THEORETICAL AND EXPERIMENTAL STUDIES OF DISSIMILAR ...

1

THEORETICAL AND EXPERIMENTAL STUDIES OF DISSIMILAR

SECONDARY METALLURGY METHODS FOR

IMPROVING STEEL CLEANLINESS

by

APRIL PITTS-BAGGETT

LAURENTIU NASTAC, COMMITTEE CHAIR

MARK WEAVER

LUKE BREWER

CHARLES MONROE

RON O’MALLEY

A DISSERTATION

Submitted in partial fulfillment of the requirements

for the degree of Doctor of Philosophy

in the Department of Metallurgical Engineering

in the Graduate School of

The University of Alabama

TUSCALOOSA, ALABAMA

2017

1

Copyright April Pitts-Baggett 2017

ALL RIGHTS RESERVED

ii

ABSTRACT

Due to a continual increasing industry demand for clean steels, a multi-depth sampling

approach was developed to gain a more detailed depiction of the reactions occurring in the ladle

throughout the Ladle Metallurgy Furnace (LMF) processing. This sampling technique allows for

the ability for samples to be reached at depths, which have not been able to be captured before, of

approximately 1.5 m below the slag layer. These samples were also taken in conjunction with

samples taken just under the slag layer as well as in between those samples. Additional samples

were also taken during the processing including multi-point slag sampling. The heats were divided

in to five key processing steps: Start of heat (S), after Alloying (A), after desulfurization/start of

pre-Rinse (R), prior to Ca treatment (C), and End of heat (E).

Sampling sets were collected to compare the effects of silicon, desulfurization rates, slag

emulsification, slag evolution and inclusion evolution. By gaining the ability to gather multiple

depths, it was determined that the slag emulsification has the ability to follow the flow pattern of

the ladle deeper into the ladle than previously seen in literature. Inclusion evolution has been

shown by numerous researchers; however, this study showed differences in the inclusion grouping

and distribution at the different depths of the ladle through Automated Feature Analysis (AFA).

Also, the inclusion path was seen to change depending on both the silicon content and the sulfur

content of the steel. This method was applied to develop a desulfurization model at Nucor Steel

Tuscaloosa, Inc. (NSTI). In addition to a desulfurization model, a calcium (Ca) model was also

developed. The Ca model was applied to target a finished inclusion region based on the conditions

up to the wire treatment. These conditions included time, silicon content, and sulfur concentration.

iii

Due to the inability of this model to handle every process variable, a new procedure was created

to provide a real time feedback via SparkDat © software installed in a ThermoFisher 4460

spectrometer.

iv

DEDICATION

To my Parents who have always encouraged me,

To my Granny and Uncle Smitty who always look over me,

And To my husband for being beside me.

v

LIST OF ABBREVIATIONS AND SYMBOLS

LMF Ladle Metallurgy Furnace

‘S' Start of heat sampling time

‘A' After Alloys sampling time

‘R' After desulfurization/start of pre-rinse sampling time

‘C' Prior to Calcium treatment sampling time

‘E' End of heat sampling time

AFA Automated Feature Analysis

NSTI Nucor Steel Tuscaloosa, Inc.

BOS Basic Oxygen Steelmaking

EAF Electric Arc Furnace

MgO Magnesium Oxide

Al2O3 Alumina

CaAl Calcium- Aluminates

MgO-Al2O3 or MgO-TiO2 Spinel

MnS Manganese Sulfides

CaS Calcium Sulfides

SEM Scanning Electron Microscope

B3 Slag Basicity

CaO Lime/Calcium Oxide

vi

SiO2 Silicon Dioxide

( ) Represents components in the slag layer

[ ] Represents components in the steel

FeO Iron Oxide

MnO Manganese Oxide

LS Sulfur distribution ratio

CS Sulfide capacity of the LMF slag

Ko Partition coefficient for oxygen

Ks Partition coefficient for sulfur

fs Activity coefficient of sulfur

Δ Change in

T Temperature

Λ Optical Basicity

%Seq Percentage of Sulfur in equilibrium

K3’ Partition coefficient for sulfur [Roy]

% Sf Final Sulfur percent in the steel

% Si Initial Sulfur percent in the steel

WM Weight of steel

WS Weight of slag

A Surface area of the slag/steel interface

ms Mass transfer coefficient

HCFeMn High Carbon Ferro-Manganese

MCFeMn Medium Carbon Ferro-Manganese

vii

FeMnN Ferro-Manganese with Nitrogen

FeNb Ferro-Niobium

FeB Ferro-Boron

FeP Ferro-Phosphorus

FeTi Ferro-Titanium

FeSi Ferro-Silicon

RSi Restricted Silicon grade

PDA-OES Pulse Discrimination Analysis by Optical Emission

Spectroscopy

ASTM American Society for Testing and Materials

IRSID Institut de Recherche de la Sidérurgie/ Iron and Steel

Research Institute

CFD

Computational Fluid Dynamics

KTH Royal Institute of Technology

PkPA Pressure

ρslag Density of slag

g Gravity

hslag Slag thickness

ρsteel Density of steel

hsteel Steel depth

HMI Human Machine Interface

SO3 Sulfur oxide

XRD X-ray Diffraction

‘F’ Far side of ladle (when referring to sampling locations)

viii

rds Desulfurization rate

t Time

R2 R-squared value - a statistical value

OT Open Tap

MW Metal Weight

Al spec Aluminum specification

CalSil Calcium-Silicon Wire

CL Liquid concentration (Scheil equation)

C0 Initial concentration (Scheil equation)

fs Fraction solid (Scheil equation)

α Constant related to dendrite arm spacing

Ds Diffusion Coefficient

λ Arm spacing

Ω Constant that incorporates α and includes back diffusion

tf Time for solidification

∆G Gibbs Free Energy

HSLA High Strength Low Alloy

ix

ACKNOWLEDGMENTS

I am honored to have this opportunity to thank the many colleagues, friends, and faculty

members who have helped me throughout this research project. I am most indebted to NSTI for

granting me this opportunity to continue my educational progress. I could not have completed this

dissertation without my advisor, Dr. Laurentiu Nastac, as a university-industry collaboration

doctoral student. I would also like to thank all of my committee members: Mark Weaver, Luke

Brewer, Charles Monroe, and Ron O’Malley for their invaluable input, inspiring questions, and

support of both the dissertation and my academic career.

This research would not have been possible without the support of the LMF team, Nucor

management team, Dr. Eugene Pretorius, Bob Williams, Dr. Qiulin Yu, co-ops (Maggie Saylor

and Garret Finnen), and of course of my friends and family who never stopped encouraging me to

persist. I could not miss the opportunity to thank the LMF operators that were lucky enough to

work the days we ran trials. So many thanks to Tyler Waldrop, Jake Miller, and Chris Lyons! I

would also like to thank the many people who helped cover for me while I was in class or listened

to me when times were tough. Daniel Green, Justin Novotny, Jake Franks, Derrick Sawyer, and

Henry Przekora your patience and help was greatly treasured. I would also like to thank Michael

Mayhall and his electrical team for creating the ‘April’ button on the HMI to make these trials run

as smoothly as they did.

x

Also, I am also humbly grateful for the relationship with Heraeus Electronite on the

sampling tube design project. Paul Turner and Joseph Hallmark, words cannot express how much

I appreciate everything you did by taking the plunge on a crazy idea and help to make it work!

The constant support and love showed by my husband is valued beyond words. His ability

to stand beside me and encourage me when things got tough was incredible. I could not imagine

tackling this world with anyone else. Even though the words thank you is not enough, thank you

for continually being my rock.

Last but not the least, I would like to thank two of the most important people in my life –

my parents Terry and Catherine Pitts for giving me unconditional love and support throughout my

dissertation work, my career, and throughout my life. They have always pushed me to succeed,

and never let me say that I could not accomplish anything that I wanted. I reached for the stars

because of them, and I will always be eternally grateful.

xi

CONTENTS

ABSTRACT .................................................................................................................................... ii

DEDICATION ............................................................................................................................... iv

LIST OF ABBREVIATIONS AND SYMBOLS ........................................................................... v

ACKNOWLEDGMENTS ............................................................................................................. ix

LIST OF TABLES ....................................................................................................................... xiii

LIST OF FIGURES ...................................................................................................................... xv

INTRODUCTION .......................................................................................................................... 1

LITERATURE REVIEW ............................................................................................................... 4

Importance of Slag .............................................................................................................. 4

Desulfurization and the Effects of Sulfur ........................................................................... 6

Inclusion Formation and Evolution .................................................................................. 10

Calcium Treatment............................................................................................................ 12

Stir Practices ..................................................................................................................... 14

Sampling ........................................................................................................................... 16

Sampling Technique ............................................................................................. 16

Sample Evaluation using PDA-OES ..................................................................... 17

Modeling Methods and Techniques .................................................................................. 18

Thermodynamic and Mathematical Models ......................................................... 18

CFD Modeling ...................................................................................................... 20

xii

OBJECTIVES ............................................................................................................................... 25

EXPERIMENTAL AND MODELING METHODOLOGY ........................................................ 27

Experimental Approach .................................................................................................... 27

Analysis Methods.............................................................................................................. 31

Modeling Technique for Segregation of Mn in Argon Purged Sampler .......................... 33

EXPERIMENTAL RESULTS AND DISCUSSION ................................................................... 36

Slag Evolution – Transient Slags ...................................................................................... 36

Sulfide Capacity and Predicting Final Sulfur ................................................................... 48

Process affects on desulfurization and model ................................................................... 56

EAF Slag Carryover .............................................................................................. 66

Inclusion Evolution ........................................................................................................... 70

Sulfur..................................................................................................................... 70

FeSi Effects ........................................................................................................... 77

Pre-rinse versus Post Rinse ................................................................................... 80

Robot samples versus multi-depth samples .......................................................... 84

Slag Emulisificaton ........................................................................................................... 90

Calcium Treatment............................................................................................................ 98

Manganese Micro-segregation in Sample ....................................................................... 105

CONCLUSIONS......................................................................................................................... 110

CONTRIBUTION OF THIS STUDY ........................................................................................ 113

FUTURE WORK ........................................................................................................................ 115

REFERENCES ........................................................................................................................... 116

xiii

LIST OF TABLES

Table 1: Total oxygen with specific alloy additions. [46] ............................................................ 11

Table 2: Sampling locations.......................................................................................................... 27

Table 3: Chemistry used for HSLA Steels. ................................................................................... 34

Table 4: Material Parameters used for the modeling conditions. [127] [128] .............................. 34

Table 5: Thermo-physical properties of HSLA steel. [127] [128] ................................................ 34

Table 6: Slags from RSI1 from the ‘S’ sampling time. ................................................................ 36

Table 7: Calculated XRF measurements error for elements of interest. ....................................... 37

Table 8: Slag Data from heat RSI4. .............................................................................................. 39

Table 9: Slag compositions for RSI3 SF, AF, RF, CF, and EF. ................................................... 40

Table 10: Slag compositions for SI3 A1, A2, A3, AF .................................................................. 41

Table 11: Ranking of heats analyzed for sulfur removal and capture. ......................................... 52

Table 12: Optical Basicity of slag components used. [40] ........................................................... 52

Table 13: Sulfur equilibrium and rate constant comparison between experimental and Fruehan. 54

Table 14: Comparison of actual %Sf to Turkdogan’s equation and Equation 18. ........................ 55

Table 15: Si-bearing heat data related to desulfurization rates. .................................................... 66

Table 16: EAF and LMF slag component comparison. ................................................................ 67

Table 17: Parameters used for deoxidation and slag calculations. ............................................... 69

Table 18: Al Consumption per heat. ............................................................................................. 80

Table 19: AFA data comparison of multi-depth versus Robot samples for processing times ‘R’,

‘C’, and ‘E’. (RSI3) ...................................................................................................................... 86

xiv

Table 20: AFA data comparison of multi-depth versus Robot samples for processing times ‘R’,

‘C’, and ‘E’. (SI3) ......................................................................................................................... 88

Table 21: Data for micro segregation calculations. [128]........................................................... 107

xv

LIST OF FIGURES

Figure 1: Ca-Al phase diagram. ...................................................................................................... 6

Figure 2: Sulfide capacity versus optical basicity. [40] .................................................................. 9

Figure 3: Typical inclusion path throughout LMF Processing. [3] .............................................. 12

Figure 4: Schematic of stopper rod control. [73] .......................................................................... 14

Figure 5: Flow pattern for ladle with offset plugs used for argon bubbling. [120] ..................... 22

Figure 6: Probability of inclusion attachment using turbulent random motion. [120] ................. 22

Figure 7: Oxygen ppm and rate constant versus stir time. [120] .................................................. 23

Figure 8: The LMF hood design at Nucor Tuscaloosa, including sampling locations. [123] ...... 28

Figure 9: Cross sectional (a) and full view (b) of the multi-depth sampler .................................. 28

Figure 10: Stir lance arm used with sample tube attached............................................................ 30

Figure 11: Key process sampling locations. [124] ........................................................................ 31

Figure 12: Example Bruker S2 Ranger XRF spectrum analysis................................................... 32

Figure 13: Example ternaries depicting different modifications. [57] .......................................... 33

Figure 14: Images of the sample side-profile view (a) and top view (b). ..................................... 34

Figure 15: Chart of density vs temperature for bulk HSLA steel. [127] ...................................... 35

Figure 16: Slag evolution trends for heat RSI3.’S’- Start, ‘A’- Alloy, ‘R’- Pre-rinse, ‘C’- Ca

Treatment, ‘E’- End. ..................................................................................................................... 37

Figure 17: Slag evolution of heat RSI4 with non-ideal stir pattern. ............................................. 38

Figure 18: Stir pattern for RSI4. ................................................................................................... 39

Figure 19: Changes in slag color from heat RSI3 at each process step. ....................................... 40

xvi

Figure 20: Changes in appearance variations across width of the ladle. ...................................... 40

Figure 21: XRD analysis of LMF slags. ....................................................................................... 42

Figure 22: XRD Analysis of RSI3 AF. ......................................................................................... 43

Figure 23: B3 comparison between Si-bearing (a) and RSi (b). ................................................... 44

Figure 24: Alumina concentration in slag comparison for Si-bearing (a) versus RSi (b). ........... 45

Figure 25: Silicon oxide concentration in slag comparison for all heats. (Dashed line heats are Si-

bearing) ......................................................................................................................................... 45

Figure 26: Comparison of Si-bearing slags vs RSi slags. ............................................................. 46

Figure 27: Ca oxide concentration in slag comparison for all heats. ............................................ 47

Figure 28: Manganese Oxide concentration in slag comparison for all heats. ............................. 47

Figure 29: Iron Oxide concentration in slag comparison for all heats. ......................................... 48

Figure 30: Sulfur concentration of ‘Far’ side slag throughout key LMF process steps. .............. 49

Figure 31: Visualization of sulfur transfer from the steel to the slag versus time from RSI2. ..... 49

Figure 32: Comparing Young and Somerville’s equations for sulfide capacity. .......................... 50

Figure 33: Sulfide capacity versus B3 using Young and Somerville’s equations. ....................... 51

Figure 34: Sulfur equilibrium prediction plot for RSI2. ............................................................... 53

Figure 35: Comparison to actual final sulfur concentration versus different models. .................. 55

Figure 36: The effects of tap oxygen (ppm) on desulfurization rate and total sulfur removal. .... 57

Figure 37: The effects of arrival Al % on desulfurization rate and total sulfur removal. ............. 58

Figure 38: FactSage calculation of reduced oxygen ppm with increased Aluminum wt%. ......... 59

Figure 39: The effects of stir rate on the desulfurization rate. ...................................................... 60

Figure 40: The effects of arrival temperature (K) on sulfur equilibrium. ..................................... 61

Figure 41 Ca – Al Phase Diagram comparing temperatures B3 ratios. ........................................ 61

xvii

Figure 42: Effects of arrival B3 on desulfurization rate and total sulfur removal (OT heat

excluded) ....................................................................................................................................... 62

Figure 43: Incoming B3 versus Prewire (‘C’) inclusion index. .................................................... 63

Figure 44: Sulfur content in the steel throughout processing time for RSi. ................................. 64

Figure 45: Sulfur content in the steel throughout processing time for Si-bearing. ....................... 65

Figure 46: FeO and MnO trending for EAF slags (~2700 heats) ................................................. 67

Figure 47: Al consumption based on oxygen ppm present. .......................................................... 69

Figure 48: Weight of related slag components. ............................................................................ 70

Figure 49: Inclusion pattern for RSi grades. (RSI3) ..................................................................... 71

Figure 50: Inclusion pattern for RSi grade – non-typical. (RSI1) ................................................ 72

Figure 51: Inclusion pattern for Si-bearing heat (SI3) .................................................................. 72

Figure 52: Inclusion pattern for SI1*. Effects of furnace slag carryover. .................................... 73

Figure 53: Inclusion pattern for Si-bearing heat with late additions. (SI4) .................................. 74

Figure 54: Inclusion evolution comparing RSi and Si-bearing heats as well as Sulfur content. .. 75

Figure 55: % Mg from AFA inclusion data using robot samples to compare effects from

desulfurization rates. ..................................................................................................................... 76

Figure 56: Comparison of two different heats that resulted in different modification. ................ 77

Figure 57: Effects of FeSi additions on AFA results. ................................................................... 78

Figure 58: Comparison of the inclusion index versus total weight added of FeSi. ...................... 79

Figure 59: Solid inclusion % for each multi-depth location and process step S-C. ...................... 79

Figure 60: Inclusion index versus pre-rinse times. [57] ............................................................... 81

Figure 61: Post Rinse inclusion evolution. ................................................................................... 82

Figure 62: Inclusion count trend for robot samples. ..................................................................... 83

Figure 63: Inclusion index trend. .................................................................................................. 83

xviii

Figure 64: Multi-Depth Sampling of RSI3 ‘R’, ‘C’, and ‘E’........................................................ 85

Figure 65: Robot Samples of RSI3 ‘R’, ‘C’, and ‘E’.................................................................... 86

Figure 66: Multi-Depth Sampling of SI3 ‘R’, ‘C’, and ‘E’. ......................................................... 87

Figure 67: Robot Samples of SI3 ‘R’, ‘C’, and ‘E’. ..................................................................... 88

Figure 68: Inclusion index comparison charts for multi-depth samples and robot samples ......... 89

Figure 69: Ca/Al Ratio comparison for multi-depth and robot..................................................... 89

Figure 70: AFA inclusion data % Al comparison for robot vs multi-depth samples ................... 90

Figure 71: AFA inclusion data % Ca comparison for robot vs multi-depth for heat RSI3 (RSi) . 90

Figure 72: Slag Droplet AFA example. ........................................................................................ 91

Figure 73: Pictures and composition of slag droplets found in AFA’s......................................... 92

Figure 74: Emulsification droplets in comparison of stir flow trends. ......................................... 93

Figure 75: Effects of Arcing time on droplet formation. .............................................................. 94

Figure 76: Slag Droplet Inclusion Index effects on desulfurization rate and total sulfur removal.

(RSIOT excluded) ......................................................................................................................... 95

Figure 77: Droplet inclusion index versus total sulfur removed................................................... 96

Figure 78: Droplet inclusion index versus initial stir rate............................................................. 96

Figure 79: Number of slag droplets per sample comparing RSi and Si-bearing. ......................... 97

Figure 80: Inclusion index per sample comparing RSi and Si-bearing. ....................................... 97

Figure 81: Four different heats with all the same wire addition. [57] .......................................... 99

Figure 82: Calcium wire model program at NSTI. ..................................................................... 100

Figure 83: Stopper Rod traces. Ideal (a) Clogging (b) [57] ........................................................ 101

Figure 84: Stopper rod schematic with ideal stopper rod trend (right). ...................................... 101

Figure 85: Stopper Rod Deviation before and after model. ........................................................ 102

Figure 86: Example of the effects of late additions in conjunction with poor stir conditions. ... 103

xix

Figure 87: Histogram of the footage difference in SparkDat versus the wire model. ................ 104

Figure 88: Solid fraction at 4 seconds. (a) Temperature profile at 4 seconds. (b) ...................... 105

Figure 89: Mn concentration change throughout sample depth for LMF (a) and Caster (b)

samples. ....................................................................................................................................... 106

Figure 90: Schematic of sample. ................................................................................................. 106

Figure 91: Grain size analysis at various locations throughout the cross section of the sample. 107

Figure 92: Micro-segregation of Mn and S comparing methods. ............................................... 108

Figure 93: Inclusion count of MnS versus time and sulfur content in the steel. [54] ................. 108

1

INTRODUCTION

The fundamentals of the Ladle Metallurgy Furnace (LMF) were initially developed in the

1980’s. [1] These fundamentals have evolved and have been further refined. Though the

fundamental principles are still relevant, much more detail is given to the investigation of the

processes and practices at the LMF due to the increased demand for cleaner steel and more

challenging applications. The initial research and investigations were focused on Basic Oxygen

Steelmaking (BOS) process since it produced about 60% of the world's total output of primary

steel production. With the increase of the Electric Arc Furnace (EAF) flat-rolled mills, capable of

making automotive and higher end applications, the need for cleaner steels has become a topic of

interest for many steel-making companies and their customers.

There are two main types of inclusions, indigenous and exogenous. [2] [3] Exogenous

inclusions are formed due to factors during the process such as slag entrapment, mold powder, etc.

These inclusion types are typically very large, and are detrimental to the finished product

properties. [4] Indigenous inclusions are formed during the process such as through deoxidization,

Ca modification, refractory (MgO) reactions from the ladle walls, reoxidation, and impurities in

alloy additions. [5] Typical indigenous inclusions studied include: alumina (Al2O3), Ca –

aluminates (CaAl), spinels (MgO-Al2O3 or MgO –TiO2), manganese sulfides (MnS), and calcium

sulfides (CaS). Though these inclusions can be relatively large in size (>15 microns), it is abnormal

for indigenous inclusions to be larger than 10 microns. These inclusions are the main focus of the

current study.

2

It has been proven that clean steel has improved fatigue properties [6], increased ductility

due to the reduction of MnS, [7] increased fracture toughness [8] [9], and reduced defects such as

slivers. [10] Another potential problem of inclusions that tend to agglomerate, [11] such as

Al2O3/CaS and MgO- Al2O3, can cause stopper rod and/or submerged entry nozzle clogging at the

continuous caster. [12] [13] [14] [15] This can lead to a flushing event that allows the

agglomeration to release from the rod or nozzle and entrap into the steel slab. These issues have

detrimental impact on the high-end application products, such as line pipes (high toughness) and

automotive parts (surface finishing and formability), etc. Understanding the types of inclusions

that are detrimental to the end application as well as the types of defects specific inclusions can

form is important. These factors combine to aid in understanding what processes are needed to

take place at the LMF to be able to meet the clean steel expectations. A general guideline for the

industry has been presented to outline the cleanliness requirements of different applications as well

as a basic discussion on the inclusion types. [16]

Evaluation techniques are also a critical part of gaining a better understanding of steel

cleanliness in relation to inclusion formation and modification. [3] [16] The limitation to most

evaluation techniques such as Automated Feature Analysis (AFA) with a Scanning Electron

Microscope (SEM) is the amount of time it takes to gain usable results. This is obviously not

efficient and sufficient for a steel production process. Due to the ever changing inclusion amounts,

compositions, and conditions in the ladle, LMF operations are evaluating effective ways to get an

online method to determine Ca wire addition and cleanliness impact at the specific time in the

ladle. [17] Hardware and software packages built into spectrometers are in the forefront to be able

to produce useful inclusion data within seconds. These methods can be highly beneficial for

3

industrial applications, after the validation and calibration to the specific application; however,

none have been developed enough to be used in the LMF operations before this study.

4

LITERATURE REVIEW

It is known that inclusion type, population, size, composition, and shape can affect the

quality of the finished products. [18] [19] [20] [21] [22] Typically these factors or a combination

of these factors are used to determine the level of steel cleanliness. During LMF processing,

numerous events can play a role in each of these factors.

IMPORTANCE OF SLAG

The slag layer plays a very important role in steel cleanliness. [23] The ability for the

molten slag to be able to capture inclusions as well as intermix with the liquid steel is a necessity

in the LMF process. [24] [25] Given enough contact time, slags can modify the solid inclusions

present within the heat because of their interactions with the calcium present in the slag. [26]

However, in a production setting, there is typically not sufficient time to completely modify all the

inclusions to the composition of the slag, which is the equilibrium condition between the steel and

the slag. The use of thermodynamic tools, which calculates the equilibrium condition, is not

sufficient to completely understand the role that slag plays in the LMF process. [27] Understanding

the relationship between the kinetics and thermodynamics of the slag layer, and its ability to react

with the molten steel is frequently being evaluated. There are also variables that can adjust the slag

properties throughout the process such as sulfur (S) and silicon (Si) contents.

The slag layer is typically quantified using a basicity calculation, B3. This equation takes

into account the acids and bases in the slag. It also produces a ratio which helps determine if the

slag is liquid during steel making temperatures. Equation 1 shows the B3 calculation.

𝐵3 =𝐶𝑎𝑂

𝑆𝑖𝑂2+ 𝐴𝑙2𝑂3 (1)

5

If the slag is not liquid during steel making temperatures, the slag will not be efficient at

completing the specific functions that are desired from the slag. LMF slags are design too:

Protect the molten steel from atmosphere

Aid in desulfurization

Capture inclusions

Act as an insulator and hold in heat

Protect the arc

The slag composition changes throughout the heat. It is desired that the initial slag be fluid to aid

in desulfurization, and thicken as the heat progresses to be able to capture inclusions as well as

maintain temperature. However, the slag can be too thick (CaO saturated) which can result in a

non-liquid slag. This is reached when the B3 exceeds approximately 2.0. The desired composition

for slags is highlighted in the blue region located between the red dotted lines in Figure 1. The red

lines indicate the range of temperatures for the LMF process.

While the B3 of the slag is important, the iron oxide (FeO) and manganese oxide (MnO)

contents of the slag are also of interest. Due to some EAF slag carryover, FeO and MnO contents

in the slag are present early in the heat. These oxides are reducible, and should be reduced out of

the slag to a minimal concentration (FeO + MnO < 1%) to aid in desulfurization and inclusion

capture. By comparing heats with differing amounts of excessive furnace slag carryover, the

effects of these components can easily be seen.

6

Figure 1: Ca-Al phase diagram.

DESULFURIZATION AND THE EFFECTS OF SULFUR

Sulfur content in the steel has an effect on the final inclusion composition within the steel

bath. [22] Through laboratory and industrial experiments, it was shown that depending on the

sulfur content of the steel bath, at the time of Ca treatment, a different reaction other than to modify

alumina was present. [28] If the sulfur content was high (> 40 ppm) the formation of CaS was the

initial reaction that took place after Ca treatment was completed. For low sulfur heats

(approximately 7 ppm), the initial formation of CaS was not present. The Ca was observed to form

CaO or CaAl. This validates the criticality of wire addition time. When the sulfur is high, the Ca

attempts to react/bond with sulfur and alumina simultaneously, which in a production environment

the reaction time with the slag is too short to modify all the alumina. This reveals the reason that

the LMF desulfurization is critical. The desulfurization process occurs by the dissolved lime (CaO)

7

in the slag interacting with the sulfur in the steel during slag/steel mixing. This reaction is shown

below in Equation 2.

3 (CaO) + 2[Al] + 3[S] (Al2O3) + 3(CaS) (2)

where ( ) represents the component is in the slag,

[ ] represents the component is in the steel.

Certain conditions are needed to achieve a high desulfurization rate. [29] Those are listed below.

High temperature

Fluid slag

High mixing conditions at the slag/steel interface

Aluminum

Low oxygen potential

If any of the conditions mentioned above are hindered, reduced, or changed it will affect the

desulfurization rate of the steel and the time required to desulfurize the steel to the target value.

Desulfurization is dependent on both thermodynamics and kinetics aspects. [30] Therefore,

producing an effective industrial model that can predict desulfurization removal rates, and sulfur

capacities of slags is very challenging. Several researchers have developed models to calculate

the sulfide capacity of slags. [31] [32] [33] [34] The sulfide capacity of the slag is dependent on

several variables. A few of them include the oxides that are present in the slag such as Iron Oxide

(FeO) and Manganese Oxide (MnO), the thickness and viscosity of the slag, the amount of arrival

sulfur, the temperature (most models assume a constant temperature), etc. [35] [36] It is also

understood that by decreasing the content of FeO and MnO in the slag, the desulfurization rate of

a heat increases. [37] This can be attributed to both kinetic and thermodynamic relationships. The

interaction of the slag/steel interface is critical; however, the availability of deoxidants at the

slag/steel interface can affect the removal of the oxygen from iron (Fe) and manganese (Mn).

8

Since the slag/steel interface interactions are so critical, the argon flow rate and porous plug design

can also heavily influence the desulfurization process. The higher the flow rate the larger the

steel/slag interface becomes, which allows for quicker desulfurization time assuming that the heat

is deoxidized, or has a low oxygen potential.

Several researchers have developed equations for sulfide capacity in slags. The KTH

sulfide capacity equation is shown in Equation 3. This shows the relationship between the sulfur

distribution ratio (Ls) and the sulfide capacity (Cs). [32] A similar equation was developed by

Steneholm et al. for the distribution ratio (Ls) shown in Equation 4, which will be used in this

study. [38]

𝐿𝑆 =(𝑚𝑎𝑠𝑠%𝑆)+∆(𝑚𝑎𝑠𝑠%𝑆) 0

[𝑚𝑎𝑠𝑠%𝑆] − ∆[𝑚𝑎𝑠𝑠%𝑆]0 = 𝐶𝑆𝑓𝑆

𝑎𝑂( [𝑚𝑎𝑠𝑠%𝑂]−∆[𝑚𝑎𝑠𝑠%𝑂]0 ).

𝐾𝑂

𝐾𝑆 (3)

𝑙𝑜𝑔(𝐿𝑆) = 𝑙𝑜𝑔(%𝑆)𝑠𝑙𝑎𝑔

[%𝑆]𝑚𝑒𝑡𝑎𝑙= −

935

𝑇+ 1.375 + 𝑙𝑜𝑔(𝐶𝑆) + 𝑙𝑜𝑔(𝑓𝑠) − 𝑙𝑜𝑔 (𝑎𝑜) (4)

where Ko is the partition coefficients oxygen at the temperature,

Ks is the partition coefficient of sulfur at the temperature,

fs is the activity coefficient of the sulfur, 0(mass%S) is the initial contents of sulfur in the slag, 0[mass%S] is the initial contents of sulfur in the steel,

∆(mass%S) is the change in mass% of sulfur in the slag due to slag/steel interactions,

∆[mass%S] is the change in mass% of sulfur in the steel due to slag/steel interactions,

ao is the oxygen activity in the steel.

The relationship between the optical basicity of the slag (Λ) and sulfide capacity is shown by

Equation 5. This equation also incorporates the effects of temperature (T). [39] Experimental

results that were published at 1600 oC is shown in Figure 2.

𝑙𝑜𝑔(𝐶𝑆) =[22690−(54640∗ 𝛬)]

𝑇+ [(43.6 ∗ 𝛬) − 25.2] (5)

9

Figure 2: Sulfide capacity versus optical basicity. [40]

Young, et al used a sulfide capacity equation that incorporated optical basicity as well as

Al2O3 and SiO2 wt % from the slag. [34] This equation has been used for the CFD modeling of

slag emulsification, and it is shown in Equation 6. [41]

𝑙𝑜𝑔(𝐶𝑆) = −13.913 + 42.84 Λ − 23.82 Λ2 −11710

𝑇− 0.02223(%𝑆𝑖𝑂2) − 0.02275(%𝐴𝑙2𝑂3) (6)

These equations will be used to compare which method is most aligned to the current study.

Sulfide capacity is not the only important calculation to develop industrial models. Being able to

calculate the predicted final sulfur percentage is also important. This is developed by predicting

the equilibrium sulfur value since the heat approaches equilibrium towards the end of the

processing time. A series of equations can be used to predict the final sulfur (or sulfur equilibrium)

10

of the steel. This set of equations started with the mass transfer coefficient as shown in Equation

7. Where W/A is the weight of the metal divided by the surface area of the slag/steel interface.

The mass transfer coefficient is then substituted into Equation 8 to solve for the kinetic constant,

k. [42] After the kinetic constant has been calculated, it is placed into Equation 9 to solve for the

sulfur equilibrium. Turkdogan also developed an equation for predicting the final sulfur. [43] This

is shown in Equation 10. Where Si is the initial sulfur percentage, Seq is the sulfur content that is

in equilibrium with the slag, and Sf is the content at the end of stir time. The methods for predicting

final sulfur will also be compared to the results of the current study.

𝑚𝑠 = 5.49 ∗ 10−10 ∗ (𝑊

𝐴)

2

− 8.27 ∗ 10−7 ∗ (𝑊

𝐴) + 1.499 ∗ 10−3 (7)

𝑘 = 𝑚𝑠∗𝜌∗𝐴

𝑊𝑀∗ (1 +

𝑊𝑀

𝑊𝑠𝑙𝑎𝑔∗3000+

ln (𝐿𝑠)

24) (8)

[%𝑆𝑒𝑞] =[𝑆𝑖]∗𝑊𝑀

𝑊𝑀+(𝐿𝑆∗𝑊𝑆) (9)

%𝑆𝑓 = %𝑆𝑒𝑞 + (%𝑆𝑖 − %𝑆𝑒𝑞)𝑒−𝑘𝑡 (10)

INCLUSION FORMATION AND EVOLUTION

For an Al killed steel, alumina is formed through the deoxidation of the steel bath. [44]

[45] At Nucor Steel Tuscaloosa, Inc (NSTI), the standard practice is to block tap a heat. Block

tapping is where additions are added to the ladle while tapping the heat. At NSTI typical additions

are manganese (Mn), lime (CaO), calcium-aluminate (CaAl), and aluminum (Al). This results in

the heat arriving at the LMF typically fully – killed (oxygen below 10 ppm), and with a

considerable amount of alumina present in both the steel and the slag. Alumina can also form

through the addition of alloys since most alloy additions at the LMF are not pure, and can contain

Al as an impurity. Furthermore, several alloys have a considerable amount of oxygen associated

with them. Table 1 shows typical oxygen ppm associated with different alloy additions. For

11

example, whenever a Mn addition occurs during the processing it will result in more oxygen being

introduced to the ladle. This will require more deoxidant additions to be made.

Table 1: Total oxygen with specific alloy additions. [46]

Therefore, since alumina is present in steel as an inclusion from either deoxidation or

reoxidation, it is an inclusion to be focused on, since it can be detrimental to the castability and

cleanliness of the final product. Throughout the processing at the LMF, alumina inclusions interact

with the refractories [47] as well as with some residual Mg in alloys to form spinels. [48] [49] [50]

The inclusion evolution throughout a heat has been studied by several researchers. [3] [48] [51]

[52] [53] [54] [37] These studies have shown to follow a very similar inclusion path from alumina

(Al2O3) to spinel to Ca-Al or a liquid phase, as depicted in Figure 3. However, there are several

processing variables that can affect the path of the inclusions, increase their size, and/or change

the population of the inclusions. This makes effective and consistent Ca treatment a challenge,

which this study attempts to rectify.

12

Figure 3: Typical inclusion path throughout LMF Processing. [3]

It has also been seen that silicon, especially when added as ferro-silicon (FeSi), has many

residual elements such as Ca. [55] [35] It has been seen that the timing of this addition within the

process effects the inclusion concentration, population, and modification. [56] [57] Though it is

not as efficient as Al, Si can also be used as a deoxidizer. For the Al-killed steel, Si is added to

increase physical properties of the steel. However, Si-bearing heats act differently than those with

a Si restriction (RSi), typically 0.04% Si maximum. [58] [59] For Si-bearing heats, it has been

seen that when Si is present Al consumption is lowered, slag composition changes, inclusion

morphology and composition changes more rapidly throughout the heat, and desulfurization is

accelerated. [60] [33] These differences in Si-bearing grades allow for better interaction with the

slag interface which can improve steel cleanliness through better inclusions removal. [35] [61]

CALCIUM TREATMENT

One researcher, Zhang, et al, used phase stability diagrams to aid in determining the

composition of the inclusion during critical processing times. [62] The commercially available

13

thermodynamic software such as FactSage and Thermo-Calc, use a similar concept to calculate

the combinations of elements that are thermodynamically favorable. [63] [58] [59]. Ca treatment

is designed to modify the solid inclusions present to liquid inclusions at the casting temperatures.

[64] [50] This process not only changes the composition of the inclusions but also modify the

morphology of the solid inclusions, which can help in the final product properties if the inclusions

are held under a small diameter (< 10 microns). Predictions for the amount of Ca needed to modify

a particular inclusion type have also been calculated using the thermodynamic software packages.

[65] [66] [67] [21] However, the addition of too much Ca can also produce unwanted inclusions

such as Ca – Sulfides (CaS) and Manganese – Sulfides (MnS). [68] [35] [69] [70]

If alumina and/or spinels are remaining in the steel, they can ultimately cause clogging at

the continuous caster. [71] [72] Incongruously, CaS resulting from too much of a Ca addition can

also cause clogging at the caster. Therefore Ca treatment is critical. The stopper rod trend is used

for real time feedback of cleanliness from the caster. The stopper rod is designed to control flow

from the tundish to the mold through the submerged entry nozzle. The stopper rod moves up or

down to reach the desired flow through the machine. A schematic of the stopper rod as well as the

axis the rod moves on is shown in Figure 4. The rod should wear slightly throughout its campaign

(travel in a slight negative direction). However, if alumina based inclusions or CaS are present

they will agglomerate to the stopper rod forcing the rod to have to raise up to maintain flow. This

is an indicator of steel cleanliness. The stopper rod trace can alert the operations team of the

cleanliness of the heat; however, it can also move upward for width changes, low tundish weight,

and speed changes. Therefore, those have to be taken in account as well when observing the stopper

rod trace. Determining what kind of inclusion type or the source of the inclusion formation that is

causing the rod to rise, cannot be solved by the stopper rod trace.

14

Figure 4: Schematic of stopper rod control. [73]

STIR PRACTICES

Stir practices are a critical part of the LMF process. Typically, argon stir is ongoing

throughout the entirety of the time at the LMF. Argon is bubbled into the ladle with porous plugs

located at the bottom of the ladle. The location for the plugs can vary; however, it is beneficial to

have two plugs that are located symmetrically to each other. This allows for an optimized flow

pattern throughout the ladle. [74] Argon flow is typically continuous; however, it does vary

throughout the process. At NSTI turbulent flow is deemed to be greater than 0.99 m3/min (35

scfm), mixed turbulent flow is 0.57 – 0.99 m3/min (20-35 scfm), laminar flow or gentle rinse is

0.14 – 0.57 m3/min (5-20 scfm), and discontinued bubbling is below 0.14 m3/min (5 scfm). These

flow rates directly affect the amount of turbulence and slag/metal interface is present in the ladle.

Upon arrival of the heat at the LMF, the bubbling of argon is desired to be at high flow (~

1.42 m3/min or 50 scfm). This allows for high interaction levels of the steel and slag which is

15

desired for the desulfurization process. However, due to the turbulence that this flow rate creates

on the slag surface (open eye), it is desired that this time to be minimized. The longer the steel is

exposed to the atmosphere by the open eye, the more reoxidation can occur at the slag eye. After

the sulfur has reduced below the target level in the steel bath, the stir is turned down to minimize

a stir eye, but sufficient enough to maintain homogenization of the bath. This time is called pre-

rinse, which is also a very crucial part of the process. This time is used to float remaining inclusions

from the steel bath into the slag. [75] [31] [57] Therefore, the stir rate should still maintain flow in

the ladle, but without creating a large slag eye. The length of pre-rinse is typically dictated by the

process time available at the LMF. The slag composition and its viscosity during this time are

critical factors in being able to capture the inclusions. This step also homogenize the temperature

and composition of the steel bath. [76] At the end of pre-rinse, the Ca treatment occurs. This is

immediately followed by the post rinse. The post rinse is used to establish a homogenous

temperature and uniform inclusion modification. This stir rate is typically near or slightly less than

the rate during pre-rinse. The post rinse is used to achieve homogeneity throughout the heat and

full inclusion modification before delivering the heat to the caster.

However, due to conditions that cannot be controlled such as: EAF slag carryover, arrival

temperature at the LMF, ladles that are returned from the caster, ladle skull falling from the LMF’s

hood into the ladle, alloy contamination, stir plugs not functioning properly, extended time at the

LMF (> 2 hours processing time), expedited time at the LMF (< 30 min processing time), etc.

Process parameters like high stir rates or pre-rinse time, which are desired, may not be able to be

achieved. These are the conditions that can limit models developed for industry settings. To be an

effective model, abnormalities that occur during operations have to be accounted for in any model

used for industry settings.

16

SAMPLING

Sampling Technique

Two sampling techniques, i.e., vacuum samplers and duck-bill samplers, have been

studied. [77] [78] [79] [80] [81] [82] The results from the studies have been mixed such as location

of sampling, timing of the sample, as well as the effects of cooling through the different sampling

techniques. The ability for Mn to segregate during solidification can affect the final inclusion data.

It depends on the sample thickness and cooling rate of the sample. [83] The main inclusion type

that has shown the effects of sampling techniques are sulfides particularly MnS and occasionally

CaS mainly based on the sulfur content. [84] Other methods of sampling have been created to

attempt to understand the slag/steel interface. Nevertheless, the results from this study did not

match the work done by K. Beskow, P. Dayal et al. [85]

Mn segregation is a known fundamental solidification effect in solidifying steels. [86] The

amount of segregation is dependent on the percentage of Mn composition in the steel. [87] Mn

often reacts with sulfur to form Manganese Sulfides (MnS). These inclusions are harmful to the

finished product. Even though MnS are typically very small and usually pushed towards the

centerline of the steel, they can have deterimental effects on the finished properties if not properly

controled. The cooling rate at which steel solidfies can also change the amount of segregation as

well as the amounts of MnS formation that is seen. Modeling MnS formation helps to understand

the effects of different variables such as cooling rates and sample thicknesses. [88] The Scheil

equation, Equation 11, is a commonly used equation for non-equilbruim solidification. [89] Brody

and Flemings developed an equation to calculate the constant α for a specific system that was

correlate the dendrite arm spacing. This equation for α is shown in Equation 12. Clyne was able to

improve the accuracy of α by further developing the equation. The extended equation is shown in

17

Equation 13, where Ω becomes the new constant used to predict the composition changes at the

solidification interface. [90] The new term Ω is then used in place of α in the Brody and Flemings

model previously determined, Equation 14. [91] The Scheil equation and the Clyne, et al. equation

will be compared to see which method maybe more in line with the current data.

𝐶𝐿 = 𝐶0

(1−𝑓𝑠 )1−𝑘 (11)

𝛼 = 4𝐷𝑠𝑡𝑓

𝜆2⁄ (12)

𝛺 = 𝛼 [1 − 𝑒−1

𝛼⁄ ] − 0.5𝑒−1

2𝛼⁄ (13)

𝐶𝑆 = 𝑘 𝐶𝑜[1 − (1 − 2𝛺𝑘)𝑓𝑠](𝑘−1)

(1−2𝛺𝑘)⁄ (14)

where CL is the liquid concentration,

C0 is the initial concentration,

α is the constant for the system, Ds is the diffusivity,

tf is the time for solidification,

λ is the dendrite arm spacing/ average grain size,

k is the partition coefficient,

Cs is the solid concentration.

One of the issues in the production setting, is when two different sampling techniques are

used throughout the melting/casting process. [77] [84] This can result in slightly different

measurement values. This is due to a difference in the cooling/solidification rates.

Sample Evaluation using PDA-OES

In the last few years, major development work mainly focused on fast evaluation of

inclusions using spectrometers has been ongoing. [92] These methods are being developed for

production use, and are very close to being used on a daily basis in determining inclusions present

within a matter of seconds compared to hours with a Scanning Electron Microscope (SEM) using

Automated Feature Analysis (AFA). [93] [78] [79] The Pulse Discrimination Analysis by Optical

18

Emission Spectroscopy (PDA-OES) method is an example in this particular research aspect. Prior

to these methods, the most common method for evaluation is by AFA analysis through the SEM.

[94] [9] This method has been used to determine the inclusion population, composition, size (μm),

and modification, as well as their distribution. The AFA method also produces a calculation that

can be used as a guide on cleanliness called the inclusion index. [3] The inclusion index is a tool

used to compare heats on cleanliness in terms of surface area of inclusions contained. Equation 15

is used to calculate the inclusion index.

𝐼𝑛𝑐𝑙𝑢𝑠𝑖𝑜𝑛 𝐼𝑛𝑑𝑒𝑥 =𝑇𝑜𝑡𝑎𝑙 𝑖𝑛𝑐𝑙𝑢𝑠𝑖𝑜𝑛 𝑎𝑟𝑒𝑎∗0.17

𝑇𝑜𝑡𝑎𝑙 𝑎𝑟𝑒𝑎 𝑠𝑐𝑎𝑛𝑛𝑒𝑑 (15)

The surface area scanned for a typically AFA is approximately 50 mm2. This index will be used

throughout as a baseline comparison between heats, and certain data sets. Wang et al, developed a

way to covert the 2-D AFA data to 3-D information. [95] Other methods, such as ASTM E45 for

product analysis and a few other classification techniques, have been used. [96] In this dissertation,

the standard AFA method will be used for inclusion analysis. In addition, this study also has a

focus on using a PDA-OES software analysis to further develop an online Ca treatment calculation

for the LMF process.

MODELING METHODS AND TECHNIQUES

Thermodynamic and Mathematical Models

FactSage has been used in conjunction with mathematical models to attempt to predict

some key process parameters, such as desulfurization rates, reaction kinetics [97] [98] [99] [100],

inclusion floatation mechanisms [101], slag – inclusion absorption [102], and slag properties [103],

etc. These methods did not attempt to take into account other process parameters such as furnace

slag carryover, processing time variation, alloy residuals, stir deficiencies, or other process upset

19

anomalies. Its outlining assumptions are not realistic to a production setting, e.g., a flat slag/steel

interface, constant temperature, as well as (in most cases) constant slag composition. [104]

However, using these tools can simulate a part of the process with validation against industrial

experiments.

Since it is not possible to see inside the ladle, nor is it possible to see most inclusions with

the naked eye trying to see and understand the all kinetics and thermodynamics of the LMF process

is difficult. This alludes to the significance of modelling in predicting information that can be

applied to industrial processes. To develop a comprehensive model, it is essential to understand

fluid dynamics, inclusion kinetics, and the flow rate during each process step at an LMF.

Processing parameters are established on the basis of cause and effect. However, it is difficult to

tie all the thermodynamic and kinetic components together since depending on where in the ladle

one may have a stronger effect. A popular method of visually understanding a portion of the

process is through water modeling. Water modeling is typically conducted using oil and water.

This is used to physically simulate the flow patterns produced from different plug locations, flow

patterns versus different flow rates, slag velocity, time to reach homogeneity, as well as some

observations on slag emulsification. [105] [106] [107] [108] [109] [110] Pistorius used water

modeling to aid the LMF operations on the design of the plug locations, and to observe how

bubbles diffuse throughout a liquid steel bath under different flow rates and conditions. [111]

However, water modeling has its limitations too. It cannot predict the effects of slag composition

or the effects of a particular element on a steel bath. It is also difficult to predict inclusion

entrapment into the slag, and the basic process of desulfurization.

Seetharaman et al. [112] completed a desulfurization study which is using micro-models

as the foundation for the process models. This particular paper focuses on the desulfurization at

20

the refining stages. The time elapse that was found is similar to the same time frame that is seen

in the current study, which is approximately 10 minutes for desulfurization. However, it is noted

that a range was found in the actual processing that inconsistent injection rates of argon flow as

the root cause of the variability. This paper discussed the difference between the two-film theory

and the use of micro-models. Where the two-film theory holds the slag – steel interface as a straight

line, and it is known to not be flat. Therefore, this causes other means of interactions, which have

to be taken into account. Furthermore, it incorporated the model developed by IRSID, a

thermodynamic model, to determine activities of alumina as well as used to determine sulfide

capacities and viscosities of slags by assuming equilibrium in the mixing zone. These calculations

excluded the effects of interfacial area and the mass transfer coefficients, which are relevant due

to findings of slag droplets deep within the liquid steel bath in the current work. Other assumptions:

one plug mimics what two plugs do together just doubled, steel close to the upper ladle wall

contains less S than the bulk early in the process, lower sulfur content steel is transported down

into the bulk of the steel where it is mixed with higher sulfur steel. [112]

CFD Modeling

The use of computational fluid dynamics (CFD) modeling can be used to simulate the water

modeling observations while incorporating slag thickness and flow rates. [113] [114] It was found

that the ideal ladle configured with 2 porous plugs at 135 degrees apart is the most efficient. [74]

In a very similar study, Shivaram [115] found that the best location was mid-radius and

diametrically opposite each other. Costa, et al [116] studied multiphase flow in argon stir ladles

with different plug configurations to quantify mixing time by means of CFD simulation (ANSYS-

CFX 12.1 ©). This was also done by using tracer elements. Marins, et al. also used a tracer method

in conjunction with a CFD and mathematical model to predict mixing times, flow rates, inclusion

21

growth and flotation as a function of rinsing time. Alexis, et al used CFD modeling to optimize

stirring condition for the LMF process. [117] All of the listed work had some underlying

assumptions built in to the model. The assumptions were things such as: frictionless side walls of

the ladle, flat slag/steel interface, constant compositions, constant temperature, and constant slag

properties.

CFD modeling is a very useful tool to simulate the ladle process; if this tool is combined

with a mathematical model and/or a thermodynamic model such as FactSage, it can become a very

powerful tool. After combining these techniques together, the CFD method becomes capable of

predicting useful process parameters, such as desulfurization [112] [118], ideal slag composition

[118], ideal slag thickness, slag emulsification [41] and its effects amongst many other process

parameter. [119] For example, Aoki, et al. used three different types of models in conjunction to

simulate the inclusion removal within the ladle refining process. [120] The first model used an

Eulerian-Lagrangian algorithm being coupled with a 3D transient multiphase flow model to

compute the flow pattern that is created by argon stirring (governing equations solved by

FLUENT), Figure 5. The second model used was a micro-scale algorithm that calculates the

probability of an inclusion to diffuse with a bubble, Figure 6. And, the final model was completed

by incorporating the inclusion probability model with the Eulerian Lagrangian model to calculate

the inclusion removal rate. This rate was calculated using the initial oxygen content (50 ppm in

this study), and only decreasing it by inclusion removal that was attached to a bubble, Figure 7.

This model focused on oxide only inclusions and did not take into account the emulsification

phenomenon and other inclusion compositional data, as well as the morphology change of

inclusion. This approach also did not consider the variation of the slag composition and its effects

on the inclusions absorption with respect to the depth of the molten steel.

22

Figure 5: Flow pattern for ladle with offset plugs used for argon bubbling. [120]

Figure 6: Probability of inclusion attachment using turbulent random motion. [120]

23

Figure 7: Oxygen ppm and rate constant versus stir time. [120]

Ramstrom et al. [32] incorporated slag-metal reactions into a mass transfer model to

increase the accuracy of their CFD calculations. The paper used a thermodynamic model to include

relevant slag – steel reactions. The activities of the oxides in the slags were calculated based on

Björkvall’s thermodynamic model, [121] and the sulfide capacities were calculated using the KTH

group’s model. [122] This data set was tested against industrial data and found to be in relatively

close agreement. This model is a more general model and it does not incorporate the temperature

and slag composition changes; it is also a 2D model. The composition used does not match the

composition ranges at NSTI. The composition that Ramstrom used was carbon 1.02%, Mn 0.28%,

Cr 1.4%, and the most critical the S at 0.023-0.012% which is an order of magnitude higher than

in the current study. The Ramstrom et al. study was also completed under a vacuum. [32]

Similarity, Andersson et al. [118] developed a 2-D three-phase fluid flow model with

thermodynamic reactions to take in account reoxidation and desulphurization. This model was

found to be agreeable with industrial data on predicting slag composition which is a big factor in

24

desulfurization and inclusion capture. However, these reactions and model were performed under

a vacuum which is not used in the current work. [118] Recently, Senguttuvan and Irons [41]

developed a model using multiphase CFD methodology to simulate two-phase entrainment and

obtain estimates for the number and size of droplets formed so that the interfacial area for the slag

– steel reaction rates could be calculated. This model was validated using water modeling. It was

found that slag viscosity affected the droplet rate and size. Also, the simulations show that the slag

conditions selected to maximize the thermodynamic refining capability (sulfide capacity) may not

maximize the interfacial area and the refining rate. [41]

25

OBJECTIVES

1. Many researchers have studied slag, and its functions throughout the LMF process.

However, understanding the variations in slag across the steel bath throughout the LMF

process has not been seen in research. A better understanding of the ideal sampling location

as well as at what point in the process does the slag composition remain uniform (i.e., slag

evolution) will be determined by comparing multiple slag sample locations during the LMF

processing.

2. The fundamentals LMF process parameters for desulfurization has been researched. It is

desired to have a better understanding of the desulfurization rate under different process

parameters and process times throughout the LMF process. It has been seen that initially

the desulfurization rate tends to be slow, followed by a much faster rate until the sulfur

level plateaus again. Therefore, a better understanding of process effects on desulfurization

and sulfur removal is desired.

3. Current sampling techniques in industry involve capturing a sample at a particular depth

(typically ~ 0.6 m into the ladle). The ability to obtain multiple steel samples

simultaneously at different depths and locations within a ladle has not been done. The

ability to capture more than one sample at multiple depths will allow for a better

understanding of inclusion flow and evolution within the ladle.

4. Slag emulsification has been studied on a macro-scale. It has been seen, using water

modeling techniques and CFD, that slag droplets can reach depths below the slag layer.

However, looking at slag emulsification on a micro-scale has not been completed. Being

26

able to understand and quantify slag droplet formation, can bring insight into the refining

process at the LMF that has not been studied.

5. The effects of silicon in Al-killed heats is a topic of interest for many researchers. Silicon

has seen to have effects on slag conditions and LMF process parameters. FeSi residuals

can also play a role in achieving the desired modification of an inclusions population.

Therefore understanding the complex nature of Al-killed heats with silicon additions, will

help aid in producing a simple industrial scale Ca treatment model to produce a ‘clean’

product, and modified inclusions.

27

EXPERIMENTAL AND MODELING METHODOLOGY

EXPERIMENTAL APPROACH

To be able to compare results from multiple slag and steel samples simultaneously, a

sampling device has been designed and developed in this study. One obstacle that had to be taken

into account is the limited accessibility of the ladle through the LMF hood. The LMF at NSTI has

no accessibility to the backside of the ladle. However, due to the ladle consisting of two porous

plugs, it was assumed that the steel and slag flows in the ladle are symmetrical on either half.

Figure 8 shows a layout of the hood design at Nucor Steel Tuscaloosa. The ladle hood openings

as well as sampling locations are listed below in Table 2.

Table 2: Sampling locations

Location Sample/Action Completed

Above Argon Plus/Stir Eyes Three-prong slag sample just outside eye edge

Robot Robot steel samples, and all temperatures

Stir Lance Apparatus Multi-depth argon purged sampling device

Wire Addition Far side slag and wire treatment

Alloy Shoot No samples, bulk alloy additions

All of the sampling locations that were used to gather samples were taken from the flow pattern of

one porous plug.

28

Figure 8: The LMF hood design at Nucor Tuscaloosa, including sampling locations. [123]

To be able to collect molten steel samples throughout the process at the LMF, several

different sampling techniques were used at different locations across the top of the ladle. The focus

was placed on developing a multi-depth sampling device that controlled the depth at which the

samples were taken. The sampler design is shown below in Figure 9. The validation of this sampler

design was completed in the previous work, Pitts-Baggett et al. [124]. The amount of argon

pressure needed to maintain the back pressure for controlling the filling of liquid steel is calculated

via Equation 16.

(a) (b)

Figure 9: Cross sectional (a) and full view (b) of the multi-depth sampler

.

Metal carrier Paperboard tube

Non-splash

refractory

Lollipop sample

molds

Refractory plug

29

𝑃𝑘𝑃𝑎 = 𝜌𝑠𝑙𝑎𝑔𝑔ℎ𝑠𝑙𝑎𝑔 + 𝜌𝑠𝑡𝑒𝑒𝑙𝑔ℎ𝑠𝑡𝑒𝑒𝑙 (16)

Values used for calculations:

𝜌𝑠𝑙𝑎𝑔 = 3100 k𝑔/𝑚3 [125] 𝑆𝑡𝑒𝑒𝑙 𝑑𝑒𝑝𝑡ℎ𝑆1(ℎ𝑠𝑡𝑒𝑒𝑙) = 0.31 𝑚

𝜌𝑠𝑡𝑒𝑒𝑙 = 7000 k𝑔/𝑚3 [125] 𝑆𝑡𝑒𝑒𝑙 𝑑𝑒𝑝𝑡ℎ𝑆2(ℎ𝑠𝑡𝑒𝑒𝑙) = 0.91 𝑚

𝑆𝑙𝑎𝑔 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 (ℎ𝑠𝑙𝑎𝑔) = 0.23 𝑚 𝑆𝑡𝑒𝑒𝑙 𝑑𝑒𝑝𝑡ℎ𝑆3(ℎ𝑠𝑡𝑒𝑒𝑙) = 1.52 𝑚

g = 9.81 m/s2

Due to the size and weight of these samplers, the location was limited, as discussed

previously in Pitts-Baggett et al. A picture of the stir lance arm apparatus used with the sample

tube attached is shown in Figure 10. The location also required the ability to have a consistent

depth into the ladle even when selected test heats were not ran consecutively. This was

accomplished through Human Machine Interface (HMI) programming, after the desired depth and

the argon purge time was acquired.

30

Figure 10: Stir lance arm used with sample tube attached.

The LMF process was divided up into five key process steps. These steps were roughly 10-

15 minutes apart depending on the total LMF processing time of the particular heat. The key

process steps were heat arrival/start (S), after alloying (A), after desulphurization/the start of pre-

rinse (R), prior to Ca treatment/end of pre-rinse (C), and after the end of post rinse/ end of heat

(E). This is depicted in Figure 11. These letters will be used throughout the study to denote the

location and timing of the sampling.

31

Figure 11: Key process sampling locations. [124]

The multi-depth samples were taken using the stir lance system at NSTI. This located the

samples to be collected on the downward side of the stir flow pattern. This sampling system

allowed for the ability to gain samples at roughly 0.15 m (6 inches), 0.76 m (30 inches), and 1.37

m (54 inches) simultaneously in the ladle. During the collection of these samples, additional

samples were taken. These included a robot sample and temperature at NSTI’s normal operational

sampling location, as well as four slag samples. The slag samples were taken with a 3-prong

sampler starting at the edge of the stir eye going out roughly 0.30 m (12 inches) from the eye. The

sample taken on the opposite side of the ladle from the 3-prong sampler, ‘Far’, was taken near the

location of the multi-depth sampling area. This was completed to see variation at the eye versus

the bulk slag in the ladle during various processing steps. The method of slag sampling used at

NSTI by use of metal pipe was validated previously by Grip et al. [126]

ANALYSIS METHODS

The slag samples collected were prepared at the NSTI lab, and XRF analysis was

performed using a Bruker S2 Ranger, XRF Instrument to determine the composition of the slag.

The components analyzed included: CaO, Al2O3, Fe2O3, MnO, SO3, SiO2, and TiO2. An example

32

XRF spectrum is shown in Figure 12. Chemistry analysis for all the steel samples were analyzed

by using a Thermofisher 4460 scientific spectrometer.

Figure 12: Example Bruker S2 Ranger XRF spectrum analysis.

Automated Feature Analysis (AFA) analysis in an ASPEX Scanning Electron Microscope

(SEM) was performed on all steel samples for inclusion information. A 50 mm2 area is scanned

by the electron beam. When an inclusion is detected, the size of the inclusion as well as the

composition is gathered. The composition is determined by Energy Dispersive spectroscopy

(EDS). Each individual inclusion’s data is collected, along with pictures of the inclusions for

further investigation. The ternary compositional diagrams generated from the inclusions measured

by the AFA process are used for visual observations of the inclusion composition and distribution

throughout the process. These diagrams aid in determining the modification range of the

inclusions. Examples of these analyses can be seen below in Figure 13. These are shown in sets to

depict what a modified, under-modified, and over-modified heat can look like using AFA analysis.

The under modified diagram shows alumina based inclusions that are not completely Ca treated

33

(tail extension from the spinel region, Mg-Al). The over modified ternary shows inclusions shifted

towards the Ca corner of the ternary. This correlates to the formation of CaS, which are also

deemed harmful in clean steel applications and castability. Ideal target conditions are also depicted

in the modified ternary shown in Figure 13.

Figure 13: Example ternaries depicting different modifications. [57]

MODELING TECHNIQUE FOR SEGREGATION OF MN IN ARGON PURGED SAMPLER

To evaluate the variation seen when using an argon purged sampling technique, ANSYS

Fluent © 17.1 was used to model the segregation present during solidificaiton. ANSYS Fluent©

is one of the most-powerful computational fluid dynamics (CFD) software tool commercially

available. The sample dimensions are 35 mm diameter and 12 mm thick. The materials

surrounding the sample are stainless steel chill plates (4 mm), and a circular ceramic to form the

sample shape. Images of the sample assembly are shown below in Figure 14. The chemistry used

is presented in Table 3.

34

(a) (b)

Figure 14: Images of the sample side-profile view (a) and top view (b).

Table 3: Chemistry used for HSLA Steels.

C Mn S Si Al Nb Ti

0.065 1.1 0.007 0.15 0.02 0.03 0.012

The material parameters used in the ANSYS’s Fluent © 17.1 software, including

parameters for the elements of concerns, are shown in Table 4, as well as thermo-physical

properties of the HSLA steel shown in Table 5. Since density is a function of temperature, the

density for the bulk steel was set up to be a piecewise-linear function using six different data points.

Figure 15 is a chart that shows the data points that were used to construct the piecewise-linear

function for the bulk steel density changes throughout the simulation. This chart used the chemistry

data located in Table 3.

Table 4: Material Parameters used for the modeling conditions. [127] [128]

Material

Molecular weight

(kg/kg.mol)

Slope of Liquidus

Line (oC/wt%)/100

Partition

Coefficient

Diffusion in

Solid (m2/s)

Mn 55 -6.25 0.77 2.43E-13

S 32 -8.9 0.1 1.00E-12

Table 5: Thermo-physical properties of HSLA steel. [127] [128]

Material Density

(kg/m3)

Specific Heat

(Cp-J/kg-K)

Thermal Conductivity

(W/m-K)

Viscosity

(kg/m-s)

Steel (HSLA) see chart 800 23 0.003

35

Figure 15: Chart of density vs temperature for bulk HSLA steel. [127]

6900

7000

7100

7200

7300

7400

7500

7600

7700

7800

0 200 400 600 800 1000 1200 1400 1600 1800

Den

sity

Kg/m

3

Temperature (C)

Density vs Temperature for HSLA Steel

36

EXPERIMENTAL RESULTS AND DISCUSSION

SLAG EVOLUTION – TRANSIENT SLAGS

To study the variation in slag composition at the stir eye of the slag layer versus the bulk

slag layer, slag samples were taken at four different locations throughout the ladle. These locations

include the three prong sampling starting on the edge of the stir eye (‘1’, ‘2’, and ‘3’ where ‘1’ is

closest to the eye), and a sample from the opposite side of the ladle farthest away from the eye,

labeled as ‘Far’. As expected, the stir eye location in the slag layer has a lot of variation across all

elements compared to the ‘Far’ side of the ladle. This is especially seen in early sampling times

within the heat due to the vigorous stir used for desulfurization. Also, due to the high stir rates

early in the process the slag at the eye is the liquid portion of the slag, whereas the far side of the

ladle is more representative of the bulk slag. This can be seen in Table 6 due to the variability seen

in the composition especially near the eye.

Table 6: Slags from RSI1 from the ‘S’ sampling time.

Named Location MgO % Al2O3 % SiO2 % CaO % MnO % FeO % SO3 % B3

Eye Edge 5.09 31.77 2.94 56.12 2.00 2.04 0.61 1.62

5" from eye 4.83 29.77 3.90 53.72 3.38 2.91 0.42 1.60

10" from eye 4.90 29.90 3.73 54.46 3.33 2.65 0.41 1.62

Far Side of

Ladle 5.22 31.50 4.29 54.12 3.29 3.01 0.43 1.51

The slag trends appear to stabilize and become more uniform after different processes

within the ladle. An example of this is shown in Figure 16. This heat followed an ideal stir practice

pattern at NSTI, and resulted in a stabilized optimum basicity (1.5-1.7 B3) after approximately 10

minutes into the heat. This does not mean that the slag did not change after reaching optimum

37

basicity. However, as shown in the charts the final two samples (C and E) are almost identical

which alludes to a low MnO and FeO (a “good” slag) prior to wire since there was no change in

the slag after the Ca wire addition. Note: Error bars are included for the ‘S’ sample type in Figure

16. Table 7 shows the calculated error for range of each element of interest used in this study.

Figure 16: Slag evolution trends for heat RSI3.’S’- Start, ‘A’- Alloy, ‘R’- Pre-rinse, ‘C’- Ca Treatment, ‘E’- End.

Table 7: Calculated XRF measurements error for elements of interest.

29.5

30.5

31.5

32.5

33.5

34.5

35.5

36.5

Eye

Edge

6" from

eye

12" from

eye

Far Side

of Ladle

%

Al2O3

S

A

R

C

E

0.3

0.5

0.7

0.9

1.1

1.3

1.5

Eye

Edge

6" from

eye

12"

from

eye

Far

Side of

Ladle%

SO3

S

A

R

C

E

0

0.5

1

1.5

2

2.5

3

Eye

Edge

6" from

eye

12" from

eye

Far Side

of Ladle

%

MnO

S

A

R

C

E1.400

1.450

1.500

1.550

1.600

1.650

1.700

Eye

Edge

6" from

eye

12"

from

eye

Far Side

of Ladle

%

B3

S

A

R

C

E

Element Error

CaO 1.447

Al2O3 0.618

SiO2 0.071

MgO 0.166

MnO 0.005

FeO 0.028

SO3 0.030

TiO2 0.007

B3 0.009

38

It was also seen that when the established LMF stir practices cannot be followed, the slag

evolution does not reach optimum basicity as quickly. An example of non-ideal stir practices and

the effects it has on slags is presented in Figure 17. The alumina present in the slag has much more

variation due to a longer high stir time, which allowed for excessive reoxidation of the heat. This

heat required approximately 34 kg (75 lbs) more Al, and 90.7 kg (200 lbs) of slag deoxidant. This

directly impacts the B3 chart.

Figure 17: Slag evolution of heat RSI4 with non-ideal stir pattern.

When evaluating heat RSI4 numerically (Table 8), it is seen that the additions to the heat

during the process, does affect the slag composition. (NOTE: This heat only had 4 sets of samples

compared to the other heats. The ‘R’ and ‘C’ samples on this heat are the same.) This heat had a

considerable amount of slag deoxidants and Al additions compared to the other heats analyzed.

Figure 18 shows the stir pattern for this heat where the blue line is the flow from the primary

porous plug and the grey line is the flow from the secondary porous plug. The reason for the

extended high stir is directly related to the sulfur removal. This heat had a very low desulfurization

rate, which slowed down the sulfur removal. This will be discussed in more detail later on.

25

28

31

34

37

40

Eye

Edge

6" from

eye

12" from

eye

Far Side

of Ladle

%

Al2O3

S

A

R

C

E 1.3

1.4

1.5

1.6

1.7

1.8

Eye

Edge

6" from

eye

12" from

eye

Far Side

of Ladle

B3

B3

S

A

R

C

E

39

Table 8: Slag Data from heat RSI4.

Step Time

(Min)

Al2O3

%

SiO2

%

CaO

%

MnO

%

FeO

%

SO3

% B3

Lime

(kg)

Al

(kg)

Slag

Deox

(kg)

HC

FeMn

(kg)

S 0 32.59 3.68 54.37 1.94 1.80 0.56 1.50 680 334 1361

A 14 36.66 2.57 53.72 0.41 0.63 1.49 1.37 272 51 91 354

C 28 33.06 2.95 56.84 0.34 0.57 1.42 1.58 10 68

E 38 34.08 2.36 56.50 0.17 0.30 1.37 1.55

Figure 18: Stir pattern for RSI4.

The slag changes not only compositionally, but it also changes in its physical appearance

during the process. To observe the progression of slag color, slag pieces were set aside on heat

RSI3. As the heat progresses the slag goes from a black glassy texture to a creamy thicker texture.

The image in Figure 19 shows the color change as well as the texture change throughout the heat.

The process steps are labeled underneath each physical piece. The compositions for each process

step is shown in Table 9.

40

.

Figure 19: Changes in slag color from heat RSI3 at each process step.

Table 9: Slag compositions for RSI3 SF, AF, RF, CF, and EF.

Sample MgO % Al2O3 % SiO2% CaO% MnO% FeO% SO3% B3

RSI3 SF 4.58 32.2 4.97 53.7 2.66 2.74 0.58 1.45

RSI3 AF 5.51 35.2 3.14 55.2 0.45 0.65 1.20 1.44

RSI3 RF 5.25 35.4 2.55 55.1 0.38 0.39 1.14 1.45

RSI3 CF 5.21 35.3 2.19 56.2 0.15 0.38 1.14 1.5

RSI3 EF 5.24 35.4 2.45 56.4 0.16 0.69 1.08 1.49

To compare the color and texture variations across the ladle, a sample set from SI3 was

taken to compare the ‘A’ sample. This was also completed to compare the differences in color for

a Si-bearing (SI3) heat versus an RSi heat (RSI3). The ‘A’ sample was picked due to timing within

the heat. It is the slag sample set that has the most variation after time at the LMF. As seen below,

there is a slight color variation between the samples taken near the eye versus the sample taken on

the far side of the ladle.

Figure 20: Changes in appearance variations across width of the ladle.

41

This color variation is due to the difference in MnO and FeO contents shown in Table 10.

Cooling rate could also have an effect on the visual differences in the slags, a faster cooling rate

retains more glassy phases. The ‘Far’ side slag has the ability to cool faster than the three prong

sampler due to how the sample is collected. The data for the set of slags across the ladle are show

in Table 10. This shows a good example of how the same B3 value can look visually different due

to other elements being present. The glassy slag (‘Far’) appears to be much darker than the other

slags, while the actual FeO + MnO levels are lower.

Table 10: Slag compositions for SI3 A1, A2, A3, AF

Sample MgO% Al2O3% SiO2% CaO% MnO% FeO% SO3% B3

SI3 A1 5.59 28.3 5.24 56.0 0.59 0.83 0.88 1.67

SI3 A2 5.15 27.6 4.95 53.1 0.62 0.89 0.83 1.63

SI3 A3 5.41 28.5 5.26 55.5 0.6 0.85 0.87 1.65

SI3 AF 5.49 28.6 5.28 55.9 0.36 0.59 0.94 1.65

X-ray diffraction (XRD) analysis was also performed on the pictured slag samples to

determine if there were any variations in phases present. This was completed using a Bruker D8

Discover with a Co-k alpha x-ray source. The samples were run at 40 kV and 35 mA. A Vantec

500 2 dimensional detector was used for data collection. The results from these tests are plotted

below in Figure 21 and Figure 22. The remaining results tested from this sample data set were

contaminated (RSI3 RF and SI3 AF) or non-crystalline (RSI3 EF), therefore producing no results.

The differences in results are believed to be due to cooling, as well as some compositional changes

throughout the process. Samples from SI3 A1, A2, and A3 showed a Ca-Al phase, Figure 21. The

test was unable to pick up other traditionally seen phases. These samples were all taken at the same

time with the three prong sampling device approximately 10 minutes into the heat. The slags taken

from RSI3 were all ‘Far’ side slags taken at different processing steps. The sample for RSI3 CF,

which is taken just before wire treatment, also showed to have just a Ca-aluminate phase. However

42

for RSI3 AF, taken after alloy addition and during desulfurization, phases commonly associated

with slag were present such as: Periclase, Lime, Calcium Aluminate, Magnetite, and

Brownmilenrite (Figure 22). A study that was completed by Vlcek, et al. study showed that cooling

also had a major role in results produced from XRD analysis. [129] The XRD analysis was

determined to be not relevant to this study, due to the fact that the temperature at which the samples

were taken, the slags were 90% liquid, so the phases present in the XRD analysis were most likely

formed during cooling.

Figure 21: XRD analysis of LMF slags.

43

Figure 22: XRD Analysis of RSI3 AF.

To take a more comprehensive view at all the heats, all the data from the heats were plotted

using the ‘Far’ side ladle slag for each process step. After plotting all the heats together, it was

noticed there were some significant differences between Si-bearing heats and RSi heats. Starting

with B3, the charts in Figure 23 where split into heats that are Si-bearing, and heats that are RSi.

The charts show a considerable difference between the progressions of slags for the two different

types of heats. It is seen that the Si-bearing heats have a much higher rate of change between

samples, whereas the RSi heats show little change throughout the LMF process. The ability for

desulfurization based on slag B3 will be discussed later.

44

(a) (b)

Figure 23: B3 comparison between Si-bearing (a) and RSi (b).

By breaking out the components of the B3 calculations, it is easier to determine which

variables have the most influence on the change in Si-bearing heats versus RSi heats. Figure 24

shows the differences in alumina content; however, this is understood further when looking at

Figure 25. Alumina trends lower on Si-bearing heats; however, SiO2 is always higher for Si-

bearing heats. The increase in SiO2 in the slag when more Si is present follows metallurgical

fundamentals. Due to the competitive deoxidation between Si and Al, the addition of Si to an

aluminum killed steel allows for more Al to remain in the heat without the need for further addition

when reoxidation is present.

45

(a) (b)

Figure 24: Alumina concentration in slag comparison for Si-bearing (a) versus RSi (b).

Figure 25: Silicon oxide concentration in slag comparison for all heats. (Dashed line heats are Si-bearing)

It is also noted that the B3 values for Si-bearing heats are higher than for the RSi heats.

Since operators add CaO additions based on visual observations of slags, a FactSage calculation

was completed to see if the Si-bearing grades may be liquid at lower temperatures. If Si-bearing

0

1

2

3

4

5

6

7

8

S A R C E

% SiO2

SI4

RSI4

SI3

SI2

RSIOT

SI1*

RSI3

RSI2

RSI1

46

grades were more liquid at lower temperatures than the slag could have a more liquid appearance

to the operator prompting them to add more CaO. Figure 26 shows that there is not a big difference

in the % liquid for RSi (red line), and Si-bearing (blue line) when plotted versus temperature. The

reason for the higher basicity levels on Si-bearing heats will require more investigation at NSTI.

Figure 26: Comparison of Si-bearing slags vs RSi slags.

Figure 27 shows a trend for CaO percentage. All of the heats follow a general trend

throughout the process. However, it was noticed that in Figure 28 and Figure 29 the dotted green

line (SI1*) has a significant more MnO and FeO present early in the heat. This is indicative of

excessive EAF slag carryover. The problems associated with this is discussed further in the effects

of furnace slag carry over.

47

Figure 27: Ca oxide concentration in slag comparison for all heats.

Figure 28: Manganese Oxide concentration in slag comparison for all heats.

48

Figure 29: Iron Oxide concentration in slag comparison for all heats.

SULFIDE CAPACITY AND PREDICTING FINAL SULFUR

One of the major mechanisms of sulfur removal is dependent on the slag. The slag is the

only phase that can remove sulfur from the steel. As shown previously, the incoming slag has a

major role in total sulfur removal. To be able to see how much sulfur has been transferred to the

slag, a plot of the sulfide percentage in the slag during the various LMF processing steps is shown

in Figure 30. It is seen that there is difference in the final amount of sulfur captured by the slag

from each heat. The incoming sulfur content of the steel can play a role in how much sulfur is

captured by the slag, especially if the incoming sulfur content in the steel is low (< 0.03 %). It can

also be seen that as the sulfur is removed from the steel, the slag increases in sulfur concentration.

This can be seen in Figure 31.

49

Figure 30: Sulfur concentration of ‘Far’ side slag throughout key LMF process steps.

Figure 31: Visualization of sulfur transfer from the steel to the slag versus time from RSI2.

Several researchers have developed sulfide capacity (Cs) equations for slags. These are

typically shown as the log (Cs). These equations do not calculate the amount of sulfur that the slag

can contain, rather they depict how well a slag can capture sulfur. The lower the log (Cs) value is,

0

0.2

0.4

0.6

0.8

1

1.2

1.4

S A R C E

% S

O3

SO3

SI4

RSI4

SI3

SI2

RSIOT

SI1*

RSI3

RSI2

RSI1

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0

0.005

0.01

0.015

0.02

0.025

0.03

0.035

0.04

0 10 20 30 40 50 60

% S

O3

in t

he

slag

% S

in t

he

stee

l

Time (minutes)

S SO3

50

the more efficient the slag is at capturing sulfur. As shown previously in Equations 5 and 6, two

different methods for calculating the sulfide capacity using optical basicity (Λ) were provided. A

comparison of the two methods is shown in Figure 32. As shown, the results are fairly similar for

the RSI heats with the exception of the open tap (RSIOT) heat. However, there is a decent amount

of variation for Si-bearing heats. To mimic the graph in Figure 2 using the data set in this study, it

was seen that at higher basicity levels of the slag obtained from the two methods were agreeable.

These results aligned relatively well with Figure 2 for CaO-Al2O3 slags.

Figure 32: Comparing Young and Somerville’s equations for sulfide capacity.

-4

-3.5

-3

-2.5

-2

-1.5

-1

-0.5

0

SI2 SI1* RSIOT SI3 SI4 RSI1 RSI4 RSI3 RSI2

Lo

g (

Cs)

Sulfide Capacity

Young Somerville

51

Figure 33: Sulfide capacity versus B3 using Young and Somerville’s equations.

The sulfide capacity for the various heats studied does not vary greatly. However, to see

which equation lines up closer with this study a table is provided that ranked the heats in order of

the most sulfur removed from the steel, the most sulfur captured by the slag, as well as the Young’s

and Somerville’s sulfide capacity rankings are shown in Table 11. Since there are so many

variables that are required for ideal sulfur removal in the steel and capture in the slag, this ranking

system does not reveal the entire picture. However, by comparing the heats with the most sulfur

capture in the slag to the ranking of the best sulfide capacity, it is seen that Young’s model is more

closely aligned to the data that was seen in this study. It shows better agreement with Si-bearing

heats than Somerville’s equation.

-4

-3.5

-3

-2.5

-2

-1.5

-1

0.75 0.76 0.77 0.78 0.79 0.8 0.81 0.82 0.83 0.84 0.85

Lo

g (

Cs)

Optical Basicity

Sulfide Capacity

Young Somerville

52

Table 11: Ranking of heats analyzed for sulfur removal and capture.

Most

Sulfur

removed

from heat

Best

Sulfur

Capture

in Slag

Best Sulfide

Capacity

(Somerville)

Best

Sulfide

Capacity

(Young)

SI4 SI4 SI2 SI1*

SI1* RSIOT SI1* SI2

RSI4 RSI4 RSIOT SI4

RSI3 SI2 SI3 RSI1

RSI2 SI1* SI4 SI3

SI2 RSI3 RSI1 RSI4

SI3 RSI2 RSI4 RSI3

RSI1 SI3 RSI3 RSI2

RSIOT RSI1 RSI2 RSIOT

To be able to develop a sulfur model, the ability to predict the final sulfur is key. To be

able to predict a final sulfur value, the ability to accurately predict the sulfur equilibrium is key. It

was also noticed that when calculating the Ls value, the optical basicity of the slag had a strong

influence on the calculations. The values used to calculate the optical basicity for this study is

shown in Table 12.

Table 12: Optical Basicity of slag components used. [40]

Oxide Optical Basicity

CaO 1.00

MgO 0.78

TiO2 0.61

Al2O3 0.61

MnO 0.59

FeO 0.51

SiO2 0.48

Fruehan incorporated the weight of the steel, sulfur distribution ratio, and the initial sulfur,

% Si as shown previously in Equation 10 to calculate the sulfur equilibrium. However, it was

noticed that the calculated sulfur equilibrium on certain heats predicted a sulfur value above the

actual final sulfur. These were seen on heats that contained Si. By using Equation 17 with

53

experimental data, a plot was created for each heat. An example of this is shown in Figure 34,

where a trend line was added to be able to determine the best R2 value for the data set while varying

the %Seq.

𝑙𝑛 ((%𝑆−%𝑆𝑒𝑞

%𝑆𝑖− %𝑆𝑒𝑞) = −𝑘′𝑡 (17)

where %S is the sulfur at time (t),

%Seq is the sulfur equilibrium,

%Si is the initial sulfur content,

k’ is the rate constant.

The slope of the trend line is the rate constant, k’. The value k’ is equivalent to the rate, k,

multiplied by the area of the slag/steel interface, and divided by the volume of the steel. Since the

area of the slag/steel interface and the volume of the steel are assumed to be the same for each

heat, k’ will be used as the desulfurization rate. The k’ and R2 value changed depending on the

%Seq used. These values are compared in Table 13 for the calculated and the experimental values.

The experimental k’ values will be used as the desulfurization rate.

Figure 34: Sulfur equilibrium prediction plot for RSI2.

y = -0.1189x

R² = 0.9719

-7

-6

-5

-4

-3

-2

-1

0

0 10 20 30 40 50 60

ln(S

-Seq

/Si-

Seq

)

Time (minutes)

54

Table 13: Sulfur equilibrium and rate constant comparison between experimental and Fruehan.

It was determined that none of the Si-bearing heats were able to be predicted by Equation

10 as shown in Table 14, because of a high [%Seq] prediction. Therefore, to be able to predict all

Si-bearing heats as well as to gain a more consistent time step for all heats a new equation was

developed. These equations are shown in Equation 18 and 19. This equation incorpoarates the %Si

content of the steel to modify the [%Seq] for each heat. The results of this equation as well as the

predicted results from Equation 10 are plotted against the actual final %S content in the steel shown

in Figure 35.

(18)

(19)

Experimental Fruehan

Heat name %Seq R2

k' %Seq R2

k'

RSI1 0.0020 0.85 -0.09 0.0021 0.84 -0.095

RSI2 0.0016 0.97 -0.12 0.001 0.83 -0.087

RSI3 0.0015 0.98 -0.11 0.0015 0.98 -0.11

SI1* 0.0002 0.98 -0.05 0.0038 0.1 -0.039

RSIOT 0.0037 0.99 -0.076 0.0022 0.85 -0.042

SI2 0.0043 0.97 -0.14 0.0069 -0.21 -0.003

SI3 0.0013 0.95 -0.175 0.0034 -0.22 -0.004

RSI4 0.0027 0.99 -0.083 0.0034 0.99 -0.091

SI4 0.0004 0.99 -0.067 0.0046 0.12 -0.066

55

Table 14: Comparison of actual %Sf to Turkdogan’s equation and Equation 18.

Heat Actual

%Sf

Turkdogan

%Sf

Time Step

Predicted

Pitts-Baggett

%Sf

Time Step

Predicted

RSI1 0.0023 0.00231 12 0.00236 10

RSI2 0.0017 0.00168 10 0.00154 9

RSI3 0.0016 0.001612 16 0.00165 11

SI1* 0.0036 -------- -------- 0.00352 12

RSIOT 0.0038 0.00386 6 0.00392 6

SI2 0.0040 -------- -------- 0.00397 11

SI3 0.0014 -------- -------- 0.00121 8

RSI4 0.0049 0.00469 9 0.00488 8

SI4 0.0036 -------- -------- 0.0038 8

Figure 35: Comparison to actual final sulfur concentration versus different models.

0

0.001

0.002

0.003

0.004

0.005

0.006

RSI1 RSI2 RSI3 SI1* RSIOT SI2 SI3 RSI4 SI4

% S

f

Prediction versus Actual %Sf

Actual %Sf Turkdogan %Sf Pitts-Baggett %Sf

56

Some of the data acquired in this step will be used to validate our recently developed CFD

and reaction kinetics modeling approach [120] [135] [137] that can be used to predict the

desulfurization kinetics in LMF under various process conditions.

PROCESS AFFECTS ON DESULFURIZATION AND MODEL

As discussed, desulfurization has several key components that are required to allow sulfur

to be removed quickly. Those items are:

Low oxygen potential in both the steel and the slag

Aluminum to drive the desulfurization equation.

High temperature

Fluid slag

High mixing conditions between steel and slag interface

Desulfurization rate is highly effected by these parameters, and which can also affect how clean

the steel is in the final processing steps. Stirring conditions are critical. If the stirring conditions in

the ladle are not maximized, ability for the slag/steel interface mixing will be reduced. This will

also reduce the ability for the steel to reduce sulfur. If the sulfur remains high in the steel, the

effects of Ca treatment will be diminished and potentially harmful due to the formation of CaS.

The desulfurization rate was calculated using experimental data in Equation 21. After the

desulfurization rate constant was calculated for each heat as shown previously in Table 13, it was

plotted against the variables listed. The initial conditions of the process are typically designed for

desulfurization. After the sulfur drops below a threshold, which can vary by process parameters

such as sulfide capacity in the slag, slag/steel interaction, oxygen presence, etc., the removal slows

down significantly as shown in Figure 44 and Figure 45

57

Total sulfur removed, is the summation of all the sulfur removed from the steel throughout

the processing time, is also important. It is related to several similar components as desulfurization

rate. However, the total sulfur removal is more dependent on the incoming sulfur content, slag

composition, as well as on the processing time. Total sulfur removed is used to correlate certain

variables.

Starting with tap ppm as shown in Figure 36, there was no direct correlation to the

desulfurization rate for tap ppm. This was expected since the practices at the LMF use Al addition

practices at tap to remove oxygen / “kill” the heat upon arrival. This chart does not depict the

dissolved oxygen of the heat which is controlled by the Al/Al2O3 equilibrium, and the oxygen level

imposed by the slag/metal interface. The dissolved oxygen in the heat does effect desulfurization.

Figure 36: The effects of tap oxygen (ppm) on desulfurization rate and total sulfur removal.

Since Al is added at tap to deoxidize the steel, the arrival Al can depict how much tap

oxygen as well as additional oxygen introduced by the Mn tap addition is present. Figure 37 shows

that the R2 value for arrival Al does play a role in the desulfurization rate, as well as the significance

R² = 0.0441

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

600 700 800 900 1000 1100 1200

Des

ulf

uri

zati

on R

ate,

k

Tap Oxygen (ppm)

Tap Oxygen vs Desulfurization Rate

58

factor in the F-test results calculates to be 0.043. To show significance through the F-test the value

needs to be below a 0.15. This can be logically followed since the desulfurization equation does

require Al to be present as shown in Equation 2. Therefore, since the heat typically arrives at a

high temperature, and the process starts with a high stir rate, having a higher arrival Al can help

increase the desulfurization rate. The higher Al present ensures a lower dissolved oxygen which

in turn promotes desulfurization. FactSage data was used to show the decrease in oxygen

concentration with an increased Al wt%, Figure 38.

Figure 37: The effects of arrival Al % on desulfurization rate and total sulfur removal.

R² = 0.6465

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

0.00 0.02 0.04 0.06 0.08 0.10 0.12

Des

ulf

uri

zati

on R

ate

% Arrival Aluminum

Arrival Al% vs Desulfurization Rate

59

Figure 38: FactSage calculation of reduced oxygen ppm with increased Aluminum wt%.

Turkdogan showed that the desulfurization rate constant was directly correlated to the stir

energy. [43] Therefore, the results presented in Figure 39 were expected. The F-test value for

significance, 0.0049 along with a p-value of 0.035 aligned well with the R2 value shown in the

chart.

0

50

100

150

200

250

0.00% 0.05% 0.10% 0.15% 0.20% 0.25% 0.30%

Oxygen

PP

M

Aluminum wt%

60

Figure 39: The effects of stir rate on the desulfurization rate.

Desulfurization reactions happen faster at higher temperatures due to the slag being more

fluid at higher temperatures. FactSage shows that desulfurization is more favorable at lower

temperatures; however, the kinetic slag effects prove to have a much greater effect than the

difference in ΔG as a function of temperature. The sulfur equilibrium is effected by the incoming

temperature through the partition coefficient, LS. By plotting the incoming temperature versus the

%Seq a correlation is seen, Figure 40. This was also confirmed by the results from conducting a

regression analysis which showed the F-test significance to be 0.077 as well as a p-value of 0.069.

The higher temperatures allow for a slightly larger chemistry range for fluid slag which can have

a direct impact as well. Higher temperatures are great for sulfur removal; however, they can have

a negative impact on refractories if the slag is not saturated with lime. Therefore, higher is not

always better.

R² = 0.7006

0.00

0.03

0.06

0.09

0.12

0.15

0.18

0.21

1.0 1.1 1.1 1.2 1.2 1.3 1.3 1.4 1.4 1.5 1.5

Des

ulf

uri

zati

on R

ate

Initial Stir Rate (m3/s)

Stir Rate vs Desulfurization Rate

61

Figure 40: The effects of arrival temperature (K) on sulfur equilibrium.

As stated, a higher temperature can relate to a slightly wider range for a fluid slag. Figure

41 shows the CaO-Al2O3 phase diagram. The steel making temperature range is designated by the

red dotted lines, as well as where the B3 ranges from this data set lie.

Figure 41 Ca – Al Phase Diagram comparing temperatures B3 ratios.

This study shows that a more fluid arrival slag will help with increased sulfur removal.

Figure 42 is the first chart that shows a linear correlation between total sulfur removal, and the

R² = 0.492

0.0000

0.0010

0.0020

0.0030

0.0040

0.0050

1830 1840 1850 1860 1870 1880 1890 1900 1910

% S

eq

Temperature, K

Arrival Temperature vs % Seq

62

arrival slag B3. This observation is repeated in the regression analysis showed a 0.021 F-test

significance as well as a p-value of 0.0033.

Figure 42: Effects of arrival B3 on desulfurization rate and total sulfur removal (OT heat excluded)

It was also seen that the incoming slag B3 aids in better inclusion capture. When plotting

the inclusion index for the prewire sample (‘C’) versus the incoming B3, it is shown to have a

relatively significant R2 value. This is shown in Figure 43. The F-test results showed this to also

be significant at 0.111 for the significance with a p-value of 0.347; however, it was not as strong

as some of the other F-test conducted.

R² = 0.6172

0.00

0.01

0.02

0.03

0.04

0.05

0.06

1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80

To

tal S

ulf

ur

Rem

oved

Arrival B3

Arrival B3 vs Total Sulfur Removed

63

Figure 43: Incoming B3 versus Prewire (‘C’) inclusion index.

To investigate the removal of sulfur throughout the heat, the chemistry of samples taken

by the robot, was plotted against the sampling time for that particular heat. As shown in Figure 44,

the initial desulfurization removal rate is fairly similar for RSi heats. One heat that does show to

be different is the dotted line heat, RSIOT. This heat was the only Open Tap (OT) heat sampled.

This heat also arrived with the lowest sulfur value sampled. This was due to the scrap mix being

used at the EAF during that particular sampling time, and not related to the heat being open tapped.

The other heats were able to take advantage of the block tap additions, which allowed the start of

the desulfurization process prior to arrival at the LMF. However, it is also seen that the arrival

sulfur varies by a considerable amount (0.02-0.045 wt %). This is due to the scrap mix being used.

Another thing to note, is that the processing time for these heats also ranged from 38 minutes to

70 minutes. This can also effect the total sulfur removal, but it also shows how important quick

desulfurization is to the LMF process. As the chart shows, the desulfurization rate (slope of the

line) happens quickly within the first 10 - 20 minutes of the heat, then it plateaus off.

64

Figure 44: Sulfur content in the steel throughout processing time for RSi.

To evaluate at the effects of Si on the initial desulfurization rate (slope of the line) as well

as total sulfur removed the Si-bearing heats were plotted separately, Figure 45. It was noticed that

these plots follow a similar pattern as the RSi plots. These heats also saw a broad range of arrival

sulfur levels, 0.048 – 0.027%, as well as a slightly wide range for processing time 36 – 57 minutes.

However, there was a significant amount of variation in the individual plots. Therefore, a further

investigation of what may be causing this variation was completed.

65

Figure 45: Sulfur content in the steel throughout processing time for Si-bearing.

EAF slag carryover can play a major role on steel cleanliness. This will be discussed more

in depth later on in this study. Nevertheless, one way furnace slag carry over effects cleanliness is

by the increased oxygen content in the slag which ultimately affects the ability for the steel bath

to remove sulfur. Heat SI1* had a considerable amount of extra furnace slag carryover, and the

effects of retarded desulfurization rate can be seen in Figure 45. It can also be seen that both SI1*

and SI4 had a significantly higher arrival sulfur than most of the heats investigated. SI4 ended with

a similar sulfur level as SI1*; however, the processing time was 25% longer for SI1* which

allowed it to reduce to similar levels as SI4.

SI2 had the lowest arrival temperature of the Si-bearing heats, as well as the lowest

incoming MnO% content. It arrived at the LMF only 25 degrees higher than its final exit

temperature compared the other Si-bearing heats which were at least 75 degrees higher. This heat

also started with a high B3 (1.7) which is not ideal for desulfurization.

Heat SI3 was an ideal heat in terms of desulfurization and sulfur removal. Due to its low

starting sulfur percent, the amount of total sulfur removal was not as high as the other heats.

66

However, the rate of removal was very good. This heat also had a 166.5 kg (367 lbs) calcium-

aluminate addition at tap that heats SI4 and SI1* did not have. This allowed for an arrival slag

lower in harmful oxide components such as MnO and FeO by dilution. The comparison of several

processing variables is shown in Table 15.

Table 15: Si-bearing heat data related to desulfurization rates.

Heat Desulfurization

Rate, k’ Arrival

Temp (K)

Sulfur

Removed

Arrival

B3

Arrival

Al%

Arrival

MnO %

Arrival

FeO %

SI3 0.175 1876 0.026 1.58 0.027 2.47 2.25

SI2 0.14 1842 0.029 1.70 0.045 2.1 2.54

SI4 0.067 1879 0.045 1.31 0.097 2.85 2.85

SI1* 0.05 1892 0.045 1.26 0.051 6.35 3.98

Since the primary data set of this study shows only a small portion of the heats produced,

a data set of over 2600 heats was compiled to compare the final sulfur of RSi heats versus Si-

bearing heats. The data set encompasses all ranges of grades, as well as various different arrival

sulfur levels due to various different scrap mixes used during the date range. The data shows that

for Si-bearing heats an average final sulfur was 0.0032 leaving the LMF, and for RSi heats the

average final sulfur was 0.0047. The standard deviation was 0.0014 and 0.0016 respectively.

Without process upsets, it is seen that the Si-bearing grades reduce sulfur more efficiently.

EAF Slag Carryover

Every heat at the LMF has some slag carryover from the EAF, estimated 453.6 kg (1000

lbs). Depending on the time in the life cycle of the tap hole the slag carryover amounts can very

some. Also, operational anomalies can cause more slag carryover as well. EAF slag contains

elevated and variable concentrations of oxygen mainly in the form of FeO and MnO. Table 16

show the difference in a typical EAF slag versus LMF target slags. Looking at FeO and MnO

67

variation in EAF slags from NSTI, over 2700 slags were compiled and plotted in Figure 46. This

shows the variation in MnO from 2.0 - 8.7%, and 13.5 – 55.2% for FeO. The variation in the

incoming reducible oxides amounts has an effect on the deoxidants usage that are required at the

LMF. In terms of desulfurization, the elevated oxygen concentrations are harmful. This will result

in a higher consumption of Al for alumina formation, which will also slow down the

desulfurization rate. This ultimately slows down all the steps of the LMF process, and can

negatively influence the cleanliness of the steel.

Table 16: EAF and LMF slag component comparison.

Composition Carryover EAF Slag Final LMF Target Slag

% MgO 10 5

% CaO 25 55-60

% Al2O3 5 25-35

%SiO2 14 5-8

% MnO 3 < 0.5

% FeO 40 < 0.6

Figure 46: FeO and MnO trending for EAF slags (~2700 heats)

0

2

4

6

8

10

12

14

16

18

20

0

10

20

30

40

50

%M

nO

% F

eO

EAF Slag

FeO MnO

FeO

MnO

68

Investigating the effects of more oxygen present on the LMF process. A series of equations

were used to calculate the Al required. Equation 20 uses the change in oxygen multiplied by the

atomic weight ratio of Al to oxygen in Al2O3. [40] This total is then multiplied by the weight of

the steel ( ). However, since the addition of Al does not product 100% recovery in the steel, an

additional calculation must be completed to know the actual weight needed to recover and maintain

a certain Al concentration (Al spec) in the ladle, Equation 21. [40] The recovery percentage used

in this study is 70%.

𝐴𝑙 𝑟𝑒𝑞𝑢𝑖𝑟𝑒𝑑 = Δ[𝑂]

1000000∗ 1.125 ∗ 𝑀𝑊 (20)

where ∆[O] is the change in oxygen (ppm),

MW is the weight of the steel.

𝐴𝑙 (𝑤𝑖𝑡ℎ 𝑟𝑒𝑐𝑜𝑣𝑒𝑟𝑦) = 𝑀𝑊((

Δ[𝑂]

100000∗1.125)+𝐴𝑙 𝑠𝑝𝑒𝑐)

𝑅𝑒𝑐𝑜𝑣𝑒𝑟𝑦 (21)

Using the data from Table 17 with the listed equations, a plot of Al consumption based on

oxygen content is shown in Figure 46. This shows a significant amount of additional Al needed

for additional oxygen present. For a typical heat at NSTI, the tap ppm ranges from 600-1000 ppm.

However, the amount of additional oxygen introduced from furnace slag carryover can vary

significantly based on the amount of furnace slag as well as the percent of FeO and MnO present

in the carryover slag. For example, heat SI1* had a tap ppm of 1074, but the Al consumption was

over 500 kg (1100 lbs). By the chart in Figure 46, this is equivalent to approximately 2500 ppm of

oxygen being present. The EAF slag composition for FeO and MnO was 42% and 2.8%

respectively. SI3 had one of the lowest FeO and MnO EAF slag composition percent at 35% and

2.7% respectively. This heat had a tap ppm of 1110, but its Al consumption was only 327 kg (720

69

lbs) total. This relates to roughly 250 ppm oxygen coming from the furnace slag carryover. Which

is 10 times less than the heat with known additional EAF slag carryover.

Table 17: Parameters used for deoxidation and slag calculations.

Tap Oxygen 900 ppm

Ladle Size 125 Tons

Final Oxygen (LMF) 4 ppm

Aluminum Target 0.03%

Al Required 114 kg

Aluminum Recovery 70%

Al required (70% recovery) 212 kg

Al2O3 to slag 622 kg

Assumed SiO2 7%

Assumed Al2O3 28%

Assumed CaO 59%

Slag Weight 1007 kg

CaO needed for 1.7 B3 268 kg

Figure 47: Al consumption based on oxygen ppm present.

By knowing the total Al addition, the alumina that is sent to the slag layer can also be

calculated, Equation 22. Using the slag composition of a particular heat, the slag mass can also be

0

100

200

300

400

500

600

700

500 1000 1500 2000 2500 3000

Alu

min

um

Wei

ght

(kg)

Oxygen PPM

Al Required (Kg) Al (with recovery)

70

calculated as well as the CaO addition required to balance the slag to the desired B3 value. This is

shown in Figure 48 for oxygen contents ranging from 600 ppm to 3000 ppm.

𝐴𝑙2𝑂3 𝑡𝑜 𝑠𝑙𝑎𝑔 = (𝑇𝑜𝑡𝑎𝑙 𝐴𝑙 𝑎𝑑𝑑𝑒𝑑 − (𝑀𝑊 ∗ 𝐴𝑙 𝑠𝑝𝑒𝑐)) ∗ 1.89 (22)

Figure 48: Weight of related slag components.

INCLUSION EVOLUTION

Sulfur

Thermodynamic tools such as FactSage, depict that inclusions will eventually match the

composition of the slag as the slag and metal approach equilibrium conditions. These are

equilibrium calculations and not time dependent whereas the conditions that exist at the LMF are

rarely at equilibrium, and will only start to approach equilibrium as the heat is processed. To

modify and change the inclusion composition as well as morphology involves time, and hence

requires kinetic consideration.

0

500

1000

1500

2000

2500

3000

3500

4000

500 1000 1500 2000 2500 3000

Wei

ght

(kg)

Al2O3 to slag (Kg) Slag Mass for 28% Al2O3 (Kg) CaO needed for 1.7 B3 (Kg)

71

This study shows that sulfur can act as a barrier to Ca transfer from the slag to the metal

for inclusion modification. For RSi heats, inclusion compositional grouping is known to typically

be very tight or compositional clustered together. A typical inclusion distribution for the ‘A’-‘R’-

‘C’ process steps is shown in Figure 49. The grouping on the sample before wire, ‘C’, shows a

slight spread. However, the grouping only starts to change after the sulfur reaches below a 0.003%.

This heat was only below a 0.003% for 6 minutes keeping the grouping fairly tight. However,

when looking at another RSi heat for the same set of AFA’s, Figure 50, the grouping starts to

spread out starting in the ‘R’ sample. This sample was taken 25 minutes into the heat, leaving 18

minutes of processing time prior to wire treatment. This resulted in a much larger tail towards the

liquid area on the Mg-Al-Ca inclusion plot. This heat was depleted of Al between ‘A’ and ‘R’

sampling which resulted in a slight increase in sulfur between the two samples. Also, this heat had

an excessive amount of arc time which could have superheated the slag causing the slag to be more

fluid. This could have potentially resulted in more slag/steel interaction.

Figure 49: Inclusion pattern for RSi grades. (RSI3)

72

Figure 50: Inclusion pattern for RSi grade – non-typical. (RSI1)

For Si-bearing heats, the grouping is rarely tight. This makes it difficult to determine

correct wire footage for Ca treatment. It also makes it easier to over treat the steel bath. This

was believed to be related to the Ca present in FeSi. The residual Ca present in FeSi can have

an effect on the inclusions especially in a late addition; however, it is also seen that sulfur can

play a role in the groupings as well for Si-bearing heats. This is much more challenging since

typically Si-bearing heats tend to get to low levels of sulfur easier and more rapidly than RSi

heats. An example of a typical Si-bearing AFA with a low sulfur percent is shown in Figure

51. Even though there was a 100 kg (220 lb) FeSi addition after the ‘A’ sample, the tail of

inclusions is very long in comparison to the other FeSi heats that did not desulfurize as low

and as quickly as SI3.

Figure 51: Inclusion pattern for Si-bearing heat (SI3)

73

As discussed previously, EAF carryover slag can have an effect on almost every function

of the LMF process especially when seen in excessive amounts. The heat SI1* as shown in Figure

52, shows a dramatic difference in the inclusion distribution due to the excessive EAF slag

carryover. This heat was very similar to SI3, in how quickly the sulfur depleted from the heat;

however, the amount of excess alumina formation on this heat caused the slag to be very fluid until

just before Ca treatment. This happened even with an excess of over 360 Kg (800 lbs) of lime

added to this heat. This heat had a fairly long process time of 57 minutes. However this heat ended

very clean. This is related to the additional stir time. By comparing this heat (SI1*) to another Si-

bearing heat (SI3) with less total time (38 minutes), the inclusion index for the pre-wire samples

was 4.7 for SI1* versus 13.8 for heat SI3.

Figure 52: Inclusion pattern for SI1*. Effects of furnace slag carryover.

As suggested, a late FeSi addition may play a role in the inclusion tail formation. SI4 never

dipped below the sulfur threshold to create the tail due to slag interaction; however, after each

alloy addition there was a shift in the tail, Figure 53. This shift is not as extreme as the shift that

was seen in the very low sulfur heat SI3; however, it does exist. This shift is closely related to the

residual Ca being added with the FeSi addition.

74

Figure 53: Inclusion pattern for Si-bearing heat with late additions. (SI4)

This study showed that inclusion distributions are affected by different means, it was

desired to plot the inclusion paths on the same chart to see variation. Figure 54 shows the inclusion

evolution broken down by RSi and Si-bearing heats. These charts were created using the AFA

average point for each process step, and connecting to form the lines seen. This figure also includes

some very basic information about the heats listed below it. For RSi grades, the inclusion path is

very dependent on sulfur. Both RSI2 and RSI3 followed the exact same inclusion path, and they

both had similar sulfur values. As the sulfur value increase, the peak of the inclusion path becomes

lower on the chart. The final average point is very similar on both sets of graphs. However, note

that the final average point for Si-bearing grades is closer to the Ca corner of the ternary. Looking

at the Si-bearing heats, the lines overlap very well on three out of four heats. These three heats all

75

have the same % Si (SI2, SI3, and SI4). The red line, SI1*, had a different % Si requirement;

however, this is also the heat that had furnace slag carryover.

Figure 54: Inclusion evolution comparing RSi and Si-bearing heats as well as Sulfur content.

Since sulfur can act as a barrier to Ca transfer from the slag to the steel, knowing the sulfur

percentage in the steel as well as the desulfurization rate is important. As it was shown, a good

desulfurization rate with a high incoming sulfur may not yield the target final % S level that is

needed depending on the processing time allowed. Also, a low incoming sulfur with a poor

removal rate can also be detrimental to the end sulfur target. Another way to evaluate the extent of

slag/metal mixing is the % Mg pickup in the inclusions. This proves to be a good indicator if a

barrier was created that reduces the slag transfer. This barrier is related to the amount of oxygen

in the steel. The higher the oxygen potential the slower the desulfurization reaction. Removing the

oxygen in the steel not only promotes sulfur removal, but also drives the increase in spinel

formation. A typical trend of the % Mg in inclusions throughout a heat can be shown in Figure 55

76

(blue). As the heat progresses % Mg increases due to interaction with the slag as well as the

refractories on the walls of the ladle. After Ca treatment the % Mg declines. This is due to the Ca

eloping the spinel inclusions causing the % Mg to be reduced as well as not being exposed as

frequently to the AFA scan. This trend was very similar on all heats that were analyzed except for

RSI4. The trend for this is also shown Figure 55, grey. As shown in the plots, the trends are very

different. When looking into RSI4, this heat had the least amount of processing time (38 minutes),

and a high incoming sulfur percentage. This heat never received an optimum pre-rinse time due to

the shortened total processing time, and the lessened desulfurization rate. This can also be seen to

alter the inclusion path during processing.

Figure 55: % Mg from AFA inclusion data using robot samples to compare effects from desulfurization rates.

Since sulfur is a main driver of the tail formation in AFA’s, it also plays a role in where

the inclusion distribution finishes. If the initial compositional range is large the final grouping after

modification will match the same large inclusions distribution. This can make it difficult to hit the

ideal target for inclusion modification with Ca-wire when the heat leaves the LMF. If more Ca is

added to the heat to shift the modification group it will only form more CaS not tighten the

0

5

10

15

20

25

S A R C E

Incl

usi

on

% M

g

RSI4

SI2

77

grouping. This is due to the distribution of Ca on the inclusion population remaining. An example

of this is shown in Figure 56.

Figure 56: Comparison of two different heats that resulted in different modification.

FeSi Effects

Si-bearing heats have been discussed and compared to RSi heats throughout the paper. It

has been shown that the effects of the residual components of FeSi, mainly Ca, can play a role in

the LMF process. Pretorius et al, [130] showed that if the FeSi bulk material can be segregated and

analyzed to determine the contained Ca% in the FeSi, then it can be used as a late addition in

replace of some or all of the Ca wire used for treatment. However, in the case where the material

cannot be physically separated the lack of control of the Ca percent in the FeSi delivered material

can cause issues when trying to use it in this way. It was determined that adding FeSi early with a

residual Ca percentage can be advantageous. When added before the end of desulfurization, it can

aid in sulfur removal. It was validated that the formation CaAl inclusions do not form after a FeSi

addition to verify that the early addition was not forming harmful inclusions, and aided in the

desulfurization process. An example of this is shown in Figure 57. After the 317.5 Kg (700 lbs)

addition of FeSi, there was no presence of CaS. A slight Ca pickup was seen in the inclusions

78

composition starting with the ‘S’ sample at 0.35% to the ‘A’ sample at 11.46%, but plateauing at

the ‘R’ sample at 11.08%. However, as shown previously, after the sulfur level is low, a significant

FeSi addition will start the modification tail as also shown in the figure below.

Figure 57: Effects of FeSi additions on AFA results.

It was determined that a Ca restricted FeSi addition early in the heat was also ideal for

cleanliness. From Pitts, et al [57], showed an improved inclusion index from adding FeSi early

(red bars - 9.7 average) versus late (blue bars - 15.3 average), Figure 58. The yellow lines on the

chart depict the average of each data set. The student T-test was used to show the probability of

significance. The results of that test for this relationship was 0.0357%. This equates to a high

likelihood that the practices had a direct effect on the data set.

79

Figure 58: Comparison of the inclusion index versus total weight added of FeSi.

To show the effects of Si on inclusions throughout a heat, a plot of the AFA’s % solid

inclusions throughout the LMF process steps shows a significant difference in the trend between

Si-bearing and RSi in Figure 59. The final step was not plotted due to it being 95+% liquid due to

Ca treatment.

Figure 59: Solid inclusion % for each multi-depth location and process step S-C.

Also, Table 18 shows that the amount of Al consumption for Si-bearing heats was always

lower than RSi heats. The only exception to this was heat SI1* which had a considerable amount

0

5

10

15

20

25

44

5

45

6

49

6

52

4

55

0

55

4

56

4

75

6

78

2

93

6

10

46

45

6

50

4

50

6

52

4

56

2

62

6

77

6

85

0

91

0

10

66

Incl

usi

on

In

dex

Kg of FeSi Addition Early FeSi addition Delayed FeSi addition

75

80

85

90

95

100

S1 S2 S3 A1 A2 A3 R1 R2 R3 C1 C2 C3

RSI3

% Solid

75

80

85

90

95

100

S1 S2 S3 A1 A2 A3 R1 R2 R3 C1 C2 C3

SI3 (Si)

% Solid

80

of furnace slag carryover. Also, the table shows the amount of tap oxygen per heat to show that

the Al consumption does not depend strictly on the tap oxygen ppm level. NOTE: This table does

not include Ca-aluminate additions to the slag which have approximately 30% alumina content.

However, the trend is still the same. Lowering Al consumption not only saves costs, but reduces

the potential for increased alumina formation.

Table 18: Al Consumption per heat.

Heat

Al Consumption

(kg/lbs)

Tap Oxygen

(ppm)

SI3 326.6 / 720 1110

SI2 331.6 / 731 627

SI4 374.6 / 826 993

RSI2 392.4 / 865 996

RSI4 395.5 / 872 966

RSI3 410.5 / 905 916

RSI1 470.4 / 1037 996

RSIOT 491.7 / 1084 961

SI1* 508.0 / 1120 1074

Pre-rinse versus Post Rinse

Previous studies have shown that pre-rinse is the best solution for inclusion removal

due to the inclusions present are still in a solid phase during this time. [28] An additional

study was completed using the FeSi addition trial data. These data sets showed very similar

results. Figure 60, the amount of pre-rinse time is plotted against the inclusion index for both

the early addition heats (red), as well as the delayed addition heats (blue). The R2 value was

enough to further investigate potential correlation. The student T-test was also completed on

this data and showed to have a high probability of significance (2.8%). An additional gain

from adding the FeSi early is the ability for it to help aid in desulfurization. After sulfur is

below its target value, the stir flow rate is reduced which ultimately starts pre-rinse. By adding

the FeSi early allows for additional pre-rinse time resulting in a potentially cleaner heat.

81

Figure 60: Inclusion index versus pre-rinse times. [57]

This study determined that post rinse does not aid in inclusion floatation. Post rinse is

very influential in the inclusion composition homogeneity. It has been verified that a

minimum of 5 minutes is required at NSTI for post rinse to reach homogeneity. This can be

seen in the ternary set shown in Figure 61. The inclusion population after wire treatment does

not decrease. Therefore, it is believed that inclusion do not float into the slag during post rinse

unless the inclusion size is very large. This is due to the inclusion being liquid, and not having

the interfacial energy to attach to the slag layer. This is discussed further in O’Malley’s paper

on inclusion evolution. [131]

0

5

10

15

20

25

30

0 5 10 15 20 25 30

Incl

usi

on

In

dex

Pre-Rinse Time (min)

Early Addition Delayed Addition

82

Figure 61: Post Rinse inclusion evolution.

To compare the results from the FeSi addition trial data to the newest multi-depth data

set, a plot of the inclusion counts on the ‘C’ and ‘E’ samples were used. Figure 62 shows that

the inclusion counts in the sample before Ca treatment (blue) is always lower than the after

wire inclusions counts. It also shows that the inclusions counts for Si-bearing grades have less

inclusions present than RSi in most cases which also matches the data that was initially

completed for the FeSi addition trials. NOTE: The open tap heat (RSIOT) and the furnace

slag carryover heat (SI1*) were excluded.

83

Figure 62: Inclusion count trend for robot samples.

Though the inclusion counts trend upward after wire treatment, the inclusion index is

typically lower. This can be seen in Figure 63, which also shows some variation between RSi

and Si-bearing heats especially early in the LMF processing.

Figure 63: Inclusion index trend.

RSI1 RSI2 RSI3 RSI4 SI2 SI3 SI4

C 563 696 807 442 602 465 318

E 860 882 1225 1409 1236 472 331

0

200

400

600

800

1000

1200

1400

1600

Incl

usi

on C

ount

Inclusions count at C and E

0

5

10

15

20

25

30

S A R C E

Inclusion Index

RSI1

SI4

84

Robot samples versus multi-depth samples

Through the information gathered from the multi-depth sampling device, the opportunity

to see inclusions at different depths simultaneously was achieved. The ability to see multiple

depths inside a ladle during processing has never been accomplished. Therefore, using this

opportunity to gain insight on inclusions populations, paths, compositional changes, slag

emulsification, etc. is exciting.

Since it has been determined that there are differences between RSi and Si-bearing grades,

this discussion will begin with RSi. Figure 64 shows the multi-depth samples from heat RSI3

where ‘R’, ‘C’, and ‘E’ are the processing step designation and ‘1’, ‘2’, and ‘3’ are the different

depths. Samples taken at depth ‘1’ is roughly 0.15 m (6 inches) below the slag layer, sample ‘2’ is

0.76 m (30 inches) below the slag layer, and sample ‘3’ is roughly 1.37 m (54 inches) below the

slag layer. As the figure portrays, there is not a significant difference in the inclusion population

location. However, there is a slight increase in tail formation towards the modification region as

the sample depth increases. In the ‘R’ and ‘C’ samples the inclusions counts diminish as the sample

depth increases. This suggest that the slag layer is the most reactive layer during processing until

Ca treatment. In processing time ‘E’, the inclusion counts do not decline from sample 1 to 3;

however, the modification region is shifted slightly more to the Ca region. This can be seen by the

average blue box located on each AFA, as well as the increase in % Ca in the inclusions

compositional data box also located on each AFA. The multi-depth samples were taken in the

downward flow portion of the stir pattern. To investigate another portion of the ladle, Figure 65

shows the robot samples taken from the same heat. Ironically, the robot samples (taken on the side

of the stir path) show a higher inclusion index, count, and much longer tail than the samples taken

from the same time using the multi-depth on sample sets ‘R’ and ‘C’. However, after Ca treatment,

85

the inclusion index was seen to be much less than the multi-depth samples. The robot sample is

taken at approximately the same depth as sample ‘2’; however, the only sample time that visually

matches the robot and the multi-depth is the final sample ‘E’. The numerical data for this heat can

be found in Table 19. This table shows the differences in the data recorded by the AFA. It can be

seen that the robot sample varies significantly from the multi-depth samples.

Figure 64: Multi-Depth Sampling of RSI3 ‘R’, ‘C’, and ‘E’.

86

Figure 65: Robot Samples of RSI3 ‘R’, ‘C’, and ‘E’

Table 19: AFA data comparison of multi-depth versus Robot samples for processing times ‘R’, ‘C’, and ‘E’. (RSI3)

Looking at Si-bearing grades, it is noticed the same trends that were present for RSi are not

present for Si-bearing. The inclusion counts do not decrease nor does the index. This was seen

across all the Si-bearing heats. The inclusion tail slightly increases as the depth increases, but not

as quantifiable as the RSi heats, i.e. the % Ca does not always increase as the depth increases. It is

noted that on the final sample ‘E’, the deepest sample is the most modified as well as the lowest

count of inclusions than the higher depths.

When comparing the multi-depth samples to the robot samples the differences are

relatively dramatic. The tail formation on the robot samples is much less than all the multi-depth

samples, as well as the inclusion counts, inclusion index, and the % liquid on the ‘R’ and ‘C’

Sample # of Incl % Solid % Liquid Avg Dia Avg Area Total Area Index % Mg % Al % S % Ca % Mn Ca/Al

R1 431 97.7 2.32 1.91 3.28 1415 3.88 23.02 68.42 1.64 4.75 2.14 0.07

R2 388 98.2 1.8 1.94 3.19 1239 3.4 26.05 67.96 0.59 5.02 0.37 0.07

R3 371 97.3 2.7 1.97 3.3 1225 3.36 25.52 67.33 0.88 5.32 0.91 0.08

R Robot 844 83.1 16.9 1.82 2.81 2372 6.53 20.61 63.34 6.39 8.23 1.28 0.13

C1 586 97.4 2.56 2.01 3.59 2105 5.77 21.55 68.43 1.17 5.84 2.95 0.09

C2 585 97.4 2.56 1.9 3.09 1808 4.95 20.35 68.8 1.5 5.47 3.84 0.08

C3 490 95.9 4.08 2.02 3.97 1946 5.33 22.13 67.48 0.98 6.78 2.62 0.1

C Robot 799 89.1 10.9 1.95 3.12 2496 6.84 21.07 64.9 3.6 8.74 1.45 0.13

E1 406 3.94 96.1 1.99 14.2 5759 15.8 6.4 41.63 3.37 47.8 0.26 1.15

E2 818 1.34 98.7 1.92 11.7 9578 26.2 4.81 41 4.28 48.9 0.14 1.19

E3 476 2.52 97.5 2.51 21.4 10210 28 4.8 40.48 2.44 50.39 0.2 1.24

E Robot 1225 0.898 99.1 1.67 3.26 3989 10.9 5.54 36.27 6.09 51.13 0.05 1.42

87

samples. When looking at the final sample ‘E’ on both the robot and the multi-depth samples, the

robot sample is more modified, has less inclusions, a better inclusion index, as well as 0% solid

inclusions. Table 20 shows the AFA numerical data collected for this particular heat.

Figure 66: Multi-Depth Sampling of SI3 ‘R’, ‘C’, and ‘E’.

88

Figure 67: Robot Samples of SI3 ‘R’, ‘C’, and ‘E’.

Table 20: AFA data comparison of multi-depth versus Robot samples for processing times ‘R’, ‘C’, and ‘E’. (SI3)

To compare RSi versus Si-bearing, a few components of the AFA’s were plotted side by

side. Inclusion index is a ratio of the area consumed by inclusions and the area that was scanned

by the AFA program. The index provides a good baseline comparison for AFA’s. Since the index

is used as a factor in rating steel cleanliness at NSTI, Figure 68 plotted the inclusions index versus

all the processing steps and samples for two different heats. As stated previously, the variation on

from the robot samples is seen throughout both sample sets. It is also see that the arrival sample

‘S’ has a significant amount variability on both heats. It is also seen that the Si-bearing heat had

less variability between each sampling time, as well as finished with a lower inclusion index than

the RSi heat.

Sample # of Incl % Solid % Liquid Avg Dia Avg Area Total Area Index % Mg % Al % S % Ca % Mn Ca/Al

R1 431 97.7 2.32 1.91 3.28 1415 3.88 23.02 68.42 1.64 4.75 2.14 0.07

R2 388 98.2 1.8 1.94 3.19 1239 3.4 26.05 67.96 0.59 5.02 0.37 0.07

R3 371 97.3 2.7 1.97 3.3 1225 3.36 25.52 67.33 0.88 5.32 0.91 0.08

R Robot 844 83.1 16.9 1.82 2.81 2372 6.53 20.61 63.34 6.39 8.23 1.28 0.13

C1 586 97.4 2.56 2.01 3.59 2105 5.77 21.55 68.43 1.17 5.84 2.95 0.09

C2 585 97.4 2.56 1.9 3.09 1808 4.95 20.35 68.8 1.5 5.47 3.84 0.08

C3 490 95.9 4.08 2.02 3.97 1946 5.33 22.13 67.48 0.98 6.78 2.62 0.1

C Robot 799 89.1 10.9 1.95 3.12 2496 6.84 21.07 64.9 3.6 8.74 1.45 0.13

E1 406 3.94 96.1 1.99 14.2 5759 15.8 6.4 41.63 3.37 47.8 0.26 1.15

E2 818 1.34 98.7 1.92 11.7 9578 26.2 4.81 41 4.28 48.9 0.14 1.19

E3 476 2.52 97.5 2.51 21.4 10210 28 4.8 40.48 2.44 50.39 0.2 1.24

E Robot 1225 0.898 99.1 1.67 3.26 3989 10.9 5.54 36.27 6.09 51.13 0.05 1.42

89

Figure 68: Inclusion index comparison charts for multi-depth samples and robot samples

Comparison charts were also completed for Ca/Al Ratio (Figure 69), % Al (Figure 70), and

% Ca (Figure 71) for the selected heats. The Ca/Al Ratio showed an increase from the multi-depth

samples to the robot samples on both heats, though the samples leading up to the final sample ‘E’

did not consistently show the same pattern.

Figure 69: Ca/Al Ratio comparison for multi-depth and robot

The charts compiled for the % Al inclusions (Figure 70) did show what was expected in

the general comparison of RSi versus Si-bearing. In that, the Si-bearing heats have less % Al

throughout except the arrival sample which appeared to have the same variation as the RSi heats.

The charts for the % Ca (Figure 71) also showed what was expected, in that the Si-bearing heats

0

5

10

15

20

25

30

35

40

S1

S2

S3

Ro

bo

t 0

A1

A2

A3

Ro

bo

t 1

R1

R2

R3

Ro

bo

t 2

C1

C2

C3

Ro

bo

t 3

E1

E2

E3

Ro

bo

t 4

Incl

usi

on I

ndex

RSI3

0

5

10

15

20

25

30

35

40

S1

S2

S3

S R

obo

t

A1

A2

A3

A R

ob

ot

R1

R2

R3

R R

obo

t

C1

C2

C3

C R

obo

t

E1

E2

E3

E R

ob

ot

Incl

usi

on I

ndex

SI3

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

S1

S2

S3

S R

obo

t

A1

A2

A3

A R

ob

ot

R1

R2

R3

R R

obo

t

C1

C2

C3

C R

obo

t

E1

E2

E3

E R

ob

ot

Ca/

Al

Rat

io

RSI3

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

S1

S2

S3

Ro

bo

t 0

A1

A2

A3

Ro

bo

t 1

R1

R2

R3

Ro

bo

t 2

C1

C2

C3

Ro

bo

t 3

E1

E2

E3

Ro

bo

t 4

Ca/

Al

Rat

io

SI3

90

tend to show more interaction throughout the heat relating to increase % Ca in the inclusion

composition.

Figure 70: AFA inclusion data % Al comparison for robot vs multi-depth samples

Figure 71: AFA inclusion data % Ca comparison for robot vs multi-depth for heat RSI3 (RSi)

SLAG EMULISIFICATON

One of the main objectives for the multi-depth sampling technique was to characterize the

slag/metal reaction zone during the LMF processing of carbon steels. By use of the multi-depth

sampling techniques a better understanding of droplet formation and size in relation to slag

composition and flow rate is achieved. Some models refer to the possibility of slag droplets being

formed; however, the depth of which they are shown to go down to in the models as well as the

0

10

20

30

40

50

60

70

80

90

100

S1

S2

S3

S R

obo

t

A1

A2

A3

A R

ob

ot

R1

R2

R3

R R

obo

t

C1

C2

C3

C R

obo

t

E1

E2

E3

E R

ob

ot

% A

l -

incl

usi

ons

RSI3

0

10

20

30

40

50

60

70

80

90

100

S1

S2

S3

Ro

bo

t 0

A1

A2

A3

Ro

bo

t 1

R1

R2

R3

Ro

bo

t 2

C1

C2

C3

Ro

bo

t 3

E1

E2

E3

Ro

bo

t 4

% A

l -

incl

usi

ons

SI3

0

10

20

30

40

50

60

S1

S2

S3

S R

obo

t

A1

A2

A3

A R

ob

ot

R1

R2

R3

R R

obo

t

C1

C2

C3

C R

obo

t

E1

E2

E3

E R

ob

ot

% C

a

RSI3

0

10

20

30

40

50

60S

1S

2S

3R

obo

t 0

A1

A2

A3

Ro

bo

t 1

R1

R2

R3

Ro

bo

t 2

C1

C2

C3

Ro

bo

t 3

E1

E2

E3

Ro

bo

t 4

% C

a

SI3

91

sizes do not match to the current sampling data. Most of the methods that are looking at

emulsification use water modeling techniques. Therefore, the droplet sizes are measured in

millimeters versus micrometers. Since most of the droplets found in this study that are being

analyzed are not large enough to be seen, SEM analysis was performed in large feature mode on

the steel samples using a large area scan only detecting large inclusion ( > 10 microns). An example

of an AFA that was completed is shown in Figure 72, where only the liquid inclusions greater than

10 μm are kept from the AFA data. These samples show droplets going deeper than 1.5 m below

the slag layer at all processing flowrates. Figure 73 shows examples of some droplets that were

found during these trials. Also, listed with the picture is the diameter as well as the composition.

This analysis will assist significantly in the development of a comprehensive droplet model that

will be used in conjunction with CFD and reaction kinetics models [120] [135] [137] to predict the

desulfurization kinetics in LMF under various process conditions.

Figure 72: Slag Droplet AFA example.

92

Figure 73: Pictures and composition of slag droplets found in AFA’s.

After the data was collected, it was challenging to determine what the cause of these

droplets were, as well as what effect, if any, on the reactions going on in the bath. No direct

correlation was found when looking at the depth versus the number of droplets. Also, no direct

correlation was found when plotting the flow rate versus the number of droplets. This seemed very

unusual. However, when the chart was laid against the flow pattern in the ladle the average number

of droplets found lined up with the change in flow patterns. An example of this is shown in Figure

74 where the blue and grey lines represent the flow rate for the primary and secondary porous

plugs respectively. Since the multi-depth samples were taken from the downward portion of the

flow pattern the conditions for capturing the droplets that were being pulled back down into the

ladle was ideal. As the slag composition changed, the flow pattern affected it differently. However,

regardless of slag composition the droplet formation was seen to increase when the flow rate

increased, and decrease when the flow rate decreased. This is not including the final sample since

there is no distinguishable way to determine if it is a large liquid inclusion or a slag droplet.

93

Figure 74: Emulsification droplets in comparison of stir flow trends.

Another variable that increased the number of droplets present was arc time. Since arcing

time varies dependent on the incoming conditions, processing time, and shop variables,

determining that the number of droplets increase with the amount of arc time is important. Figure

75 shows that the arc time is related to the droplets that are seen in the samples. After conducting

the F-test on this data set it was seen to have a significance factor of 0.165; however, after

excluding RSIOT and RSI3, which were outliers in the data set, the F-test significance dropped to

0.104 with a p-value of 0.0135 which validates a correlation between arc time and droplet count.

This could also be related to the slag being superheated while arcing, which will make the slag

layer more fluid. If the slag layer is more fluid the opportunity to produce more droplets will be

present.

0

10

20

30

40

50

60

70

80

90

100

17 17 17 45 45 45 14 14 14 25 25 25 15 15 15

Dro

ple

ts F

ound

Flow Rate (scfm)

RSI3

94

Figure 75: Effects of Arcing time on droplet formation.

Since AFA methods were used to measure the droplets within the sample, the ability to use

the inclusion index was available. Using the inclusion index, it was seen that there was a correlation

between the slag droplet index and the desulfurization rate as well as the total sulfur removed. The

plot of the droplet inclusion index versus the desulfurization rate is shown in Figure 76. This shows

that the desulfurization rate is higher for a smaller inclusion index. However in Figure 77, it is

shown that the total sulfur removed followed an opposite trend. It shows that the larger the

inclusion index the more sulfur is removed from the heat. These variables were both validated by

conducting F-tests. The F-test significance for the desulfurization rate versus the droplet inclusion

index was 0.0075 showing high correlation with a p-value of 0.00019. For the total sulfur removed

versus the droplet inclusion index the F-test significance was 0.043 with a p-value of 0.005.

Typically, liquid inclusions early in the heat would be considered harmful since most liquid

inclusions do not float out into the slag. The slag droplets found do appear could have some

positive influence on the total sulfur removed in the heat. However, with the desulfurization rate

showing an opposite trend, the potential correlation of initial stir rate versus the droplet inclusion

R² = 0.5738

0

5

10

15

20

25

0 2 4 6 8 10 12 14 16 18 20

Dro

ple

t C

ount

Arc Time (minutes)

Arc Time

95

index was evaluated. This is shown in Figure 78. Though the R2 value is not quite as high as the

desulfurization rate correlation, the F-test showed a significance of 0.069 along with a p-value of

0.00002. Since the stir rate changes throughout a heat, the R2 value could have been influenced by

the change in the stir rate throughout the data set sampling time. It appears the opposite trend in

the desulfurization rate may be influenced by the stir flow rate.

Figure 76: Slag Droplet Inclusion Index effects on desulfurization rate and total sulfur removal. (RSIOT excluded)

R² = 0.7899

0

2

4

6

8

10

12

14

16

18

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20

Deo

ple

t In

clusi

on I

ndex

Desuflurization Rate

Droplet Inclusion Index vs Desulfurization Rate

96

Figure 77: Droplet inclusion index versus total sulfur removed.

Figure 78: Droplet inclusion index versus initial stir rate.

Comparing slag droplets present in RSi versus Si-bearing heats showed a significant

difference. For RSi heats large amounts of slag droplets (> 50) were found on all but one sampling

time. However, for the Si-bearing heat at no point in time during the process did the slag droplet

R² = 0.5932

0

2

4

6

8

10

12

14

16

18

0.010 0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050

Dro

ple

t In

clusi

on I

ndex

Total Sulfur Removed

Droplet Inclusion Index vs Total Sulfur Removed

R² = 0.5144

0.00

2.00

4.00

6.00

8.00

10.00

12.00

14.00

16.00

18.00

1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4 1.45 1.5

Dro

ple

t In

clusi

on I

ndex

Initial Stir Rate (m3/s)

Droplet Inclusion Index vs Initial Stir Rate

97

count even exceed 30. Variability between sampling depths as well as location shows to be

unpredictable at this point, and it will be continued to be investigated.

Figure 79: Number of slag droplets per sample comparing RSi and Si-bearing.

Similar trends were found when comparing the inclusion index of RSi versus Si-bearing

heats. It appears to be even more significant than droplet count since the inclusion index

incorporates the sizes of the droplets found. This could potentially explain the very low sulfurs at

the end of the heat on RSI3 even though the heat was RSi. These heats both ended the LMF process

with very low sulfur concentration. The RSI3 heat finished processing with a 0.0016 wt % S, and

the SI3 heat finished processing with a 0.0014 wt % S.

Figure 80: Inclusion index per sample comparing RSi and Si-bearing.

0

10

20

30

40

50

60

70

80

90

S1

S2

S3

S R

obo

t

A1

A2

A3

A R

ob

ot

R1

R2

R3

R R

obo

t

C1

C2

C3

C R

obo

t

E1

E2

E3

E R

ob

ot

# s

lag d

rop

lets

RSI3

0

10

20

30

40

50

60

70

80

90

S1

S2

S3

S R

obo

t

A2

A3

A R

ob

ot

R1

R2

R3

R R

obo

t

C2

C3

C R

obo

t

E1

E2

E3

E R

ob

ot

# s

lag d

rop

lets

SI3

0

50

100

150

200

250

S1

S2

S3

S R

obo

t

A1

A2

A3

A R

ob

ot

R1

R2

R3

R R

obo

t

C1

C2

C3

C R

obo

t

E1

E2

E3

E R

ob

ot

AF

A I

ndex

RSI3

0

50

100

150

200

250

S1

S2

S3

S R

obo

t

A2

A3

A R

ob

ot

R1

R2

R3

R R

obo

t

C2

C3

C R

obo

t

E1

E2

E3

E R

ob

ot

AF

A I

ndex

SI3

98

It appears that the formation of the droplets is potentially different for RSi than Si-bearing.

The number of droplets for Si-bearing heats was typically less than the number of droplets found

for RSi heats. The effects of emulsification on desulfurization and steel cleanliness are to be

investigated further.

CALCIUM TREATMENT

Ca treatment is an essential part of clean steel practices. The timing of the treatment as well

as the amount can alter the inclusions in the liquid steel positively or negatively. Inevitably,

inclusions will always be present. The goal is to control the composition, morphology, and size to

improve castability and to not harm the physical properties desired in the finished product. Each

heat produced is slightly different; therefore, logically Ca treatment would not be the same for

every heat. Until recently, that was the condition at NSTI. Each heat received the same Ca

treatment regardless of the condition of the heat. Examples of some heats are shown below, Figure

81.

99

Figure 81: Four different heats with all the same wire addition. [57]

Using a large data set, a mathematical model was developed using NSTI’s data, and with

assistance from Nucor’s Steel Making Technology Manager that merged the effects of three main

components: time, sulfur, and silicon. This model calculates the amount of Ca wire needed to

modify the heat based on the three main components listed (patent pending). A screen shot of the

100

model at NSTI is shown below in Figure 82. As discussed previously, each of these parameters

play a major role in the evolution of inclusions.

Figure 82: Calcium wire model program at NSTI.

This Ca addition model was verified using AFA data, as well as a live feedback approach.

The live feedback approach was tracking the stopper rod movement at the caster. Though the LMF

is not always the culprit to stopper rod movement, it definitely can have a negative effect. The

stopper rod function is to control the flow of steel into the mold. This flow needs to be very

consistent to not produce turbulence or voids in the flow of steel from the tundish to the mold. The

stopper rod trend is desired to gradually move downward, example Figure 83 (left). This is a sign

of slight wear on the rod, and has been accepted as an ideal practice. However, when the rod has

to move upward, it is due to something clinging to the rod which restricts the flow. The rod then

has to move upward to continue to allow the set flow rate into the mold. This causes the stopper

rod trace to go up, example Figure 83 (right). The stopper rod data is recorded on a per heat basis.

Since the installation of the mathematical model the stopper rod upward movement has reduced

by 25%. This model has been very successful with reducing positive stopper rod movement, as

well as controlling modification of inclusions consistently at the LMF. Some examples of stopper

rod trends of heats that were ran before and after the model are below in Figure 83. Figure 84 shows

101

the stopper rod schematic [73] with a recent stopper rod sequence trend. The stopper rod trend is

in blue, and the ladle exchanges are shown above it in yellow.

(a) (b)

Figure 83: Stopper Rod traces. Ideal (a) Clogging (b) [57]

Figure 84: Stopper rod schematic with ideal stopper rod trend (right).

Figure 85 shows a histogram of the stopper deviation (throughout the heat) for 2900 heats

before and after the model was implemented. The data shows that the stopper rod deviation

improved after the model was implemented. This shows that there was a 25% reduction in heats

that caused the rod to rise after the model was implemented (rise greater than 1%).

102

Figure 85: Stopper Rod Deviation before and after model.

The Ca mathematical model is currently being used at NSTI, and it has been proven to be

85+% accurate on determining the amount of wire needed to make the heat successfully leaving

the LMF. This model is only limited to the amount of silicon contained, the amount of sulfur, and

the time at the LMF. However, if any anomaly occurs during processing such as furnace slag

carryover, inadequate processing time, heat being sent back from the caster, etc. this model is not

always adequate. An example is shown in Figure 86. This particular heat had several process

variables that were not ideal, but was sent to the caster against better judgment. This heat clogged

off due to the lack of wire with the combination of being excessively dirty.

-5 or

less(-4)-(-5) (-3)-(-4) (-2)-(-3) (-1)-(-2) 0-(-1) 0-1 1-2 2-3 3-4 4-5

After 18 6 7 27 105 751 1603 294 49 13 11

Before 16 5 2 36 172 706 1472 369 80 18 5

0

200

400

600

800

1000

1200

1400

1600

1800

Fre

quen

cy

Stopper Rod Deviation (mm)

103

Figure 86: Example of the effects of late additions in conjunction with poor stir conditions.

Since AFA takes well over two hours to complete, a real-time model with rapid process

feedback is desired. With a PDA-OES technique, all production situations could be predicted with

a much higher accuracy since it is a measurement, and not a semi-empirical calculation. Therefore

while using the mathematical model as a baseline, an empirical LMF Ca injection model using

SparkDAT (PDA-OES) techniques for rapid inclusions analysis was created. Numerous papers

have been written in determining if a PDA-OES technique could be a useful tool in an operational

setting. However, none of the publications have used this tool in production as a Ca weight

determination method. While running the methods simultaneously, regression analysis was

completed on the SparkDAT data correlating it to the mathematical model data. The equation that

was formulated from the analysis was added to the calculation completed by the spectrometer. The

data was captured for over six months and compared. Both the wire mathematical model and the

SparkDAT model have been edited to better align with production variables. SparkDAT has been

deemed to be an accurate source of information for wire footage. Now with the use of this live

tool, the ability for accurate wire determination can be done within approximately one minute.

104

To show the variation in the SparkDat analysis and the wire model that was developed,

Figure 87 is a histogram that shows the percentage of samples test out of approximately 3200

samples that had a difference of wire footage request from SparkDat versus the wire model. When

looking at the differences in wire, it is hard to believe that 22.9 m (75 ft) of difference is not a lot.

The previous practice at NSTI was to add 152.5 m (500 ft) of wire per heat. This equates to roughly

17.2 Kg (38 lbs) of Ca (CalSil wire is 0.11 Kg/m or 0.0765 lbs/ft contain Ca). This means that

22.9 m (75 ft) of wire is roughly 2.3 kg (5 lbs) of Ca going into 125 Tons of steel. Out of

approximately 3200 heats over 80% of the heats showed the model and SparkDat to be within 22.9

m (75 ft) of each other. The heats that are outside these ranges are being investigated. The root

cause of the differences so far have found to be two-fold. The wire model needs an upper limit to

be introduced. However, the biggest difference is typically related to sample preparation. Means

of improving sample preparation that is efficient enough for operational interest is being

investigated.

Figure 87: Histogram of the footage difference in SparkDat versus the wire model.

0.5% 0.9%

2.9%

7.0%

14.4%

22.3%

20.1%

13.6%

8.4%9.8%

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

% o

f sa

mp

les

wit

hin

ran

ge

Feet of wire difference between wire addition vs Sparkdat

105

MANGANESE MICRO-SEGREGATION IN SAMPLE

Determination of the sample thickness and its effects on the quality of inclusions

characterization (MnS, CaS) by using a CFD model for an industrial sampler was investigated. It

is known that Mn is an element that segregates in steel, as well as it forms MnS during the

solidification process especially at longer solidification time and higher concentrations of sulfur.

By using CFD modeling (Ansys’s Fluent), the goal was to determine the appropriate sample

thickness needed to minimize the resulting Mn segregation for production samples, as well as to

compare the current two sampling techniques to the resulting data to determine sample thickness

segregation effects.

The initial simulation showed that the sample solidified at approximately 4 seconds, Figure

88a. This resulted in no macro segregation of Mn in the sample. The temperature profile is shown

in, Figure 88b, after the 4-seconds interval. The sample solidified as expected, and with a valid

time frame for this application. These results were also similar to the results found in Zhang et al.

[132]

(a) (b)

Figure 88: Solid fraction at 4 seconds. (a) Temperature profile at 4 seconds. (b)

106

Since Mn variation throughout the thickness of a sample occurs in the production

application as shown in Figure 89, the need to understand how much micro-segregation is present,

as well as if the segregation has an effect of the compositional changes. To accomplish this, an

argon purged sample was cut in half, schematic shown in Figure 90. The inside surface was

polished and etched to gather grain size information through various locations throughout the

sample. Grain size analysis on the cross section of the sample is shown in Figure 91.

(a) (b)

Figure 89: Mn concentration change throughout sample depth for LMF (a) and Caster (b) samples.

Figure 90: Schematic of sample.

0.96

0.97

0.98

0.99

1

1.01

0 2 4 6

% M

n

HSLA

LMF Caster

0.82

0.83

0.84

0.85

0.86

0 2 4 6

% M

n

A36

LMF Caster

107

Figure 91: Grain size analysis at various locations throughout the cross section of the sample.

The average grain size is approximately 70 microns in diameter. Using the grain size as

previously discussed to determine the α and Ω values, Equation 12 and 13. These were then used

to solve for the solid concentration (Cs), Equation 14. The data for the micro-segregation

calculations is listed in Table 21. Where Ds is the diffusivity of sulfur, tf is the time to solidify, λ

is the average grain size, α and Ω are calculated from Equations X and X as a constant, k is the

partition coefficient, and C0 is the starting composition of the element (Mn).

Table 21: Data for micro segregation calculations. [128]

Ds tf λ α Ω k (Mn) C0 (Mn)

m2/s s m %

2.47E-13 5 7.00E-05 -1.01E-03 -1.01E-03 0.77 1.1

The results were plotted in Figure 92. This showed the differences in using the Scheil

equation and Clyne equation differences. [90] The methods for this data set were very similar for

both Mn and Sulfur. This is not unexpected due to the low amounts of sulfur present in the steels

108

produced today. Figure 93 shows a comparison of sulfur concentration over time correlated to the

number of actual MnS inclusions. A very limited amount of MnS inclusions were found in the

current project data, as well as an even smaller presence of CaS inclusions.

Figure 92: Micro-segregation of Mn and S comparing methods.

Figure 93: Inclusion count of MnS versus time and sulfur content in the steel. [54]

0.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0.0030

0.0035

0.0040

0.0045

0.0050

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

CS

Fraction Solid

Coarsening Segregation

Mn

S Clyne

S Scheil

109

More analysis should be completed in the future including SEM analysis in the grain

boundaries to investigate potential MnS in the sample, as well as the repeatability in the

spectrometer results. Overall, it was determined that the Mn variation seen in the current method

sampling technique is not related to micro- or macro-segregation, and it is more related to the

variation within the spectrometer along with the sample preparation variation.

110

CONCLUSIONS

The functions of the slag at the LMF are critical to producing a high end product with

higher cleanliness. Slag evolves throughout the LMF process to be able to adapt to its changing

functions throughout the heat while remaining as an insulator and as a barrier to the atmosphere.

Initially in the heat, the slag fluidity is important for desulfurization; however, as the heat

progresses the removal of alumina based inclusions becomes more of a focus. This can be seen

through the slag sampling technique developed. The three-prong device to take samples at the eye

provided insight on the reactivity of oxygen at the exposed point in the ladle, furthermore deeming

it more reliable to take samples away from the stir eye.

LMF slag is also dependent on the amount of EAF slag carryover that takes place. It was

seen how dramatically excessive slag carryover can impact the entire LMF processing sequence.

It was shown how it affects the incoming slag composition, the desulfurization, the total sulfur

removal, the Al consumption, the inclusion counts, as well as on inclusion modification. Though

it was also seen that with enough processing time, this anomaly was overcome.

It is known that Si effects the fluidity of the slag as well reoxidation of Al in the steel bath.

This in turn effects the inclusions captured in the slag, the desulfurization rate, and the morphology

of the inclusions within the molten steel. When comparing the inclusion evolution of Si-bearing

grades versus RSi grades several differences arose. RSi heats typically have a very tight cluster of

inclusions; however, it was noticed that when RSi grades dropped below a 0.003% S a

compositional tail formation occurred. This is similar to what is seen in Si-bearing grades without

necessarily having the low %S present. The inclusion paths were plotted and showed significant

111

differences based off of the sulfur content. Si-bearing grades are known to have a wider

compositional range of inclusions or a longer tail formation. Since Si-bearing heats have a

tendency to desulfurize faster as well as to lower levels of sulfur, longer tail formations are

expected on these grades. However looking further into the inclusion evolution on Si-bearing

grades, it was noticed that the compositional range/tail formation is more closely correlated to the

FeSi addition after desulfurization due to the residual Ca % contained in the FeSi. It was noted that

the tail formation was heavily effected when both conditions, %S > 0.003 and FeSi was added

after desulfurization, were present. It has been seen that the longer the processing time after the

sulfur reaches below 0.003 wt %, the more of a tail formation / modified the heat becomes

regardless of grade type. FactSage shows fast Mg and Ca pickup in steel from slag. It is seen that

the Mg inclusion composition will increase overtime; however, it was found that the Ca pickup is

directly related to the S% and Si% in the steel. Below a certain level the sulfur stops acting as a

barrier, and drives the Ca to be able to come out of the slag for inclusion modification. Si-bearing

heats typically decrease sulfur faster; therefore, this effect is seen more vividly on Si-bearing heats.

It was also seen that the pre-rinse stir always had a lower inclusion count than the post rinse

stir. This in turn confirms that the post rinse portion is not efficient for inclusion removal, but

rather for homogenization of the heat.

One of the major insights from this study was the differences between the robot samples

taken versus the multi-depth samples. The Si-bearing heats had very little change from sampling

time and location with respect to inclusion populations and composition. However, it was seen that

the inclusion population and compositional distribution was affected by location on RSi heats. It

was also noticed that the Si-bearing heats typically had a better inclusion index/cleanliness as well

as lower inclusions counts and fewer slag droplets. The variation of inclusions from different

112

locations in the ladle needs to be investigated further in relation to flow rates and sampling

locations.

The multi-depth sampling technique provided information on slag droplets that has not

been seen before. Slag emulsification has been investigated previously, but not on a micro-scale.

These slag droplets showed a correlation to the total sulfur removal, and were more predominant

in the RSi grades. Variables that showed correlation in predicting the droplet amounts: included

stir rate, % Si, and arc time.

A significant contribution from this study was the Ca treatment model. This model has

improved steel cleanliness, saved costs, and increased operational consistency at NSTI. The

model’s ability to predict a more accurate wire footage has proved to be successful as well. With

its success, the ability to implement the PDA-OES method through SparkDat is possible. This

methods alone has reduced the amount of quality issues due to inclusion related events

dramatically at NSTI.

Overall, the LMF process has many transient kinetic components. Looking strictly at the

thermodynamic components is not sufficient for understanding all of the reactions taking place

during LMF processing. The more the LMF process is investigated the more that is understood

about desulfurization, Si-bearing versus RSi, effects of ultra-low sulfur (%S < 0.003), slag

droplets, inclusion formations, etc. By incorporating these variables into improved dynamic

models will help make a more competitive product and cleaner steels.

113

CONTRIBUTION OF THIS STUDY

Slag composition

The evolution of slag composition was studied. It was determined that the ideal sampling

location for slag is located on the opposite side of the ladle from the stir eyes. It was also seen that

excessive furnace slag carryover can negatively affect the desulfurization rate, total sulfur removal,

as well as increases aluminum and lime consumption which adds to the cost of the process. It was

also seen that the effects of excessive slag carryover can be overcome with more processing time.

Desulfurization

A new equation was developed to calculate final sulfur content in the steel by incorporating

%Si in the steel. This equation had a standard deviation of 0.013% on all the heats run when

compared against the actual final sulfur content. Whereas, the equation developed by Turkdogan

had a 0.012% standard deviation, however it was unable to predict any of the Si-bearing heats.

Inclusion Evolution and Distribution

It was seen that sulfur had a dramatic effect on inclusion evolution for Restricted Silicon

grades. Whereas for Si-bearing grades, the FeSi addition timing played a bigger role on the AFA

tail formation in the inclusion distribution. It was seen that the barrier to the slag created by Sulfur

concentration in the steel is diminished when the sulfur level is below 0.003%. This allows the

Calcium in the slag to interact with the solid inclusions causing a tail formation/partially modified

inclusions. This was seen for both grade types.

114

Calcium Treatment

Due to the increased demand for low sulfurs, and the challenges faced with targeting the

exact calcium addition, a Ca treatment model was developed (patent pending). This model is

currently being used at NSTI with the ongoing development of the PDA-OES method, SparkDAT.

This is to incorporate a real-time feedback for exact calcium needed for inclusion modification.

Multi-depth Sampling

A multi-depth argon purged sampling device was created (patent pending), to be able to

simultaneously take samples during various processing times. Through AFA data, the samples

from this device showed the effects of argon flow on inclusions.

Emulsification

By use of the multi-depth sampling device, micro-scale emulsification was captured. It was

seen that the flow rate, arc time, as well as Si content in the steel had an effect on the presence of

these slag droplets. It was also seen that the increase in slag droplets correlates to an increased

total sulfur removal from the heat.

115

FUTURE WORK

Further development is needed to improve the accuracy on process anomalies for the PDA-

OES SparkDAT Ca-treatment method. Also, a deeper understanding on the formation and control

of the slag droplets on the sulfur removal process. The ability to quantify flow rate when comparing

data from different ladles would also be useful in further understanding the droplet formation.

Also, the data for this study will be used to further develop a 3D CFD model to aid in predicting

LMF processing steps.

116

REFERENCES

[1] R. Fruehan, Ladle Metallurgy Principles and Practices, Chelsa: Iron & Steel Society, 1985.

[2] L. Zhang and B. Thomas, "Inclusions in Continuous Casting of Steel," in XXIV National

Steelmaking Symposium, Morelia, Mexico, 2003.

[3] E. Pretorius, H. Oltmann and B. Schart, "An Overview of Steel Cleanliness From an Industry

Perspective," AIST Transactions, vol. 12, no. 3, 2015.

[4] P. Kaushik and H. Yin, "Thermodynamics, Engineering, and Characterization of Inclusions in

Advanced High Strength Steels," in 8th International Conference on Clean Steel, Budapest, 2012.

[5] W. Tiekink, A. Overbosch, P. ten Bras, J. de Weerdt, C. van Hoek and R. Kooter, "Aspects of

Traces in Alloys and Some Effects on Steelmaking and Casting," in 8th International Conference

on Clean Steel, Budapest, Hungary, 2012.

[6] E. Damm, P. Glaws, D. Gynther and D. Quikel, "Clean Steels Improve Gear Fatigue Properties,"

FORGEmag.com, vol. April , pp. 32-38, 2016.

[7] K. Yamamoto, H. Yamamura, D. Maeda and Y. Suwa, "The Effects of Nonmetallic Inclusions on

Local Ductility," in AISTech Conference Proceedings, Indianapolis, 2011.

[8] P. Kaushik, M. Lowry, H. Yin and H. Pielet, "Inclusion Characterisation for Clean Steelmaking

and Quality Control," Ironmaking and Steelmaking, vol. 39, no. 4, pp. 284-300, 2012.

[9] V. Singh, "Inclusion Modification in Steel Castings Using Automated Inclusion Analysis, A

Masters Thesis," Missouri S & T, Rolla, 2009.

[10] M. Cornille, M. Schad and S. Chakraborty, "A new technique for the characterization of sliver

defect orgins in the steel industry," in 8th International Conference on Clean Steel, Budapest,

Hungary, 2012.

[11] Y. Kang, B. Sahebkar, P. Scheller, K. Morita and D. Sichen, "Observation on Physical Growth of

Nonmetallic Inclusion in Liquid Steel During Ladle Treatment," The Minerals, Metals & Materials

Society and ASM International, vol. 42B, no. June, pp. 522-534, 2011.

[12] R. Morales and L. Garcia-Demedices, "Anticlogging Stopper Rod," in AISTech Proceedings -

Volume I, Indianapolis, 2011.

[13] R. Liu, J. Sengupta, D. Crosbie, M. Yavuz and B. Thomas, "Effects of Stopper Rod Movement on

Mold Fliud Flow at ArcelorMittal Dofasco's No. 1 Continuous Caster," in AISTech Conference

Proceedings, Indianapolis, 2011.

117

[14] R. Rastogi and A. Cramb, "Inclusion Formation and Agglomeration in Aluminum Killed Steels,"

in Ironmaking Conference Proceedings, Baltimore, 2001.

[15] B. Harcsik, P. Tardy, G. Karoly and R. Jozsa, "Investigation of Deposits on Submerged Entry

Nozzles," in 8th International Conference on Clean Steel, Budapest, Hungary, 2012.

[16] L. Zhang, B. Thomas, X. Wang and K. Cai, "Evaluation and Control of Steel Cleanliness -

Review," in 85th Steelmaking Conference Proceedings, Warrendale , 2002.

[17] H. Weerts, D. Senk, L. Sattler, C. Beiler and S. Janosch, "On-Line Determination of Steel

Cleanliness During Ladle Metallurgy," in 8th International Conference on Clean Steel, Budapest,

Hungary, 2012.

[18] S. Abraham, R. Bodnar and J. Raines, "Inclusion Engineering and the Metallurgy of Calcium

Treatment," Iron and Steel Technology, no. July, pp. 57-69, 2014.

[19] A. Wilson, "Clean Steel Technology - Fundamental to the Development of High Performance

Steels," in Advances in the Production and Use of Steel with Improved Internal Cleanliness, vol.

ASTM STP 1361, West Conshohocken, PA, American Society for Testing and Materials, 1999,

pp. 73-88.

[20] A. Kaijalainen, P. Karjalainen, D. Porter, P. Duikkanen, J. Komi, V. Kesti and T. Saarinen, "Effect

of Inclusions on the Properties of Ultra-High Strength Low-Alloy Steel with a Martensitic-Bainitic

Microstructure," in 8th Internaional Conference on Clean Steel , Budapest, Hungary, 2012.

[21] S. Nafisi, J. Jordan, C. Souza, L. Collins and T. Drake, "A Study of Ca-Modification Process in

Al-Killed Steels," in AISTech Conference Proceedings, Atlanta, 2012.

[22] W. Wilson and A. McLean, Desulfurization of Iron and Steel and Sulfide Shape Control,

Warrendale, PA: Iron & Steel Society of AIME, 1980.

[23] O. Adaba, M. Harris, R. O'Malley, S. Lekakh, V. Richards and N. Sutcliffe, "An SEM/EDS

Statistical Study of the Effect of Mini-Mill Practices on the Inclusion Population in Liquid Steel,"

in 9th International Conference and Exhibition on Clean Steel, Budapest, 2015.

[24] M. Jiang, X. Wang and W. Wang, "Study on Refining Slags Targeting High Cleanliness and

Lower Melting Temperature Inclusions in Al-Killed Steels," Iron and Steelmaking, vol. 39, no. 1,

pp. 20-25, 2012.

[25] C. Tse, S. Lee, S. Sridhar and A. Cramb, "In Situ Observations Relevant to Clean Steel:

Dissolution of Alumina Particles in Slags," in Steelmaking Conference Proceedings, Pittsburgh,

2000 .

[26] S. Michelic, M. Hartl and C. Bernhard, "Thermodynamic and Experimental Study on the

Modification of Nonmetallic Inclusions Through the Contact with CaO-Al2O3-MgO Slags," in

AISTech Conference Proceedings, Indianapolis, 2011.

[27] H. Gaye, "Inclusion Formation in Steels," in ASM Casting Volume, Pittsburgh, PA, The AISE

Steel Foundation, 2003, pp. 1-20.

118

[28] E. Pretorius, H. Oltmann and B. Schart, "An Overview of Steel Cleanliness from an Industry

Perspective," in AISTech, Pittsburgh, 2013.

[29] J. Lee and K. Morita, "Effect of Surface Active Element Sulfur on the Interfacial Energies

Between Gas, Solid and Molten Fe-C-S Alloys," in 85th Steelmaking Conference Proceedings,

Nashville, TN, 2002.

[30] N. El-Kaddah and J. Szekely, "Mathematical Model for Desulphurization Kinetics in Argon-

Stirred Ladles," Ironmaking and Steelmaking , vol. No. 6, pp. 269-278, 1981.

[31] K. Graham and G. Irons, "Toward Integrated Ladle Metallurgy Control," AIST Transactions, vol.

6, no. 1, pp. 164-174, 2009.

[32] E. Ramstrom, Mass Transfer and Slag - Metal Reaction in Ladle Refining - a CFD Approach

(Licentiate Thesis), Stockholm: KTH School of Industrial Engineering and Management - Division

of Micro-Modeling, 2009.

[33] D. Roy, "Effect of Silicon on Desulfurization of Al-Killed Steels," Carneige Mellon University ,

Pittsburgh, 2012.

[34] R. Young, J. Duffy, G. Hassail and Z. Xu, "Use of Optical Basicity Concept for Determining

Phosphorus and Sulfur Slag-Metal Partitions," Ironmaking and Steelmaking, vol. 19, pp. 201-219,

1992.

[35] S. Story and R. Asfahani, "Control of Ca-Containing Inclusions in Al-Killed Steel Grades," in

AISTech, Pittsburgh, 2013.

[36] S. Story, N. Gupta and M. Molnar, "Effect of Oxygen Sources on Steel Cleanliness in Ti-Stablized

Ultra-Low Carbon Steels," in 8th International Conference on Clean Steel, Budapest, Hungary,

2012.

[37] J. Mendez, A. Gomez, C. Capurro, R. Donayo and C. Cicutti, "Effect of Process Conditions on the

Evolution of MgO Content of Inclusions During the Production of Calcium Treated, Aluminum

Killed Steels," in 8th International Conference on Clean Steel, Budapest, Hungary, 2012.

[38] K. Steneholm, M. Andersson, P. Jonsson, A. Tilliander and M. Nzotta, "Removal of Hydrogen,

Nitrogen, and Sulfur from Tool Steel during Vacuum Degassing," in 7th International Conference

on Clean Steel, Hungary, 2007.

[39] J. Sosinsky and I. Sommerville, "The Composition and Temperature Dependence of the Sulfide

Capacity of Metallurigcal Slags," Metallurgical Transactions B, vol. 17, no. 2, pp. 331-337, 1986.

[40] E. Pretorius, "Fundamentals of EAF and Ladle Slags and Ladle Refining Principles," [Online].

Available: http://etech.lwbref.com/Downloads/Theory [Accessed May 2016].

[41] A. Senguttuvan and G. Irons, "Model Studies on Slag Metal Entrainment in Gas-Stirred Ladles,"

AIST Transactions, vol. 11, no. 1, pp. 230-241, 2014.

119

[42] G. Kor and P. Glaws, "Ladle Refining and Vacuum Degassing," in Making, Shaping, and Treating

of Steel, Pittsburgh, The AISE Steel Foundation, 1998, pp. 662-713.

[43] E. Turkdogan and R. Fruehan, "Fundamentals of Iron and Steelmaking," in Making, Shaping, &

Treating of Steel, Pittsburgh, PA, AISE Steel Foundation, 1998, p. Steelmaking and Refining

Volume.

[44] E. Turkdogan, Fundamentals of Steelmaking, London: The Institute of Materials, 1996.

[45] W. Tiekink, "Initial Alumina Formation: Some Additional Throughts Based on Labratrory and

Steel Plant Findings," AIST Transactions, vol. 11, no. 6, pp. 294-298, 2014.

[46] S. Ueda, H. Shi and A. Cramb, "Observation of the Effect of Droplet Size on the Contact Angle

Between Liquid Iron and a Single Crystal, Alumina Substrate at 1873KI," in ISSTech Conference

Proceedings, Indianapolis, 2003.

[47] W. Tiekink, A. Overbosch, P. ten Bras, J. deWeerdt, C. van Hoek and R. Kooter, "Aspects of

Traces in Alloys and Some Effects on Steelmaking and Casting," in 8th International Conference

on Clean Steel , Budapest, Hungary, 2012.

[48] J. Wang, L. Zhang, S. Uang, Y. Chen and J. Li, "Interaction Between Molten Steel, Lining

Refractory, and Slag Phase," in 8th International Conference on Clean Steel, Budapest, Hungary,

2012.

[49] N. Verma, C. Pistorius, R. Fruehan, M. Potter, H. Oltmann and E. Pretorius, "Calcium

Modification of Spinel Inclusions in Aluminum-Killed Steel: Reaction Steps," Metallurgical and

Materials Transactions B, vol. 43B, no. August, pp. 830-840, 2012.

[50] S. Story, N. Gupta, G. Casuccio and M. Potter, "Analysis of Ti, Mg, and Zr Bearing Inclusions in

Ultra-Low and Low Carbon Steels," in AISTech Conference Proceedings, Pittsburgh, 2010.

[51] C. Seo, S. Kim, S. Jo, M. Suk and S. Byun, "Modification and Minization of Spinel

(Al2O3.xMgO) Inclusions Formed in Ti-Added Steel Mills," Metallurgical and Materials

Transactions B, vol. 41B, no. August, pp. 790-797, 2010.

[52] T. Dubberstein, O. Volkova, S. Lachmann and P. Scheller, "Invesitgation on Steel Cleanliness of a

Heat Resistant Steel Grade - Characterization of Non-Metallic Inclusion in Ladle Treatment," in

8th International Conference on Clean Steel, Budapest, Hungary, 2012.

[53] M. Harris, V. Richards, R. O'Malley and Lekakh, "Evolution of Non-Metallic Inclusions in

Foundry Steel Casting Processes," in SFAS Technical and Operations Conference , 2015.

[54] S. Yang, L. Zhang, Q. Wang and Y. Chen, "Transient Phenomena of Inclusions in Alloy Steels

During RH-VD-CC Process," in AISTech Conference Proceedings, Atlanta, 2012.

[55] S. Yang, L. Zhang, Q. Wang and Y. Chen, "Variation of Non-Metallic Inclusions in Molten Steel

During Refining Processes," in 8th International Conference on Clean Steel , Budapest, 2012.

120

[56] O. Wijk and V. Brabie, "The Purity of Ferrosilicon and Its Influence on Inclusion Cleanliness of

Steel," ISIJ Internationl , vol. 36, no. Supplement, pp. 132-135, 1996.

[57] S. Story, G. Goldsmith, R. Fruehan, G. Casuccio, M. Potter and D. Williams, "Study of Casting

Issues using Rapid Inclusion Identification and Analysis," in AISTech Conference Proceedings,

Cleveland, 2006.

[58] A. Pitts, B. Tate and L. Nastac, "FeSi Residuals and Their Effects on Steel Cleanliness," Iron &

Steel Technology, vol. 12, no. 10, pp. 93-101, 2015.

[59] D. Roy, P. Pistorius and R. Fruehan, "Effect of Silicon on the Desulfurization of Al-killed Steels:

Part II. Experimental Results and Plant Trials," Metallurgical and Materials Transactions B, 2013.

[60] D. Roy, C. Pistorius and R. Fruehan, "Effect of Silicon on the Desulfurization of Al-Killed Steels:

Part I. Mathematical Model," Metallurgical and Materials Transactions B, vol. 44B, no. October,

pp. 1086-1094, 2013.

[61] P. Iwamasa and R. Fruehan, "A Kinetic Model Demonstrating the Effect of FeO in the Slag and

Silicon in the Metal on Hot Metal Desulfurization," in Steelmaking Conference Proceedings,

Toronto, 1998.

[62] H. Asth, R. Tavares, S. Silva, R. Sampaio, O. Neto, K. B. F. Schwerdtfeger and L. Silva,

"Microcleanliness Improvements in VM2011 Steel and Evaluation of Benefits by Increasing the

Alumina Content in the Ladle Furnace Slag at the V&M do Brasil," in 8th International

Conference on Clean Steel, Budapest, Hungary, 2012.

[63] L. Zhang, Y. Ren, D. Hoajian, W. Yang and L. Sun, "Stability Diagram of Mg-Al-O System

Inclusions in Molten Steel," Metallurgical and Materials Transactions B, vol. 46B, August, pp.

1809-1825, 2015.

[64] W. Bielefeldt and A. Vilela, "Study of Formation and Modification of Inclusions in Al-Killed Ca-

Treated Steel," in 8th International Conference on Clean Steel, Budapest, 2012.

[65] G. Yang, X. Wang, F. Huang, D. Yang, P. Wei and X. Hao, "Influence of Calcium Addition on

Inclusions in LCAK Steel with Ultrlow Sulfur Content," Metallurgical and Materials Transactions

B, vol. 46B, no. February, pp. 145-154, 2015.

[66] S. Yang, L. Zhang, L. Sun, J. Li and K. Peaslee, "Investigation on MgO-Al2O3 Based Inclusions

in Steels," AIST Transactions, no. August, pp. 245-258, 2012.

[67] N. Verma, R. Fruehan and P. Pistorius, "Modification of Alumina and Spinel Inclusions by

Calcium in Liquid Steel," CISR Spring Meeting, Carnegie Mellon University, 2011.

[68] S. He, G. Chen, Y. Guo, B. Shen and Q. Wang, "Morphology Control for Al2O3 Inclusion without

Ca Treatment in High-Aluminum Steel," Metallurgical and Materials Transcations B, vol. 46B,

April , pp. 585-594, 2015.

[69] F. Zhang, G. Li, B. Wang and Y. Zhang, "Effects of Ca Treatment on Magnetic Properties of High

Efficient NO Si Steel Sheets," in AISTech Conference Proceedings, Indianapolis, 2011.

121

[70] A. Dash and P. Kaushik, "A Laboratory Study of Factors Affecting Sulfide Shape Control in Plate

Grade Steels," in AISTech Conference Proceedings Volume II, Indianapolis, 2011.

[71] S. Yoshida, M. Kubota, K. Miyanishi, Y. Abe and K. Ushioda, "Influence of Manganese and

Sulphur Concentration on Formation of Fine MnS," in 8th International Conference on Clean

Steel, Budapest, Hungary, 2012.

[72] N. Verma, C. Pistorius, R. Fruehan, M. Potter, M. Lind and S. Story, "Transient Inclusion

Evolution During Modification of Alumina Inclusions by Calcium in Liquid Steel: Part I.

Background, Experimental Techniques and Analysis Methods," Metallurgical and Materials

Transactions B, vol. 42B, no. August, pp. 711-719, 2011.

[73] N. Verma, C. Pistorius, R. Fruehan, M. Potter, M. Lind and S. Story, "Transient Inclusion

Evolution During Modification of Alumina Inclusions by Calcium in Liquid Steel: Part II. Results

and Discussion," Metallurgical and Materials Transactions B, vol. 42B, August, pp. 720-729,

2011.

[74] AISE Steel Foundation, The Making, Shaping, and Treating of Steel, 11th Edition, Pittsburgh: The

AISE Steel Foundation, 2003.

[75] A. Marins, W. Martins, C. Silva, I. Silva, J. Neto and V. Seshadri, "Assessing Inclusions Removal

in a Steelmaking Ladle: Plant Assessment and Modeling," in AISTech Conference Proceedings,

Atlanta, 2012.

[76] X. Wang, Q. Li, F. Huang, H. Li, J. Yang and X. Kang, "Control of B Type Non-Metallic

Inclusions in Linepipe Steel Plates," in 8th International Conference on Clean Steel, Budapest,

Hungary, 2012.

[77] D. Mazumdar and R. Guthrie, "Mixing Times and Correlations for Gas Stirred Ladle Systems,"

ISS Transactions, no. September, pp. 89-96, 1999.

[78] F. Ruby-Meyer, A. Karasev, P. Jonsson, M. Perez-Alonso and S. Baragiola, "Optimisation of

Sampling for Inclusion Assessment at Liquid Steel Stage," in 8th International Conference on

Clean Steel, Budapest, Hungary, 2012.

[79] S. Baragiola, A. Braconi, V. Colla and G. Nastasi, "Prediction of Final Cleanness Based on the

Assessment of Liquid Steel Inclusion Population," in 8th International Conference on Clean Steel ,

Budapest, 2012.

[80] P. Kaushik, J. Lehmann and M. Nadif, "State of the Art in Control of Inclusions, Their

Characterization, and Future Requirements," Metallurgical and Materials Transactions B, vol.

43B, August, pp. 710-725, 2012.

[81] R. A. Falk.United States of America Patent 5,014,561, 14 May 1991.

[82] T. Junker and A. Reilly, "Revolutionary Approach to Consistently Obtain the Cleanest and Most

Representative Steel Samples from Molten Metal," in AISTech Conference Proceedings,

Indianapolis, 2007.

122

[83] W. Souronis, P. Turner and R. Rote, "Improvement in Sampling Process Precision by Means of an

Inert Atomsphere Sampling System at ArcelorMittal Indiana Harbor East #2 Steel Producing," in

AISTech Conference Proceedings, Atlanta, 2012.

[84] M. Li, F. Wang, C. Li, Z. Yang, Q. Meng and S. Tao, "Effects of Cooling Rate and Al on MnS

Formation in Medium-Carbon Non-Quenched and Tempered Steels," International Journal of

Minerals, Metallurgy, and Materials, vol. 22, no. 6, pp. 589-597, 2015.

[85] O. Ericsson, M. Lionet, A. Karasev, R. Inoue and P. Jonsson, "Changes in Inclusion

Characteristics During Sampling of Liquid Steel," Iron and Steelmaking, vol. 39, pp. 67-75, 2012.

[86] K. Beskow, P. Dayal, J. Bjorkvall, M. Nzotta and D. Sichen, "A New Approach for the Study of

Slag - Metal Interface in Steelmaking," Iron and Steelmaking, vol. 33, February, pp. 74-80, 2006.

[87] Y. Won and B. Thomas, "Simple Model of Microsegregation during Solidification of Steels,"

Metallurgical and Materials Transactions A, pp. 1755-1767, 2001.

[88] M. Lage and A. Costa e Silva, "Evaluating Segregation in HSLA Steels Using Computational

Thermodynamics," Journal of Materials Research and Technology, pp. 353-358, 2015.

[89] S. Gerasin and D. Kalisz, "Modeling of the Mn and S Microsegregation During Continuous

Casting of Rail Steel," Archives of Foundry Engineering, pp. 35-38, 2015.

[90] E. Scheil, "Bemerkungen zur schichtkristallbildung," Zeitschrift Metallkunde, vol. 34, no. ., pp. 70-

72, 1942.

[91] T. Clyne and W. Kurz, "Solute Redistribution During Solidification with Rapid Solid State

Diffusion," Metallurgical Transactions, p. 12A:965, 1981.

[92] H. Brody and M. Flemings, "Measurement of Solute Redistribution in Dendritic Solidification,"

Trans. Met. Soc., vol. 236, pp. 624-633, 1966.

[93] D. Janis, A. Karasev, P. Jonsson, T. Engkvist and G. Runnsjo, "Evaluation of Inclusion

Characteristics in Stainless Steel Samples Using PDA/OES at Outokumpu Stainless AB in

Avesta," in 8th International Conference on Clean Steel, Budapest, Hungary, 2012.

[94] A. Karasev, K. Nakajima and P. Jonsson, "Possibilites to Obtain Feedback on Inclusion

Characteristics During Ladle Refining," in 9th International Conference on Molten Slags, Fluxes,

and Salts, Beijing, 2012.

[95] M. Harris, O. Adaba, S. Lekakh, R. O'Malley and V. Richards, "Improved Methodology for

Automated SEM/EDS Non-Metallic Inclusions Analysis of Mini-Mill and Foundry Steels," in

AISTech Conference Proceedings, Cleveland, 2015.

[96] Y. Wang, X. Zuo, L. Zhang, S. Li, A. Dong and L. Damoah, "Entrapment of Inclusions in

Continuous Casting Billet: Industrial Observation and Modeling," in AISTech Conference

Proceedings, Pittsburgh, 2010.

123

[97] S. Abraham, J. Raines and R. Bodnar, "Development of an Inclusion Characterization

Methodology for Improving Steel Product Cleanliness," AIST Transactions, vol. 11, no. 2, 2014.

[98] J. Peter, K. Peaslee, D. Robertson and B. Thomas, "Experimental Study of Kinetic Processes

During the Steel Treatmeent at Two LMF's," US DOE DE-FC36-03ID14279.

[99] A. Harada, G. Miyano, N. Maruoka, H. Shibata and S. Kitamura, "Composition Change of

Inclusions During Ladle Treatment by the Reaction with Slag and Refractory," in 8th International

Conference on Clean Steel, Budapest, Hungary, 2012.

[100] Q. Shu, S. Lachmann, O. Volkova and P. Scheller, "Inclusion Development in Steel During

Metallurgical Ladle Treatment - A Process Simulation Model," in MetTec InSteel Conference,

Dusseldorf, 2011.

[101] Q. Shu, O. Volkova, S. Lachmann and P. Scheller, "Modification of Inclusion Composition in

Steel During Secondary Metallurgical Ladle Treatment - A Comprehensive Process Simulation

Model," in AISTech Conference Proceedings, Indianapolis, 2011.

[102] S. Wang and S. Louhenkilpi, "EMS on Larger Sized Inclusions in the Billet/Bloom Casting," in

METEC, InSteelCon, STEELSIM, Dusseldorf, 2011.

[103] S. Yang, J. Li, C. Liu, L. Sun and H. Yang, "Motion Behavior of Nonmetal Inclusions at the

Interface of Steel and Slag. Part II: Model Application and Discussion," Metallurgical and

Materials Transactions B, vol. 45B, no. December, pp. 2453 - 2463, 2014.

[104] K. Graham and G. Irons, "Elaboration of the Cell Model for Application to Ladle Metallurgy

Slags," in AISTech Conference Proceedings, Pittsburgh, 2008.

[105] J. Perez and G. Irons, "Inclusion Control Model in the Ladle Metallurgy Furnace," in AISTech

Conference Proceedings, Pittsburgh, 2013.

[106] J. Rogler, L. Heaslip, D. Xu and M. Mehrvar, "Modeling of Ladle Bubbling for the Elmination of

Thermal Stratification: Part I - Without Heating," AIST Transactions, vol. April , pp. 81-93, 2004.

[107] J. Rogler, L. Heaslip, D. Xu and M. Mehrvar, "Modeling of Ladle Bubbling for the Elmination of

Thermal Stratification: Part II - With Heating," AIST Transactions, vol. May, pp. 93-99, 2004.

[108] K. Krishnapisharody and G. Irons, "A Model for Slag Eyes in Steel Refining Ladles Covered with

Thick Slag," Metallurgical and Materials Transactions B, vol. 46B, February, pp. 191-198, 2015.

[109] K. Swartz, L. Hibbeler, B. Joyce and B. Thomas, "Numerical Investigation of Slag Entrainment,"

Association for Iron and Steel Technology, vol. July, pp. 79-90, 2015.

[110] K. Krishnakumar and G. Irons, "Water Model Study of the Spout and Eye Formation in a Gas-

Stirred Ladle," in ISSTech Conference Proceedings, Indianapolis, 2003.

[111] S. Yang, L. Zhang, J. Li and K. Peaslee, "Fluid Flow and Steel Cleanliness in a Continuous

Casting Tundish and Mold," in AISTech Conference Proceedings - Volume II, Indianapolis, 2011.

124

[112] C. Pistorius, "Bubbles in Process Metallurgy," in Treatise on Process Metallurgy, Volume 2,

Elsevier Ltd. , 2014, pp. 179-196.

[113] S. Seetharaman, D. Sichen, P. Jonsson and L. Jonsson, "Modeling of Metallurgical Reactions

Towards Process Simulations," in 6th Japan-Nordic Countries Joint Symposium, Nagoya, 2000.

[114] B. Li, H. Yin, C. Zhou and F. Tsukihashi, "Modeling of Three-phase Flows and Behavior of

Slag/Steel Interface in an Argon Gas Stirred Ladle," ISIJ International, vol. 48, no. 12, pp. 1704-

1711, 2008.

[115] L. Li, Z. Liu, M. Cao and B. Li, "Large Eddy Simulation of Bubbly Flow and Slag Layer Behavior

in Ladle with Discrete Phase Model (DPM) - Volume of Fluid (VOF) Coupled Model," JOM,

July, 2015.

[116] P. Shivaram, "CFD Modeling to Simulate Gas Stirring Process Using Bottom Plugs in a Steel

Ladle," in AISTech Conference Proceedings, Cleveland, 2015.

[117] L. Costa, R. Tavares, B. Braga, D. Lima and H. Lino, "Mathematical Model of Multiphase Flow in

Argon Stirred Ladle and a New Technique for Prediction of the Mixing Time," in AISTech

Conference Proceedings, Atlanta, 2012.

[118] J. Alexis and J. Bjorkvall, "Mathematical Modeling of Stirring for an Optimized Ladle Furnace

Process," in AISTech Conference Proceedings, Indianapolis, 2011.

[119] M. Andersson, L. Jonsson and P. Jonsson, "A Thermodynamic and Kinetic Model of Reoxidation

and Desulphurisation in the Ladle Furnace," ISIJ International , vol. 40, no.11, pp. 1080-1088,

2000.

[120] L. Nastac, Q. Cao, D. Zhang, A. Pitts and R. Williams, "3D CFD Modeling of the LMF System,"

The Minerals, Metals and Materials Society, vol. CFD Modeling and Simulation in Materials

Processing 2016, pp. 187-194, 2016.

[121] J. Aoki, L. Zhang and B. Thomas, "Modeling of Inclusion Removal in Ladle Refining," in ICS

Proceedings, Montreal, 2005.

[122] J. Björkvall, Thermodynamic Study of Multicomponent Slags - A Model Approach, Stockholm,

Sweden: PhD Thesis, 2000.

[123] M. Nzotta, D. Sichen and S. Seetharaman, "A Study of the Sulfide Capacities of Iron-Oxide

Containing Slags," Metallurgical and Materials Transaction B, vol. 30B, no. October, pp. 909-

920, 1999.

[124] A. Pitts, J. Hallmark, P. Turner, B. Williams and L. Nastac, "Inclusion Evolution Mapping during

the LMF Process via a Multi-Depth Sampling Technique," in AISTech Conference Proceedings,

Pittsburgh, 2016.

[125] A. Pitts-Baggett, P. Turner, J. Hallmark, B. Williams and L. Nastac, "Method for Simultaneous

Sampling of Steel at Multiple Depths in the Refining Ladle," Iron and Steel Technology, January,

2017.

125

[126] G. Frigm, R. P. Stone, J. Plessers, R. S. Bogan and E. T. Turkdogan, "Accuracy of Hydrogen

Measurement in Liquid Steel with In-Situ Microprocessor-Based System," Process Controls &

Instrumentation, 1991.

[127] C. Grip and L. Jonsson, "Physical Behavior of Slag in a 107-Tonne Ladle: Production Scale

Experiments and Theoretical Simulation," Scandinavian Journal of Metallurgy, vol. 32, pp. 113-

122, 2003.

[128] J. Vlcek, V. Tomkova, H. Ovcacikova, F. Ovcacik, M. Topinkova and V. Matejka, "Slags from

Steel Production: Properties and Their Utilization," Metallurgia, vol. 52, no. 3, pp. 329-333, 2013.

[129] E. Pretorius, "Refinement of Steel". USA Patent WO 2008/070360 A2, B22D 11/116 (2006.01),

PCT/US2007 /083125, 12, 2006.

[130] R. O'Malley, "Inclusion Evolution and Removal in Ladle Refining," in AISTech Conference

Proceedings, Nashville, 2017.

[131] Z. Zhang, A. Tilliander, A. Karasev and P. Jonsson, "Simulation of the Steel Sampling Process,"

ISIJ International , vol. 50, no. 12, pp. 1746-1755, 2010.

[132] E. Pretorious, H. Oltmann and T. Cash, "The Effective Modification of Spinel inclusions by Ca

Treatment in LCAK steel," Iron and Steel Techonology, no. July, pp. 31-44, 2010.

[133] M. Hemmerlin, L. Paulard and G. and Schotter, "Determination of ultra low carbon and nitrogen

contents in steel: combustion versus electrical spark source optical emission spectrometry for steel

making process control," Journal of Analytical Atomic Spectrometry, pp. 282-286, 2003.

[134] S. Story and R. Asfanhani, "Control of Ca-Containing Inclusions in Al-Killed Steel Grades,"

United States Steel Corporation.

[135] L. Nastac, D. Zhang, Q. Cao, A. Pitts and R. Williams, "3D CFD Modeling of the LMF System,"

TMS, vol. CFD Modeling and Simulation in Materials Processing , pp. 187-194, 2016.

[136] A. Ishii, M. Tate, T. Ebiswawa and K. Kawakami, in Electric Furnace Conference Proceedings

(40th), Kansas City, 1982.