝑘2ℎ eq. (1) [6]

Where 𝑑𝑑ℎ𝑑𝑑𝑑𝑑

is the change in foam height, Qg is the gas flow [m3 s-1], h is the foam height [m], k1 is a formation constant and k2 is a rupture constant. Steady-state foam height is when 𝑑𝑑ℎ

𝑑𝑑𝑑𝑑= 0.

2

1.2.1 Foaming index

There are a lot of theories and models on foam formation, many of these are based on a foam index Σ, quantifying the foaming behavior [5] [7] [8]. Σ is defined as the ratio between foam volume and the gas flow, as is shown in the following expressions eq. (2):

Ʃ = 𝛥𝛥ℎ𝛥𝛥𝑉𝑉𝑔𝑔𝑠𝑠

where 𝛥𝛥𝑉𝑉𝑔𝑔𝑠𝑠 = 𝑄𝑄𝑔𝑔𝐴𝐴

eq. (2) [6]

Where Δh is the change in foam height [cm], Qg is the gas flow [cm3 s-1] and A the cross section area of the vessel [m2]. Note that the unit for the foam index is time; this can be interpreted as the time it takes for the process gases to pass through the foam [6]. Therefore the foam height is directly proportional to the foam index. Earlier studies from literature show that the foam index is influenced by: apparent viscosity (μ) of the slag, surface tension (σ), density (ρ) and bubble diameter (db). The relationship between the foam index and these variables have yet to be verified but has been suggested for certain cases to be as is shown in eq. (3) and (4): According to Lahiri and Pal [5].

Σ = 115 ∗ µ1.2

σ0.2ρ𝑑𝑑𝑏𝑏0.9 eq. (3)

According to Jiang & Fruehan [9]: Σ = 115 ∗ µ

�ρσ eq. (4)

A higher apparent viscosity leads to a higher foam index and an increase in foam height follows. The major reason for slopping in today’s industry is slopping induced by over-viscous slags [1]. This is due to the prolonged residence time of the metal droplets in the slag allowing more droplets to be in emulsion simultaneously and more gas is produced thereafter [10].

1.2.2 Process gases

As mentioned earlier, the amount of gas generated effects the growth of the foam. Process gas is a byproduct of decarburization, which consists of the following gases and reactions, eq. (5)-(8) [4]:

• Direct oxidation at the liquid metalsurface: C + 1

2O2 (g) => CO (g) eq. (5)

• Oxidation in the foam, between metaldroplets and iron oxide:C + FeO => CO (g) + Fe eq. (6)where FeO is the product ofFe + 1

2O2 => FeO eq. (7)

• Reaction between dissolved oxygen andcarbon in the melt:C + O => CO (g) eq. (8)

1.3 The Slopping phenomenon

1.3.1 Oxygen blowing

When blowing oxygen in BOF, it can prove difficult to get a good balance between oxygen in the slag and the melt. Generally there are two kinds of blowing that deviates from a balanced blow [4]: Hard blowing meaning there is a hard impact from the jet on the surface since the lance height is low, which will result in an under oxidized slag because eq. (5) and eq. (8) will be dominant. Soft blowing is instead when the height of the lance is high therefore having a softer impact, resulting in an over-oxidized slag following eq. (7), as well as a slower de-carburization of the melt. The optimum height of the lance is at which the oxygen added to the slag balances the consumption of iron oxide, when the metal droplets are ejected into the foam and decarburized. Every steel mill has its variations and so does the preferred height of the oxygen lance for the mill, it is therefore hard to generalize an optimum lance height [4].

Soft blowing has been compared to a time-bomb, due to “hyper-reactive conditions” that builds up as the oxygen does not react with carbon, even the smallest of changes in converter environment can trigger a dramatic gas generation which leads to extreme foam growth which pours over the lips of the ladle resulting in slopping [4]. In order to avoid slopping one has to have control over the slags composition, as can be seen in figure 2, the distance

3

between under-oxidized slag and over-oxidized slag is a narrow corridor. As a slag becomes under-oxidized the apparent viscosity increases leading to a “dry” viscous foam. This is due to a low amount of iron oxide at the start of the main decarburization period. However, if over-oxidized, an overproduction of process gas is generated increasing the gas velocity within the foam [4]. Today slopping registration is mostly based on images taken by cameras situated against the bottom of the vessel or the vessel mouth [3] [11]

[12].

Figure 2: Diagram connecting the state of oxidation of the slag with slopping. The black dots indicate the best slag trajectory in order to avoid slopping [10].

1.3.2 Slopping causes

Mainly there are two different groups of slopping causes, static and dynamic. Static factors are such factors that can be influenced before the process, these include: vessel design, charge quality as well as the blowing scheme controlling the lances positioning, time for adding materials and oxygen flow. Dynamic factors are instead factors that take place during blowing, such as deviating from the blow scheme and bottom stirring [4]. According to Shakirov et.al slopping can be divided into three different types: dry, volcano and common. Dry and volcano are the most dangerous although avoidable by having good control over the quality of the charge materials as well as using consistent blowing schemes [2].

1.4 Slopping prediction

When the blow has started one can initiate dynamic slopping control measures, however one must first be able to predict if the batch will slopp, this can be done two ways, either by measuring the foam level or by monitoring secondary information that has a connection with foam growth. There are several techniques to do this; some of these are described below, however during blowing some methods will not be accurate during certain periods due to factors not related to slopping. Therefore it is best to combine several methods for the most secure data [4].

1.4.1 Modelling

Today most industries have a dynamic process control system, which calculates temperature while running chemical composition analysis of the slag, melt and gases. By combining this with thermodynamics and kinetics models one can calculate the state and composition of the foaming slag and forecast the risk of slopping [2] [4] [12].

1.4.2 Vibration measurements

A method used by many steel plants is estimating the foam height by lance vibrations. During the blow kinetic energy will be transferred from the foam to the lance resulting in vibrations spreading out through the lance structure. By using an accelerometer these vibrations can be measured, the amplitude of the vibration will increase proportional to the contact area and certain patterns of this amplitude foretell an imminent slopping. This method will not work the first few minutes of blowing as the foam will have to rise before it covers the lance or if there is a sturdy centralized process gas flowing up around the lance. Another issue is if there is metal solidifying on the lance ,sculling, as this will change the behavior of the vibrations, therefore the frequency band will have to be confirmed when changing the lance [4] [13].

4

1.4.3 Audiometry

Another common method is audiometry, which is done by placing microphones by either the mouth or on the lower stack of the vessel, these record the sounds from the oxygen jet. The emulsion of slag works as a dampener for sound waves originating from the jet. The noise level will decrease as the foam level increases. In 2010 this was regarded as the most accurate method for height estimation, as it, like lance vibration, only reacts to the height of the foam. Despite this it has its drawbacks, the method requires the lance nozzle to be submerged in the foam and the foam needs to cover the nozzle horizontally, the instruments are also dust sensitive, not optimal for an industrial environment [4] [14].

1.4.4 Radio wave interferometry

By the use of radio wave interferometry (RWI) one can directly measure the height of the separate layers in the LD-converter. As microwaves encounter an interface between two phases a portion of them will reflect and another portion will refract and travel through the material. If these waves are sent in directly from above down toward the bath reflections will be coming from both the gas-slag interface and the slag-metal interface these will have different properties and by precise treatment of these signals the height of the interfaces can be measured. However as the foam has a great number of interfaces between gas-liquid, gas-solid and liquid-solid the radio waves have an uncertainty to them. There are also some practical problems as the RWI unit needs to be placed close to the vessel mouth in order to work, this will often lead to a damaged unit as it is exposed to molten material ejected up through the vessel mouth [4] [11] [14].

1.4.5 Vessel vibration

By placing an accelerometer on the trunnion vibrations of the vessel can be measured. The vibrations have a direct correlation with the foam height and can therefore be used for estimations. The vibration phenomenon is shown in figure 3 along with the soundwaves used in the audiometry method. However certain process disturbances may reduce the accuracy of this method [4].

1.5 Slopping prevention

The reasons for slopping are well documented and mapped, however slopping is not easily avoidable since the steel mills steelmaking setup probably already is in place and installation of additional instruments would cost time and money or simply cannot be done.

1.5.1 Static slopping control

Static slopping control includes all preventive measures that can be taken before the blow, these are among others: Defining the vessels volume -most of today's vessels are overloaded due tohigh demands on productivity, ensuring thequality of the hot metal, the fluxes and the scrap,choosing the right coolants, as well as predictingslag properties. It is also of utmost importance tohave a blowing scheme that minimizes the risk ofslopping, this includes: the lance profile, theoxygen profile and the flux profile [4] [15].

Figure 3: The simplified principle of vessel vibration.

5

1.5.2 Dynamic slopping control

If slopping is detected countermeasures for slag suppression can be initiated. The parameters that can be altered are: the oxygen lance position, oxygen flow rate, bath agitation and material addition [3] [16] [13].

1.5.2.1 Oxygen lance, -flow and bath agitation

Once slopping is detected the most common thing to do if the slag is over-oxidized is to move the lance closer to the metal bath. As a result the oxygen content in the foam will be lowered; this is often combined with a lower oxygen flow rate and additional bottom stirring –bath agitation. If the oxygen flow is to be reduced without first lowering the lance it would result in a soft blow which would just worsen the situation. However if the foam is under-oxidized the oxygen flow should be lowered while the lance should not, it can in some cases be correct to instead elevate the lance. Bath agitation only has a limited effect and to achieve this minimal effect the gas flow for bath agitation has to be significantly increased [4][10].

1.5.2.2 Foam suppressing soundwaves

A newer method under development is acoustic defoaming where the foam is suppressed by the use of sound waves. Studies have given the following conclusions: the defoaming effect of sound waves is dependent on the sound frequency, intensity and the viscosity of the slag. Lower frequency waves is more effective, certain resonance frequencies amplifies defoaming and there is a threshold value, dependent on viscosity and frequency, at which the steady state foam height abruptly decreases [17].

1.5.2.3 Additives By adding material, depending on circumstance, it is possible to suppress slopping. If the foam is under-oxidized fluorspar (CaF2) or alumina (Al2O3) are good choices, however the steel maker has to be sure not to add too much as to endanger the composition of the batch. In the case of an

over-oxidized foam additions that create channels, through which trapped gases can escape, or additions that simply reduce the volume of the slag are suitable. For the latter coke or coal are ideal [4] [16] [13].

1.6 Turbulent jets

When a moving fluid encounters another body of a resting fluid a shear of velocity is created between them leading to turbulence and mixing, this is most notably seen when observing jets. Experiments have shown that when a fluid penetrates into such a resting body of the same density it will adopt a near conical shape when exiting a circular nozzle. The radius of this jet cone is proportional to the distance that the jet has been traveling since exiting its discharge location. Studies indicate that the angel of travel is 11.8°, giving the correlation between jet radius and distance traveled shown in eq. (9) [23]. 𝑅𝑅(𝑥𝑥) = tan(11.8°) ∗ 𝑥𝑥 eq. (9) Where: R is the radius of the cone and x the distance from the virtual source. The virtual source being inside the nozzle, a distance 5𝑑𝑑

2 from

the opening where d is the diameter of the circular exit hole [18].

2. Aim of work

A lot of work has been put into estimating the height of the foam, the causes for slopping and foam control in BOF systems. However slopping is still a major problem for the steel industry. Systems such as audiometry and lance vibration measurement exist but have certain flaws, are not applicable and/or do not give a good enough picture of the slopping phenomenon to allow proper slopping prevention actions to be taken. Therefore there is a need for a reliable foam control system. This study aims to be a proof of concept that slopping can be prevented using mechanical force in the form of pressurized gas blowing and to map some of the parameters that effect the suppression of foam as well as assessing the economic viability of this technique.

6

3. Experimental section

3.1 Experimental preparation

A cold model was constructed to simulate the slag with a Plexiglas cylinder and 3D-printed parts, which were added for extra functionality. A nozzle height control device with bib functionality was made to ensure that all nozzles were at the same starting height and to prevent silicon oil spillage in case of slopping which can be seen in figure 4. It was modelled using “COMSOL” with a 22 [mm] diameter circular hole, to which all nozzles were designed to fit snuggly into. A total of five nozzles were modelled in “Solid Edge” and their respective design can be seen in Figure 5-9. As shown in figure 5 nozzel 1 was designed with an Ø6 [mm] circular outlet. Nozzel 2 had a smaller circular outlet of Ø2 [mm] shown in figure 6. Nozzle 3 was designed to mimic the look of an oxygen lance with six Ø1.27 [mm] circular outlets as can be seen in figure 7. Square outlets were integrated onto nozzle 4 where three square outlets were placed parallel to each other, the one in the middle having an area of 1 [mm] times 6.8 [mm] and the two surrounding with the dimensions 1.3 [mm] and 4 [mm] displayed in figure 8. The appearance of nozzles 3 was redesigned for nozzle 5, as can be seen in figure 9, to have four Ø4 [mm] circular outlets. The height control device, as well as all nozzles, was 3D-printed using the software “Cura” and an Ultimaker2Extended 3D printer with Poly-lactic-acid (PLA) as printing material and solid as print setting.

Figure 4: Nozzle height control device with a Ø22 [mm] circular hole for nozzle placement and a bib to prevent oil spillage.

Figure 5 (ttl) and 6 (ttr): Screenshots of nozzle 1 to the left with a Ø 6[mm] circular outlet and nozzle 2 to the right with a Ø2 [mm] circular outlet.

Figure 7 (ttl) and 8 (ttr): Screenshots of nozzle 3 to the left designed as an oxygen lance and nozzle 4 to the right with three rectangular outlets of different dimensions.

Figure 9: Modified oxygen lance design with four circular outlets placed sector symmetrically with a diameter of 4 [mm].

3.2 Experimental Setup

An argon gas tank was outfitted with an AGA pressure valve which was set to 2 [bar] connected by a tube to a Bronkhorst flow bus model F-201CV-1K0-AAD-33-V, calibrated for 1 [ln min-1] Ar, 2 [bar] at 293 [K]. The flow bus was in turn connected to an argon flow control device “Bronkhorst High-Tech B.V.E-7000”. The Plexiglas cylinder, with a height of 280 [mm] and 93 [mm] in inner diameter, was outfitted with a gas outlet in the bottom through which argon gas could flow and a silica membrane was placed

7

10 [mm] from the bottom of the cylinder, with a porosity of 10–16 [μm]. The nozzle height control device was placed on top of the Plexiglas cylinder and a ruler was positioned parallel to the cylinder. The nozzles were then placed into the circular hole of the height control device and connected through a tube to a Bronkhorst flow bus which was controlled using a computer. This flow bus was in turn connected to a pressure valve set to 2 [bar]. The pressure valve was then connected to a pressurized air source. A schematic sketch of the setup can be seen in figure 10.

Figure 10: Schematic drawing of the experimental setup

3.3 Measurement method

250 [ml] of silicon oil was poured into the Plexiglas cylinder and an argon flow of 0.5 [ln min-1] was turned on inflating the silicon oil. As the height of the foam pillar had stabilized the height was recorded from the bottom of the cylinder by adding more oil and a nozzle was set in place into the height control device, with the air flow bus set to 5 [ln min-1]. A reduction in foam height was noticed immediately, although a waiting time of three minutes was given for the surface to fully stabilize. The new foam height was measured using a ruler parallel to the cylinder and another ruler held perpendicular to determine the height this is shown in figure 11. The flow of air was then gradually increased from 5 [ln min-1] with a 1 [ln min-1] interval up to 10 [ln min-1], then from 10 [ln min-1] up to 40 [ln min-1] with a 2 [ln min-1] interval with measurements taken at

every interval. As the surface became unstable and started to splash the experiment was discontinued. The impact on the height of the foam pillar by argon flow was tested by gradually increasing argon flow from 0.5 [ln min-1] to 1.5 [ln min-1] with a 0.1 [ln min-1] interval. A constant airflow of 5 [ln min-1] was used for each nozzle and measurements were taken in the same fashion as above. A total of 270 measurements were made.

Figure 11: Schematic drawing of the measurement method

8

4. ResultsTo determine the effectiveness of the five nozzles the ratio between reduction in foam height, reduction in foam height will furthermore be referred to as ΔL, as well as the ratio of exit velocity and ΔL were investigated.Figure 12 depicts how the difference in foam height varies with air flow and nozzle type for the silicon oil with a viscosity of 0.1 [Pa s] and an argon gas flow of 0.5 [ln min-1]. It can be observed that nozzle 2, with a Ø2 [mm] circular hole, has the overall highest ΔL to flowrate ratio of all nozzles having reduced the foam pillar with 71 [mm] at 20 [ln min-1] giving it a ratio of 3.55 [min m-2]. The relative maximum ΔL for nozzle 2 is 69.6% compared to the other nozzles in falling order: nozzle 3 61.8%, nozzle 4 56.9%, nozzle 1 52.9%, nozzle 5 38.2%. An asymptotic con-verging behavior is observed for all nozzles at different values. Similarly figure 13 presents ΔL as a function of exit velocity for the same setup. It can be seen that the ΔL to exit air velocity ratio is highest for nozzle 1, with a 6 [mm] circular hole, having reduced the foam with 61 [mm] at 17.7 [m s-1] a ratio of 3.45e-3 [s]. Resembling behaviors are observed in figure 14 and 15 where silicon oil with a dynamic viscosity of 0.2 [Pa s] was investigated with the same setup as in figure 12 and 13. Nozzle 2 again exhibits the highest ΔL to [ln min-1] ratio, reaching a maximum ΔL of 68 [mm] a relative reduction of 66.7 % for a flow of 9 [ln min-1] giving it a ratio of 7.56 [min m-2] and nozzle 1 the highest ΔL to [m s-1] ratio.

Figure 12: Nozzles 1-5 air flow and ΔL for silicon oil with a dynamic viscosity of 0.1 [Pa s].

Figure 13: Nozzles 1-5 exit velocity and ΔL for silicon oil with a dynamic viscosity of 0.1 [Pa s].

Figure 14: Nozzles 1-5 air flow and ΔL for silicon oil with a dynamic viscosity of 0.2 [Pa s].

Figure 15: Nozzles 1-5 exit velocity and ΔL for silicon oil with a dynamic viscosity of 0.2 [Pa s].

0

20

40

60

80

5 15 25 35

ΔL [m

m]

Air flow [ln min-1]

Nozzle 1 Nozzle 2 Nozzle 3Nozzle 4 Nozzle 5

40

50

60

70

0 20 40 60

ΔL [m

m]

Velocity [m s-1]

Nozzle 1 Nozzle 2 Nozzle 3Nozzle 4 Nozzle 5

0

20

40

60

5 10 15 20

ΔL [m

m]

Air flow [ln min-1]

Nozzle 1 Nozzle 2 Nozzle 3Nozzle 4 Nozzle 5

30

40

50

60

70

0 10 20 30 40

ΔL [m

m]

Velocity [m s-1]

Nozzle 1 Nozzle 2 Nozzle 3Nozzle 4 Nozzle 5

9

A comparison between 0.1 [Pa s] and 0.2 [Pa s] is seen in figure 16 where a generally higher ΔL to air flow ratio is observed for the silicon oil with a dynamic viscosity of 0.2 [Pa s], although nozzle 2 has a higher absolute value of ΔL of 67 [mm] at 14 [ln min-1] for 0.1 [Pa s] compared to 64 [mm] at 14 [ln min-1] for 0.2 [Pa s]. A converging behavior for all nozzles at both dynamic viscosities is perceived. The influence of the argon flow was also tested by varying the argon gas flow and keeping the air flow constant at 5 [ln min-1] for 0.1 [Pa s] and 0.2 [Pa s]. An increase in argon flow resulted in bigger bubbles and, as can be seen in figure 17 and 18, ΔL is proportional to the flow of argon gas. A higher dynamic viscosity of the silicon oil led to a greater ΔL as can be seen by comparison of figure 17 and 18 as well as in figure 19, where ΔL is plotted against argon flow for nozzle 2 comparing the two dynamic viscosities. A total of 270 measurements were taken.

Figure 16: Nozzles 1-5 air flow and ΔL for both dynamic viscosities, where red represents a dynamic viscosity of 0.1 [Pa s] and blue 0.2 [Pa s].

Figure 17: Nozzles 1-5 argon flow and ΔL for silicon oil with a dynamic viscosity of 0.1 [Pa s].

Figure 18: Nozzles 1-5 argon flow and ΔL for silicon oil with a dynamic viscosity of 0.2 [Pa s].

Figure 19: Nozzles 2 argo flow and ΔL for both dynamic viscosities. Red represents 0.1[Pa s] and blue 0.2 [Pa s].

0

10

20

30

40

50

60

70

5 15 25

ΔL [m

m]

Air flow [ln min]

Nozzle 1-5(0.1 Pa s)

Nozzle 1-5(0.2 Pa s)

0

20

40

60

80

0,5 1 1,5

ΔL [m

m]

Ar flow [ln min-1]

Nozzle 1 Nozzle 2 Nozzle 3Nozzle 4 Nozzle 5

01020304050607080

0,5 0,7 0,9 1,1 1,3 1,5

ΔL [m

m]

Ar flow [ln min-1]

Nozzle 1 Nozzle 2 Nozzle 3Nozzle 4 Nozzle 5

45

55

65

75

0,5 0,7 0,9 1,1 1,3 1,5

ΔL

[m

m]

Ar flow [ln min-1]

Nozzle 2 (0,1 Pa s) Nozzle 2 (0,2 Pa s)

10

A logarithmic relation between ΔL and air flow is apparent for all nozzles with both dynamic viscosities by analyzing figure 20 where the air flow is plotted in log scale along with the R2-values tabled in table 1.

Figure 20: Nozzles 1-5 log air flow and ΔL for both dynamic viscosities where red represents a dynamic viscosity of 0.1 [Pa s] and blue 0.2 [Pa s]. Logarithmic trendlines as well as the associated R2-values are displayed.

Viscosity 0.1 [Pa s] R2-value

Nozzle 1 0,8537 Nozzle 2 0,952 Nozzle 3 0,9921 Nozzle 4 0,9848 Nozzle 5 0,9891 Viscosity 0.2 [Pa s] R2-value

Nozzle 1 0,9725 Nozzle 2 0,9202 Nozzle 3 0,9235 Nozzle 4 0,8971 Nozzle 5 0,9706 Table 1: The to every trendline associated R2-value.

0

10

20

30

40

50

60

70

5

ΔL [m

m]

Air flow log [ln min-1]

Nozzle 1-5(0.1 Pa s)

Nozzle 1-5(0.2 Pa s)

11

5. Discussion

5.1 Impact of nozzle geometry

5.1.1 Jet area

Using eq. (9) described in 1.6 “Turbulent jets” the impact area of the gas jet on the foam surface was calculated although due to restrictions in the mathematical model nozzles with inclined or non-circular outlets were not analyzed. Nozzle 1 and Nozzle 2, with circular outlets of Ø2 [mm] and Ø6 [mm] respectively were within the borders of themathematical model and will be further examinedas the jet area is important to identify as with it thework per area/pressure that the jets deliver can becalculated.

5.1.2 Kinetic energy vs momentum

In order to distinguish whether or not foam rupture correlates to the work delivered onto it or the pressure applied by the nozzle, the kinetic energy and momentum was calculated by determining the mass flow and the exit velocity of air through the nozzles.

By dividing momentum and kinetic energy with the jet impact area and time, the momentum per area and time [kg m-1 s-2] = [Pa] (pressure), and the kinetic energy per area and time, [J m-2 s-1] = [W m-2] (power per area), could be calculated and plotted against ΔL, see figure 21 and 22. A linear trendline was added to the momentum plot and an exponential trendline [x2] was added to the kinetic energy plot due to the mathematical nature of the underlying equations. By identifying which of these plots that had the least amount of deviation, by examining the respective R2-value, it could be concluded which parameter foam rupture is dependent of. It was found that the R2-value for momentum was 0.8182 and for kinetic energy 0.939, therefore it was concluded that foam rupture is more dependent on the kinetic energy and is influenced more by velocity of the gas since Ek ∝ mv2 than mass of the gas since p ∝ mv.

The same analysis was made for nozzle 1 although the R2-values were less than 0.9 and a statistical enforced conclusion could not be made and was therefore discarded .The sudden jump in value for ΔL = 65 [mm] is interpreted as a measuring error.

Figure 21: Momentum per area per second [kg m-1 s-2] plotted against ΔL [mm] with a linear trendline.

Figure 22: Kinetic energy per area per second [J m-2 s-1] plotted against ΔL [mm] with an exponential trendline.

5.1.3 Converging behavior

For figures 12-20, except in figure 17, a converging behavior can be observed for all nozzles. The value towards which ΔL converges is different for all nozzles. This indicates that the asymptotic value is dependent on nozzle geometry and therefore gas exit velocity as well as jet impact area. Even as the area density of kinetic energy dramatically increase as flow increases so

R² = 0,8182

0

5

10

15

20

25

30

55 60 65 70

Mom

entu

m [k

g m

-1 s-2

] ΔL [mm]

Nozzle 2

R² = 0,9395

0

200

400

600

800

1000

1200

1400

55 60 65 70

Kin

etic

ene

rgy

[J m

-2 s-1

]

ΔL [mm]

Nozzle 2

12

will the foam generation due to lower foam height, this foam generation will counter the foam rupture and establish a steady state. This goes to show that the more the foam has been ruptured the more energy it will take to rupture it further, as its height deviates from the natural state. This can be seen in figure 20 where the logarithmic behavior can be observed. This pattern might be the effect of film draining as mentioned in 1.2, since the foam on the top will be older than the one newly formed at the bottom, therefore it will be more likely to rupture [19]. This convergence is also dependent of the viscosity, which can be seen when comparing figure 12 with figure 14, note that ΔL converges much faster when using a higher viscosity, indicating that it is easier to suppress.

5.2 Oxygen as blowing gas

The atmosphere, at sea level, is a mixture of gases many of which are not desirable to expose a steel bath to. However the slag is exposed to the atmosphere protecting the steel bath from it. Blowing pressurized air could be a viable choice as long as it does not penetrate the foam entirely reaching the steel bath. Yet steel droplets are suspended in the foam and should they be exposed as bubbles rupture it will most surely impact the composition of the steel bath as they return from suspension. However, as mentioned earlier, an over-oxidized slag will produce more gas and therefore more foam, blowing pure oxygen into an over-oxidized slag might worsen the situation by promoting the chemical reactions producing process gases inflating the foam even more. Although in the case of a under-oxidized slag using oxygen blowing as foam suppression might help to stabilize the slags oxygen rate. Oxygen gas does deliver more kinetic energy at a given flow compared to air since it has a higher density. The blowing gas could be optimized for low composition impact with high kinetic energy for a given flow. However the cost per kinetic energy should also be considered.

5.3 Foam analysis

According to the foam index, gas flow is proportional to the foam height, however during the trials that were carried out the foam height decreased as argon gas flow increased, an inverse proportionality, as can be seen in figure 18. During measurements it was also noted that an increase in argon flow increased the bubble diameter of the foam. It has been calculated that the energy needed to deform a foam consisting of smaller bubbles, with Ø2 [mm], the same amount as a foam consisting of larger bubbles, with Ø4 [mm], is 2.65 times higher, due to the greater surface area of the foam with smaller bubbles [20]. This indicates that the negative impact that the bubble size has on foam height weights more than the positive impact of additional argon gas flow. During trials it was found that a higher dynamic viscosity allowed for more foam suppression at a given flow. This can be seen in figure 16 where all nozzles provide a greater foam reduction for the silicon oil with 0.2 [Pa s] compared to the silicon oil with 0.1 [Pa s]. It has been gathered that a higher dynamic viscosity of the liquid provides foam with a higher apparent viscosity and that a smaller bubble size also contributes to a higher apparent viscosity [20]. By increasing the argon gas flow the bubble size increased lowering the apparent viscosity making the foam harder to rupture. However the value of reduction increases with argon flow indicating that the influence of bubble size and therefore surface area seems to be more influential over foam rupture than the effect of viscosity decrease due to bubble size increase.

5.4 Kinetic energy vs momentum assessment by argon gas flow variation

By varying the argon gas flow, while keeping air flow constant, the foam height was lowered, as can be seen in figure 17. Since bubble diameter increased with an increasing argon flow the energy required to deform the bubble dwindled, as

13

has been discussed above. However the decrease in energy needed to deform bubbles could also be a conformation that it is indeed the kinetic energy that foam rupture is dependent on. Height reduction due to an increase in argon gas flow was plotted against kinetic energy per area and time, as can been seen in figure 23, in order to further investigate and strengthen the conclusion made in 5.1.3 Kinetic energy vs momentum. Yet again the correlation between kinetic energy and foam rupture is to be seen.

Figure 23: Kinetic energy per area per second [J m-2 s-1] plotted against Argon gas flow induced height reduction ΔL [mm] with an exponential trendline.

5.5 Jet momentum number and penetration depth

When blowing on a liquid surface with a gas jet, a cavity will form in the surface; this cavity will have a penetration depth dependent on many factors. Literature has shown that if all other parameters are held constant: decreasing the nozzle diameter, increasing the gas flow rate and decreasing the lance height will increase the penetration depth. It has also been shown that the penetration depth increases as the jet momentum number N increases. This is calculated using eq. (10). N = P

g𝜌𝜌𝑙𝑙h3 eq. (10) [21]

Where: 𝑃𝑃 = πρ𝑔𝑔𝑣𝑣2𝑑𝑑2

4, g is the gravitational

acceleration, ρl is the density of the fluid, v the velocity of the gas exiting the nozzle, ρg is the

density of the gas, h the lance height and d the diameter of the nozzle. Results, from experiments done in this study, show that a smaller nozzle diameter and a higher gas flow rate gives a lower foam height. As can be seen in figure 23 the area density of kinetic energy is lowered as the distance from the nozzle to the foam is increased, as mentioned this foam suppression is due to another factor, if this factor was to be consistent it would most surely be found that ΔL would decrease the further from the foam level the nozzle was positioned. Therefore it is concluded that there might be a correlation between penetration depth and foam suppression by gas jet. This would be fortunate if true as there has been a lot of research done on penetration depth.

5.6 Model basis in reality

The cold model is not a perfect representation of the foaming slag, lacking in surface tension, with only argon gas and no chemical reactions producing foam. A CaO-FexO-SiO2 slag has about 20 times higher surface tension compared to the silicon oil foam [22]. Silicon oil foam, as well as foaming slag, exhibits a non-Newtonian behavior making the apparent viscosity of the foaming slag difficult to reproduce and subsequently making the cold model suffer in reliability. [20]. The connection between surface tension and height reduction indicates that the model most likely impacts the real oxygen flow required for slag suppression in a positive manner, meaning the flow required for slopping prevention is lower in the cold model than in reality.

5.7 Economic viability

Even for a working concept there must be an economic incentive for businesses to adopt a change and invest in it. There are a number of factors that influence whether an investment is viable. “Hard factors” such as installation-, running- and staff costs, increase in yield and loss/gain in production time must be weighed against each other. Although there are “soft

R² = 0,9995

1517192123252729313335

50 55 60 65 70 75

Kin

etic

ene

rgy

[J m

-2 s-1

]

ΔL[mm]

Nozzle 2

14

factors” that are hard to quantify which should also be taken into consideration such as worker environment, risk of injury and lifespan increase of equipment. This analysis only sets the increase in yield in contrast to marginal running costs to loosely determine the economic viability of this method. Nozzle 2 is the most effective nozzle having the highest height difference to airflow ratio, requiring 20 [ln min-1] with a velocity of 106.16 [m s-1] to reduce the foam height by 71 [mm]. By comparing the Plexiglas cylinder with an inner diameter of 93.8 [mm] to a ladle with an inner diameter of 4.35 [m] this experiment is roughly a model scaled 2.16e-2 times the real size. By rescaling the nozzle diameter with this scale factor the real flow to achieve the same velocities is calculated to 0.7171 [m3 s-1]. As discussed above foam rupture most likely depends on the kinetic energy applied to the bubbles therefore if oxygen gas is used the flow should be weighed against kinetic energy since oxygen gas has a higher density of 1.4284 [kg m-3], calculated using the general gas law at SPT, compared to air which has a density of 1.1993 [kg m-3] [23]. The kinetic energy weighted flow is 0.605 [m3 s-1]. However, as already discussed, due to the cold models limitations this is still not the true value, the true flow is likely to be higher. Nevertheless this flow value will still be used as a baseline for slopping prevention in this economic analysis. For BOS-systems the FeO content is approximately 25% [24] therefore as the slag spills over iron is lost and the yield decreases. If slopping yield losses is comparable to that of deslagging yield losses it ranges from 0.5-1.5%. Other sources indicate that this number is closer to 0.5 % [1][25]. A loss in yield leads to a lower tonnage of end product. Calculated for a ladle heat size of 140 tons a 0.5% increase in yield would mean a 0.7 [ton] yield increase. If applied to the end product “Hardox Extreme”, produced at SSAB which carries a cost of 20 000 [SEK ton-1] [26], it would translate into a gain of 14 000 SEK per batch. With an oxygen gas cost of 0.5 [SEK m-3] [27], it would be economically defendable to run the oxygen suppression system for 771 minutes

per batch. If applied to the end product “S355 construction steel” produced at SSAB which carries a cost of 6 000 [SEK ton-1] [26], it would translate into a gain of 4 200 SEK per batch. It would be economically defendable to run the oxygen suppression system for 231 minutes per batch only considering oxygen gas costs. The breakeven oxygen gas flow calculated on a blow time of three minutes is 156 [m3 s-1], 47 [m3 s-1] respectively.

5.8 Sustainability

Steel production has a large environmental impact mainly due to the hefty volume of carbon emissions. It is estimated that steel production is accountable for 6.7% of the world’s total carbon emissions a value of 1.8 tons of CO2 is released per ton of steel with roughly 1.6e9 tons of steel produced per year as of 2014 [28]. Even the slightest of overall increases in productivity would mean a great deal for the amount of CO2 emitted. By preventing slopping a 0.5% productivity increase would lead to a reduction of 1.44e7 tons per year in carbon emissions further facilitating efforts toward a more sustainable steel production. Note that this calculation presumes that every batch slopps, in reality this is not true but no sources can be found on the percentages of slopping batches in the industry, as this often is a company secret.

6. Conclusion

All results indicate that by blowing gas on foam its steady state height will decrease even at relatively low flows. The geometry of the nozzle has an impact on the amount of foam that will rupture at a certain flow as the kinetic energy per area per second that the nozzle applies to the foam surface changes with geometry. A higher kinetic energy per area and time allows for a greater amount of volume to rupture. The flow rate besides nozzle geometry is also an important factor. The kinetic energy per area and time is proportional to the flow rate in tandem with

15

geometry since these factors decide the exit velocity of the gas and thereafter the amount of ruptured foam. As the distance between nozzle and foam surface increases, due to foam rupture, the kinetic energy per area will reach a point where it is no longer sufficient to overcome the foam generating parameters resulting in a converging behavior. The height at which this occurs varies with nozzle. It would seem that a high gas flow for maximum foam rupture would be preferable, although due to the logarithmic behavior of foam rupture vs flow rate the marginal gain of spending more resources to achieve a higher flow rate decreases. The point of breakeven was calculated for nozzle 2 by weighing the cost of oxygen flow against gain in yield resulting in an economically viable flow of 47 [m3 s-1] which is above the flow rate required for maximum foam rupture. A successful slopping prevention system would increase the sustainability of steelmaking by increasing the efficiency thus reducing the carbon footprint and energy need for steel production. Considering the practical success of this study and the loosely confirmed economic viability, this study is indeed a proof of concept for oxygen gas blowing as a slopping prevention system.

7. Future work

To further strengthen the concept of oxygen blowing as a slopping prevention method there are a number of things that requires additional research. The most obvious is to try this method on a warm model where slopping prevention can be simulated using a foaming slag that has all the associated reactions producing process gases for foam inflation creating a foaming slag with the appropriate surface tension and apparent viscosity. Although most likely an intermediary research step is required before warm modelling. A numeric description simulating slopping in a warm system with cold model experiments as validation could be such an intermediary step. Jet momentum number indicates the nozzles depth penetration ability. According to our results nozzles with a high jet index also decreased the

height of the foam pillar the most. For future work where focus lies on optimizing nozzles the jet momentum could be used as nozzle design parameter to get a rough idea of the effectiveness of the nozzle. Conducting studies on the impact of oxygen as blowing gas on the composition of the slag could also prove to be useful in determining if oxygen is indeed the gas that should be used.

8. SummaryCold model experiments for slopping prevention, with silicon oil foam to simulate foaming slag in BOS-systems, were carried out by blowing pressurized air onto a foam surface thereby reducing the foam pillar height. Experiments concluded that the steady state foam height reduction:

-Increases with a decreased nozzle diameter-Increases with an increased air flow-Decreases with increased nozzle height-Increases with an increase in Argon flow-Increases with an increase in foam viscosity

proving that mechanically blowing air onto a foam surface can be used for slopping suppression. Oxygen was proposed as the choice for suppression gas although whether it is a viable choice remains to be investigated. Preventing slopping is of both economic and environmental interest for the producer, as oxygen blowing, as a slopping prevention method, increases yields thereby increasing profit and lessens the environmental impact of steel production. Oxygen blowing is a concept that requires further development and research, although, that it has potential as a slopping prevention method is very clear.

16

9. Acknowledgements

We would like to extend our most sincerest of thanks to our supervisors and mentors Dr. Mikael Ersson and PhD student Johan Martinsson at the Dept. of Material Science and Engineering at KTH, for their never wavering patience when faced with the questions of lesser men and for lending us their knowledge and guiding us during this project.

A special thanks goes out to PhD student Martin Berg at KTH for his technical support and Oscar Hessling for his help during the design stage of this project.

17

10. References

[1] J. Stubbles, "The Basic Oxygen Steelmaking(BOS) Process," 2000. [Online]. Available:http://www.steel.org/making-steel/how-its-made/processes/processes-info/the-basic-oxygen-steelmaking-process.aspx. [Accessed 15May 2016].

[2] M. B. A. G. G. S. B. Shakirov, "Prediction andPrevention of Slopping the BOF," ZapTechCorporation, Atlanta, 2003.

[3] M. M. A. Evestedt, "Model-based SloppingWarning in the LD Steel Converter Process,"Elsevier, vol. 19, no. 6, pp. 1000-1010, 2009.

[4] M. Brämming, "Avoiding Slopping in Top-blownBOS vessels," Luleå University of Technology,Luleå, 2010.

[5] A. P. S. Lahiri, "Foaminess of slag: cause andcontrol," in Proc. 7th Int. Conf. on Molten SlagsFluxes and Salts, Cape Town, 2004.

[6] I. F. Kimihisa, "Study on the foaming of CaO-SiO2-FeO slags: Part I. Foaming parameters andexperimental results," MetallurgicalTransactions B, vol. 20, no. 4, pp. 509-514, 1989.

[7] S. F. R. Jung, "Foaming Characteristics of BOSSlags," ISIJ Int., vol. 40, no. 4, pp. 1225-1232,2000.

[8] P. B. C. R. Misra, "Dynamic Model of SlagFoaming in Oxygen Steelmaking Converters,"ISIJ Int., vol. 38, no. 11, pp. 1225-1232, 1998.

[9] R. J. R. Fruehan, "Slag foaming in bath smelting,"Metall. Trans., vol. 22B, pp. 481-489, 1991.

[10] J. A. Castillo, "Optimisation of the bottom stirringpraxis in a LD-LBE converter," UppsalaUniversitet, Uppsala, 2015.

[11] R. J., "Special measurements and control modelsfor a basic oxygen furnace (BOF)," University ofOulu, Oulu, 2012.

[12] C. Stroomer-Kattenbelt, "Modeling andoptimization of slopping prevention and batchtime reduction in basic oxygen steelmaking,"University of Twente, Enschede, 2008.

[13] Y. E. K. O. M. M. Y. O. M. a. Y. H. Iida, "FullyAutomatic Blowing Technique for Basic OxygenSteelmaking Furnace," Trans. ISIJ, vol. 24, no. 7,pp. 540-546, 1984.

[14] J. O. S. B. L. L. K. Ruuska, "Possibilities to UseNew Measurements to Control LD-KG-converter," in Proc. 5th EOSC, Aachen, 2006.

[15] O. A. S. B. D. D. S. K. Deo B., "Control of slagformation, foaming, slopping, and chaos in BOF,"Transactions of the Indian Institute of Metals,vol. 66, no. 6, pp. 543-554, 2013.

[16] Y. R. F. Zhang, "Effect of Carbonaceous Particleson Slag Foaming," Metallurgical and MaterialsTrans. B, vol. 26B, no. 4, pp. 813-819, 1995.

[17] S. V. K. M. S. M. Komarov, "Suppression of slagfoaming by sound wave," Tohoku University,Sendai, 2000.

[18] C.-R. B., "Turbulent Jets," in Enviromental FluidMechanics, Lebanon, New Hampshire, USA,Thayer School of Engineering, 2014, pp. 141-145.

[19] W. A., Principles of Foam Formation andStability, Berlin: Springer-Verlag, 1989.

[20] S. D. Martinsson J., "Study of apparent viscosityof foam and droplet movement using a coldmodel," Department of Materials Science andEngineering, Royal Institute of Technology,Stockholm, Sweden, 2015.

[21] F. A., A Study of Top Blowing with Focus on thePenetration Region, Stockholm: Department ofMaterial Science and Engineering Divion ofApplied Process Metallurgy Royal Institute ofTechnology, 2010.

[22] Verein Deutscher Eisenhuttenleute, Slag Atlas,Düsseldorf: Verlag Stahleisen mbH, 1995.

[23] F. E. Jones, "The Air Density Equation and theTransfer of the Mass Unit," Journal of Researchof the Notional Bureau of Standards , vol. 83, no.5, pp. 419-428, 1978.

[24] T. J. J. S. A. G. D. Miller, "Oxygen SteelmakingProcesses," in Oxygen Steelmaking Processes,Pittsburgh, The AISE Steel Foundation, 1998, pp.475-524.

[25] NIIR Board of Consultants & Engineers, TheComplete Technology Book on Hot Rolling ofSteel, Delhi: NATIONAL INSTITUTE OFINDUSTRIAL RESEARCH, 2010.

[26] Augustsson, Interviewee, Price of Hardox steel atSSAB. [Interview]. 13 May 2016.

[27] A. Nilsson, Interviewee, Cost of Oxygen gas atAGA. [Interview]. 13 May 2016.

[28] World Steel Association, "STEEL’SCONTRIBUTION TO A LOW CARBONFUTURE," World Steel Association, ISBN 978-2-930069-83-8, 2015.

www.kth.se