Solubility of hydrogen in slags and its impact on ladle refining

Transcription

1 Solubility of hydrogen in slags and its impact on ladle refining Jenny Brandberg Licentiate Thesis School of Industrial Engineering and Management Division of Micro Modelling Royal Institute of Technology SE Stockholm Sweden Akademisk avhandling som med tillstånd av Kungliga Tekniska Högskolan i Stockholm, framlägges för offentlig granskning för avläggande av Teknologie licentiatexamen, fredagen den 20 oktober 2006, kl i sal B2, Brinellvägen 23, Kungliga Tekniska Högskolan, Stockholm ISRN KTH/MSE--06/57--SE+MICROMODMETU/AVH ISBN

2 Jenny Brandberg. Solubility of hydrogen in slags and its impact on ladle refining KTH School of Industrial Engineering and Management Division of Micro Modelling Royal Institute of Technology SE Stockholm Sweden ISRN KTH/MSE--06/57--SE+MICROMODMETU/AVH ISBN The Author

3 ABSTRACT The aim of the present work was to clarify the mechanisms of hydrogen removal during vacuum degassing. The main reason for this was because the primary source of hydrogen pick-up in steel-making is the moisture in the furnace atmosphere and the raw material charged into the ladle furnace. Previous studies showed that the presence of hydroxyl ions in the ladle slag results in hydrogen transfer from the slag back into the steel bath. The main focus of this thesis was therefore to gain deeper knowledge of the ladle slag and its properties. For this purpose a number of slag compositions were examined in order to clarify whether these slags were single liquids at 1858 K. 14 out of 27 compositions in the Al 2 O 3 -CaO-MgO-SiO 2 system was completely melted. These results were in disagreement with the existing phase diagrams. Water solubility measurements were carried out by employing a thermo gravimetric technique. The temperature was found to have negligible effect on the water solubilities. The experimental results showed that the water capacity values varied between and in the majority of the composition range. However, for compositions close to CaO saturation the water capacity value could reach higher than The experimental determined water capacity was further used to develop a water capacity model for the quaternary slag system Al 2 O 3 -CaO-MgO-SiO 2. The model was constructed by considering the affects of the binary interactions between the cations in the slag on the capacity of capturing hydroxyl ions. The model calculations agreed well with the experimental results as well as with the literature data. An attempt was made to develop a preliminary process model for dehydrogenation by using the results from CFD calculation. For this purpose industrial sampling was made during vacuum treatment. The hydrogen concentrations decreases fast in the initial stages of the degassing, but is slowed down in the final stage. The model calculations fit the initial stage of the dehydrogenation process well. In the final stage of the process the predicted values are somewhat lower than the plant data. v

4 The results from the model prediction showed that a dynamic process model could be satisfactorily constructed using the results from CFD calculation. The present work aimed at determining how big impact hydroxyl ions in the slag have on the final hydrogen concentration in the liquid steel. It was found that the effect is of less importance regarding the final concentration of the metal after the degassing treatment. Keywords: ladle slag, dehydrogenation, water solubility, water capacity, model, hydrogen, vacuum degassing, steel, ladle treatment vi

5 ACKNOWLEDGMENTS I will start by acknowledge the man who made this thesis possible, my supervisor Professor Du Sichen. Thank you for your encouragement and excellent guidance. Special thanks to Professor Seshadri Seetharaman for his fruitful discussions and valuable comments. Jernkontoret and Brinell centre are gratefully acknowledged for their financial support. I choose to express a special thank you to everyone who has contributed to this work in one way or another. Thanks to all the anonymous reviewers who contributed with valuable advice and for their vital comments and encouraging support. Thank you to all my colleagues at the Department of Materials Science and Engineering. Especially, to Peter Kling who with his positive spirit always helped me with my troubles during the experimental work. Thank you also for the bakeries and ice-creams. I am grateful for the experimental facilities and equipments for this work provided by Uddeholm Tooling AB. Thank you, Tech. Lic. Karin Stenholm, PhD. Mselly Nzotta and Andreas Norberg, all from Uddeholm Tooling AB, for your help with the industrial sampling and analysing. A very special thanks is owed to my father, Sten and my mother, Ewa for their love and support. My brother, Joakim and my sister, Anna-Lena, is also gratefully acknowledged for always letting me know I was the youngest, pushing me to perform a little bit more. Without you I would not be here. Finally, I would like to thank my better half, Hans, who with his patience and support helped me through all the difficult times. Not least, I am really grateful to him for the physical manifestation of our love, our beautiful and loving son, Alvin. vii

6 The thesis is based on following supplements: I. Characterization of melting of some slags in the Al 2 O 3 -CaO-MgO-SiO 2 quaternary system F. Dahl, J. Brandberg, Du Sichen ISIJ International, vol. 46, no. 4, pp , April II. Water vapour solubility in ladle-refining slags in the Al 2 O 3 -CaO-MgO-SiO 2 quaternary slag system J. Brandberg, Du Sichen Metallurgical and Materials Transactions B, vol. 37B, no. 3, pp , June III. IV. Water capacity model of Al 2 O 3 -CaO-MgO-SiO 2 quaternary slag system J. Brandberg, L. Yu, Du Sichen Accepted for publication in Steel research March Process model for dehydrogenation during vacuum degassing J. Brandberg The contribution by the author to the different supplements of the thesis: I. Part of literature survey, assisted during experimentation and major part of the writing. II. Literature survey, experimentation and major part of writing. III. Literature survey, assisted during experimentation, major part of modeling and major part of the writing. IV. Literature survey, sampling, modeling and writing. The following work has been carried out by the author but are not a part of the thesis Hydrogen Pick-up in an Argon Bubble J. Brandberg, E. Johansson, M. Magnelöf, T. Niemi, E. Rutqvist, D, Sichen, U. Sjöström. Steel Grip Journal of Steel and Related Materials, in press. Stockholm in August 2006 The Author viii

7 CONTENTS 1 Introduction Previous work The objectives of the present work 2 2 Experimental Work Laboratory work 5 Materials and sample preparation 5 Experimental set-up and procedure 5 Characterization of melting of slags 5 Water vapour solubility experiments Industrial data 9 Plant description 9 Sampling 9 3 Modeling work Hydroxyl capacity model 11 Construction of the model 11 Hydrogen capacity data Process model 13 Metal-gas reaction 14 Metal-slag reaction 15 Model calculation 16 4 Results Laboratory work 17 Characterization of melting of slags 17 Water vapour solubility Industrial sampling Modeling work 21 Hydroxyl capacity model 21 Process model 23 ix

8 Contents 5 Discussion Laboratory work 25 Characterization of melting of slags 25 Water vapour solubility Industrial work Modeling work 29 Hydroxyl capacity model 29 Process model 30 6 Conclusions Proposal for future work 34 Bibliography 35 x

9 Chapter 1 INTRODUCTION Steel is the worlds most produced metallic material and has been a very important construction material for a long period of time. It is relatively cheap to produce and the properties of the steel can easily be varied by the added alloys. From the beginning the ladle was used only as a transport vessel between the melting furnace and casting. The demands of cleaner steel led to development in steel refining and ladle metallurgy was from now on used for various kinds of refining operations. Hydrogen is a non-metallic element that causes embrittlement, pipes and reduction of mechanical properties in the final steel products. Even the presence of a few parts per million of hydrogen in molten steel can have serious effect upon the end material. A recent study has shown that a hydrogen concentration as low as 1-2 ppm would have detrimental impact on the mechanical properties of tool steels. 1 The primary source of hydrogen pick-up in steel making is the moisture in the furnace atmosphere and the raw materials charged into the ladle furnace. Although most hydrogen is removed as vapour, low pressure is required to bring down the hydrogen concentrations below 1 ppm. An efficient dehydrogenation process is necessary in order to meet the demand of cleaner steel. The most common industrial process for dehydrogenation is the vacuum degassing treatment. In many industrial practices, synthetic ladle slags are based on the Al 2 O 3 -CaO-MgO-SiO 2 quaternary system. The slag composition plays an important role in ladle refining. While both the level of vacuum and argon flow rate is of great importance in determining the rate of hydrogen removal, the partition of hydrogen between slag and molten steel plays another crucial role. 1

10 Introduction 1.1 Previous work Dor et. al. have reported that the water content in the slag after vacuum treatment could be 200 times higher than the value predicted from the metal-gas-slag equilibrium. 2 This implies that the dehydrogenation kinetics in the steel is much faster than it is in the slag. 3 The hydrogen solubilities of binary, ternary and quaternary slag system have been carried out by different research groups. 3-8 Researchers have found the solubility of water to be proportional to the square root of the water pressure. 3,9-12 This finding supports that the water vapour is present as free hydroxyl ions in basic slags. 9 By increasing the basicity, the solubility of water is increased. 3-4,10,13 The temperature on the other hand has been found to have negligible effect on the solubility of water vapour in metallurgical slags ,14 This can be attributed to low heat of solution of water vapour in metallurgical slags. 11,15 Some effort has been put forward to describing the hydrogen capacity as a function of temperature and slag composition. Sosinsky et al. 10 relates hydroxyl activities in CaO-MgO-SiO 2 melts to the activity of silica and I. D. Sommerville tries to relate hydroxyl capacities to basicities of slags. 16 Ban-ya et al. 17 used the regular solution model of Lumsden on binary and ternary slag systems in order to predict the hydroxyl capacities. Jo and Kim 3 extended this model to the quaternary system Al 2 O 3 -CaO-MgO-SiO The objective of the present work The major oxide components of a ladle slag are usually Al 2 O 3, CaO, MgO and SiO 2. The phase diagram information from slag Atlas 18 of the Al 2 O 3 -CaO-MgO- SiO 2 quaternary system involves considerable uncertainties since many liquidus temperatures are given as dotted lines. Disagreements between the phase diagram and the reality have been found by both the present research laboratory and the steel industry 19. One aim with this thesis is therefore to study a number of slag compositions to examine whether these slags are single liquids at 1858 K. Water vapour dissolves in liquid slag in a form of hydroxyl ion or hydroxyl radical. The presence of hydroxyl ions in the ladle slag will result in hydrogen transfer from the slag back into the steel bath. 9 Hence, the initial concentration of hydroxyl ions in ladle slag would seriously affect the kinetics of the dehydrogenation process and therefore the hydrogen content in the final product. In order to optimize the dehydrogenation process, a good knowledge of the hydrogen solubility of the ladle slag is essential. The slags studied in the Al 2 O 3 -CaO-MgO-SiO 2 system are mostly in the composition region having relatively higher SiO 2 contents. 3 On the other 2

11 Chapter 1 hand, many ladle processes operate with slags having less than 10 mass% SiO 2 nowadays. To meet the demand of process optimization of ladle treatment, the present work focuses on the determination of hydrogen solubilities of some Al 2 O 3 - CaO-MgO-SiO 2 quaternary slags having lower SiO 2 concentrations. To meet the demand of cleaner steel, an efficient dehydrogenation process is required. In ladle refining, slag plays a crucial role in desulphurization, dephosphorization and inclusion removal. Optimization and the development of a process model of the ladle treatment necessitates good knowledge of the ladle slag with respect to, among other properties, its capacities of taking up sulphur, phosphorus and hydrogen. Very often, descriptions of these capacities as a function of temperature and slag composition are required, since the slag composition varies during refining. Jo and Kim 3 described the hydrogen capacity for the quaternary Al2O3-CaO-MgO-SiO2 system as a function of temperature and slag composition. However, the model by Jo and Kim 3 predicts much lower values when compared to the experimental data in the high basicity region. Therefore, this thesis aims at developing a water capacity model for quaternary ladle slags. Producing clean steel is important for the steel producers as well as increased productivity. Vacuum treatment is a time-consuming process step and is therefore a limiting step during steel production. Since no sampling can be made during the degassing, the process time is usually longer than necessary in order to make sure that the level of hydrogen meets the requirements. This means that if the dehydrogenation rate could be predicted, the vacuum degassing time could be reduced, thereby increasing the productivity of the steel plant. Another aim of this work is to develop a process model for dehydrogenation by using the results from a CFD model 20 that describes the flow in the ladle. For this purpose industrial sampling during vacuum treatment was made at Uddeholm Tooling AB. 3

12 4 Introduction

13 Chapter 2 EXPERIMENTAL WORK 2.1 Laboratory work Materials and sample preparation In many industrial practices, synthetic slags are based on the Al 2 O 3 -CaO-MgO-SiO 2 quaternary system. These four oxide compounds were therefore used in the present work. The oxides were calcined separately for 24 hours at 1073 K. After calcinations, the oxides were ground and thoroughly mixed in predetermined proportions. The mixtures were pressed into pellets and platinum crucibles were used as sample container during the experiments. The pellets were kept in a sealed glass bottle that was kept in a desiccator before the experiment. Experimental set-up and procedure Characterization of melting of slags A number of ladle slag compositions were studied to examine whether these slags are single liquids or multiphase mixtures at 1858 K. For this purpose a quenching technique was employed. A horizontal furnace with an alumina reaction tube was employed for most of the experiments. The experimental setup is schematically shown in Figure 1. The furnace had KANTHAL SUPER 1800 heating elements controlled by a regulator using a Pt-30 pct Rh/Pt-6 pct Rh thermocouple. An alumina holder was used to support two crucibles containing the samples. In a typical run, the samples were 5

14 Experimental Work carefully introduced into the reaction chamber and placed in the even temperature zone. A constant argon gas flow was passed through the reaction tube at low flow rate. The samples were heated to 1858 K and kept for more than 5 hours. The sample holder was rapidly pulled out of the furnace. After a quick visual examination, to see if the sample was completely molten, the samples were quenched in liquid nitrogen. In order to confirm the reproducibility of the experimental method, each slag composition was investigated four times Figure 1: Schematic figure of the horizontal furnace. 1. silicone rubber stopper; 2. alumina reaction tube; 3. furnace; 4. gas inlet; 5. thermocouple; 6. alumina crucible holder; 7. platinum crucible containing the sample; 8. gas outlet. In many ladle treatments, slags of high basicities are employed. In order to examine whether these high basicity slags were still single liquid at lower temperatures that might be encountered in ladle refining, these slags were quenched from 1793 K. To avoid the slight possibility of solidification of the sample during quenching, a vertical furnace was employed. The vertical furnace also had KANTHAL SUPER 1800 heating elements. As shown in Figure 2, an alumina reaction tube was used as the reaction chamber. The slag sample was placed in a platinum crucible welded in both ends. The platinum crucible was hanged by a piece of platinum wire in the even temperature zone of the furnace. The sample was heated and kept at the experimental temperature, 1793 K for 7 hours. The platinum wire was cut off and the sample was quenched in liquid nitrogen. The samples were embedded in bakelite under vacuum for microscopic examination. 6

15 Chapter Figure 2: Schematic figure of the vertical furnace. 1 gas outlet; 2 silicone rubber stopper; 3 platinum wire; 4 samples; 5 alumina reaction tube; 6 silicone rubber stopper; 7 thermocouple; 8 gas inlet. Water vapour solubility experiments The thermo gravimetric technique was used for determination of the hydrogen solubilities of some CaO-SiO 2 binary slags and Al 2 O 3 -CaO-MgO-SiO 2 quaternary slags. Thermocouple Gas outlet with bubbler Flow meter Three way switch Three way switch Argon Flow meter Glass balls Sample Furnace Alumina transducer Alumina tube Water column Gas inlet (heated) Figure 3: Connection to prevent condensation Chamber Balance Schematic diagram of the experimental setup. The experimental setup used in this study is schematically illustrated in Figure 3. The setup consists of three major components, the water column, the balance 7

16 Experimental Work chamber and the furnace. The water column has heating elements and is thereby acting as a boiler. Argon gas is introduced at the bottom of the water column. The whole column is filled with small glass balls in order to provide good contact between water and argon gas. By adjusting the temperature of the water column, the water pressure in the H 2 O-Ar mixture can be controlled on the basis of the thermodynamic data. The balance (METTLER TOLEDO AG285) is placed in an air tight box and it is connected to a computer, which follows the weight change in situ. An alumina transducer sitting on the balance is used to hold the sample in the even temperature zone of the furnace. The furnace is a high temperature furnace with an alumina tube employed as the reaction chamber. A pair of B type (Pt-30 pct Rh/Pt-6 pct Rh) thermocouple is used to control the furnace temperature. Another pair of B type thermocouple is placed just above the sample to measure the temperature. H 2 O-Ar gas mixture is introduced into the reaction chamber by a stainless steel tube connected to an alumina tube, the tip of which is about 5 cm below the sample. In order to prevent water condensation from the moist gas, the inlet stainless steel tube is always kept at 363 K. In a general run, a sample along with the platinum crucible was placed on the top of the alumina transducer before the reaction tube was sealed. An argon flow of Nl/min was introduced into the reaction chamber. The furnace was heated up to the experimental temperature at a heating rate of 5 K/min. After a period of stabilization, H 2 O-Ar gas mixture was introduced into the reaction chamber. In most of the experiments, the water column was kept at 327 K. Repetitions of measurements of some slag compositions were carried out at different water pressure. The reproducibility was also checked using higher flow rate of argon flow. Calibration was made by making experiment using platinum crucible containing a piece of pure re-crystallized alumina under exactly identical conditions. To confirm the reliability of the experimental data, two slag samples were equilibrated in a horizontal furnace using the same water column. After equilibrating the slag samples with the H 2 O-Ar mixture they were quenched in liquid nitrogen. The samples were weighed before the experiment and just after quenching. The weight change would give the solubility of H 2 O or hydroxyl ions of the slag at the experimental temperature. 8

17 Chapter Industrial data Plant description The industrial sampling were carried out at the scrap based steel plant Uddeholm Tooling AB, Hagfors, Sweden. A schematic illustration of the production route at Uddeholm Tooling AB can be seen in Figure 4. The recycled steel is melted in an electric arc furnace, EAF. The molten steel is transferred to a 35 to 65 ton ladle furnace where it is deslagged, deoxidized and alloyed. A synthetic slag is added and the steel is heated to the right temperature before the ladle is moved to the vacuum degassing station. After vacuum degassing, the steel is transferred to the up-hill ingot casting. Figure 4: Flow chart of the production route at the steel plant. Sampling Measurements of hydrogen content along with slag samples were taken from the tool steel grade ORVAR 2M (Fe-0.39C-1.0Si-0.4Mn-5.3Cr-1.3Mo-0.9V). The slag samples were taken with a sampling scope and they were cooled down immediately before they were sealed in a glass bottle and placed in a dessicator. Hydrogen concentration was measured with Hydris (Hydrogen Direct Reading Immersion System), Heraeus Electro-Nite before, during and after vacuum degassing. Hydris is an on-line system that measures the hydrogen content of steel in the ladle, tundish or ingot. The carrier gas absorbs hydrogen until equilibrium is attained between gas and steel. The hydrogen partial pressure is analysed by a thermal conductivity detector placed in the pneumatic unit. The hydris processor unit converts the partial pressure to the hydrogen content by Sieverts law. When HYDRIS malfunctioned, hydrogen samples were taken manually with a sampling scope. 9

18 Experimental Work From the scope steel was collected in a small glass tube that immediately was put in the freezer and then sent to the laboratory for analyses. The slag samples were sent to NILAB in order to evaluate the hydrogen content. 10

19 Chapter 3 MODELING WORK The existing model for the quaternary system Al 2 O 3 -CaO-MgO-SiO 2 developed by Jo and Kim 3 predicts much lower values when compared to the experimental data in the present study. This disagreement could be attributed to the fact that the experimental measurements made by Jo and Kim 3 in the quaternary system are mostly in the high silica content region. The measurements in the present work are in the same system but mainly in the high basicity region. A hydroxyl capacity model for quaternary ladle slags has therefore been developed. 3.1 Hydroxyl capacity model Construction of the model The hydroxyl capacity is introduced on the basis of the following slag-gas reaction, 2 H O(gas) + O (slag) = 2OH (slag) (1) 2 The hydroxyl capacity of a slag can be expressed as: C OH 0.5ΔG = exp RT o 1 f a O OH 2 (2) o where Δ G 1 is the standard Gibbs energy of reaction (1), R is the gas constant and T is the temperature in K. f and a 2 stand for the activity coefficient of OH - and OH O the activity of O 2- respectively. 11

20 Modeling work The exponential term in (2) is system independent, while a 2 and f depend on O OH the slag composition and temperature. It should be mentioned that the standard states for ion activities are difficult to define. Hence, eq. (2) would have very little practical use. On the other hand, both a 2 and f would relate to their partial O OH Gibbs energies exponentially. It is reasonable to expect that a O ψ' 2 = exp (3) f RT OH where ψ' is a function of composition and temperature. Inserting eq. (3) into eq. (2) leads to C OH 0.5ΔG = exp RT o 2 exp ψ' RT ψ = exp RT (4) ψ in eq. (4) is a new function, which depends on the slag composition and temperature. In line with the activity model 21 and sulphide capacity model 22 developed in the present department, the slag could be considered to have two sub-groupings. A silicate melt containing m oxides, C1 c1 O a1, C2 c2 O a2,...ci ci O ai,...c cm O am can be expressed as: v1 v 2 vi vm 2 [( C1 ) y, ( C2 ) y,, ( Ci ) y,, ( Cm ) y ]( O ) q c1 c 2 (5) ci where C vi stands for cations, the superscript vi denotes the electrical charge. Even Si 4+ ion is included in the cation group. q in eq. (5) are stoichiometric coefficient. The different ions would have effects on each other with respect to the capacity to accommodate the OH - ions. Hence, ψ in eq. (4) can be formulated as a function of the interactions between different cations in the presence of O 2-. In a multi-component system, ψ is expressed by: m 12 cm = ( X i ψ i ) ψ mix (6) ψ + 1 where X i is the mole fraction of the oxide component i. ( Xi ψ i) in (6) describes the linear variation of ψ from the pure components in the absence of the interaction between different species. ψ mix arises from the mutual interaction between the different cations in the presence of O 2-. ψ mix can be described by:

21 Chapter 3 m 1 m ψ = y y D (7) mix 1 i+ 1 D is the parameter for the binary interaction between cations C C i C j i and C j. It is expressed as polynomials, D C i C i C C o o 1 1 ( D + D T) + ( D + D T)( y y )... Ci Cj Ci Cj + i j = (8) where y i stands for the ionic fraction of cation within the cation grouping. Previous studies 10-11,14 have shown that hydroxyl capacities of slags have very little dependence on temperature. Hence, the terms related to temperature can be omitted. The interaction parameters can be optimized by using the experimentally determined C. OH Hydrogen capacity data The present hydroxyl capacity model is semi-empirical. The use of reliable experimental data in the optimization would be crucial for the validation of the model. Hence, a careful evaluation of the literature data is needed in order to optimize the model parameters. In the binary system CaO-SiO 2 three publications was found. 6-8 Unfortunately, considerable discrepancy is noticed. While SiO 2 is a network former, CaO is a basic oxide. It is expected that the interaction between Ca 2+ and Si 4+ in the presence of oxygen would have strong effect on the hydroxyl capacity of the slag. New experimental measurements were made in order to decide if any of the documented data should be incorporated in the model. The literature data 3-5,7-8,10,13 in the other systems were found to be in good agreement, and were therefore incorporated in the model calculation. Totally, 171 documented data were used for the optimization. 3.2 Process model An attempt is made in the present work to predict the dehydrogenation process using the velocities of the melt evaluated by CFD calculation. 20 Both the reaction between metal and gas plume as well as the reaction between metal and slag are considered. 13

22 Modeling work Metal-gas reaction Hydrogen is eliminated from the steel phase in three steps: 1. Mass transport of the dissolved gas to the surface of the gas bubble. 2. Reaction at the interface. 2H H2(g) (9) 3. Diffusion from the interface into the bulk of the gas phase (in each bubble). It was found in an earlier study, 23 that the mass transfer of H 2 in the argon gas bubble was very fast. Therefore, it can be assumed that equilibrium is achieved at the interface between the gas phase and the metal phase. If we assume equilibrium at the interface eq. (10) is valid. P ( X + ΔX ) tot H2 H = (10) [ H] int erf + [ ΔH] int erf 2 K 9 K 9 is the equilibrium constant for reaction (9), P tot is the sum of the total pressure, X H 2 is the mole fraction of hydrogen in the gas phase and [ H ] int erf is the hydrogen concentration at the interface. Δ symbolises the change of concentration after one time step. Eq. (10) can be used for calculation of the amount of hydrogen that is removed from the liquid metal during each time step of the model. The initial hydrogen concentration in the steel bulk is known from the industrial sampling, and after each time step the bulk concentration is updated according to eq. (11): i [ H] i [ H] bulk msteel m ΔH = (11) + 1 bulk msteel m steel is the amount of steel in the vessel and m Δ H is the amount hydrogen that is removed by the argon gas. The superscripts i and i+1 denote steps i and i+1, respectively. The bulk concentration can be used in order to calculate the new (step i+1) hydrogen concentration at the interface according to eq. (12): i [ H] i i [ H] ( m m t ) + [ H] ( m t ) + 1 interf steel 1 step bulk 1 interf = msteel step (12) In eq. (12) 1 m is the mass of steel that comes into the gas plume per time step. The mass flow is calculated by using eq. (13). 14

23 Chapter 3 = ( v A) (13) m ρ steel r v r is the velocity, of which the steel is transported towards the gas plume and A the interface area between the plume and the melt. On the basis of the [ H ] interf and [ H] int erf Δ, X H is calculated using eq. (10). 2 Metal-Slag reaction There is a chemical reaction between the slag phase and steel phase according to reaction (14). 2H + O + O OH 2 2 (14) The exchange of hydrogen between the slag and metal is controlled by the partition coefficient, L, and by the mixing fraction between the slag and metal. L (wt%oh) K a 14 O = = COH fh (15) [ wt%h] K 1 Since L can be evaluated by the water capacity of the slag, it is possible to calculate j the amount hydrogen that is transferred from the slag, Δ, back into the metal. H back ΔH j back m = 2 t step L M j j ( L M [ H] M ( H) ) H m H slag bulk + M OH m OH 2 t step slag (16) m 2 is the amount of metal that reacts with the slag per time step and it can be described according to eq. (17). m ( A ) (17) = fmix fmetal ρsteel y 2 v f mix in eq. (17) is the mixing factor between the slag and the metal, f metal is the fraction of metal in the gas-metal plume, A is the area of the cross section of the plume at the level of the slag-metal interface. v y is the velocity, at which the metalgas plume pass trough the cross slag-metal interface. As a reasonable approximation 24, the metal brought up by the gas-metal plume is assumed to have good contact when it is descending back to the metal bulk. Hence, f mix is taken as 1 in the model calculation. The value of f metal is taken as The hydrogen concentration in the slag is updated according to eq. (18) 15

24 Modeling work j+ 1 j ( H) ( ) ( ) j slag = H slag ΔH back (18) where j and j+1 stand for steps j and j+1, respectively. The final hydrogen concentration in the steel bulk is calculated according to eq. (19). j j [ H] [ ] [ ] j tot = H bulk + ΔH back (19) Model calculation The following assumptions are made in the model calculation. The temperature in the ladle is considered to be constant 1800 K and no temperature gradient in the slag or in the steel is considered. The density of pure Fe is used for the liquid metal. The argon gas is assumed to be injected in the ladle through nozzle located at the bottom of the vessel. The gas phase is assumed to initially consist of pure argon. The density of pure Ar is used for the Ar-H 2 mixture. No reaction between the slag phase and the gas phase is considered. The two gas plumes are assumed to function the same way, viz. each is responsible for the dehydrogenation of half of the melt. The activity coefficient, f H, is approximated to 1. The ladle furnace consists of a 65-ton ladle with two porous plugs for argon stirring. The height of the furnace is approximately 3.0 m and the diameter of the same is 2.4 m. The amount of slag was approximated to 800 kg. For the model calculation, the ladle is divided into four zones along the height of the metal bath. The total process time is 1800 s and the time step used for the modelling is 0.01s. The water capacity of the slag is calculated by the water capacity model. The equations were solved by iteration and the initial values were known from the experimental work. The flow incorporated in the model is the results from the CFD calculation for the argon gas flow rate of 50 NL/min. The amplitude of the velocity depends on the position in the ladle. 16

25 Chapter 4 RESULTS 4.1 Laboratory work Characterization of melting of slags The compositions of the investigated slags along with the results of the experiments are presented in Table I. Table I: Compositions of the samples and the experimental results Slag no. mass%al 2 O 3 mass%cao mass%mgo mass%sio 2 Result 1858 K 1793 K multi phase multi phase multi phase multi phase multi phase multi phase liquid liquid liquid liquid liquid liquid liquid liquid liquid multi 17

26 Results phase multi phase multi phase liquid liquid liquid liquid multi phase multi phase multi phase liquid liquid liquid liquid liquid multi phase As can be seen in Table I, 14 samples were completely melted of the investigated slag samples. The four compositions that were investigated further with a fast quenching method were all found to be completely melted. Water vapour solubility Three slag compositions in the binary CaO-SiO 2 system were studied. These compositions are listed in Table II. The experimentally determined H 2 O pick-up along with the evaluated water capacities and hydroxyl capacities of the slags are also presented in the table. Table II: The experimental conditions along with the results and the calculated water and hydroxyl capacities. Slag no. CaO:SiO 2 (mass-%) Temperature (K) P H2O (kpa) ppm H 2 O Water capacity (mass-%/atm 1/2 ) Hydroxyl capacity (mass-ppm/atm 1/2 ) 1a 38.05: b 38.05: : a 55.51: b 55.51: c 55.51: In the CaO-Al 2 O 3 -MgO-SiO 2 quaternary slag system, 12 slag compositions were investigated. The slag compositions along with the experimental conditions are presented in Table III. The mass percentages of H 2 O and the water and hydroxyl capacities in the slags evaluated from the gravimetric measurements are also listed in Table III. 18

27 Chapter 4 Table III: Water solubilities, water capacities and hydroxyl capacities obtained in the present study. Slag no. %Al 2 O 3 %CaO %MgO %SiO 2 T H2O (K) T sample (K) C H2O x10-3 C OH %H 2 O 1A Heating Heating 1B Quenching 2A Heating Heating 2B Cooling Heating 4A Heating Heating 4B Quenching Heating Heating Cooling Heating Heating Heating Heating Heating Cooling 9A Heating Heating Cooling 9B Heating Heating Cooling 9C Heating Heating Cooling Heating Heating Cooling Heating Heating Cooling Heating Heating Cooling Cooling 19

28 Results Most of the experiments were carried out through both heating and cooling cycles. Table III shows that the water solubilities obtained in heating cycle and cooling cycle agree very well. In Tables II and III it can be seen that the reproducibility is good, irrespective of the argon flow rate or water pressure. For slag compositions 1 and 4, the water solubilities were measured by both gravimetric technique and quenching technique. The good agreement of the experimental results using the two techniques shown in Table III indicates again the reliability of the gravimetric technique. It is seen in Table III that irrespective of the slag compositions, the effect of temperature on the water solubilities is negligible. This result is in accordance with the observation reported in the literature , Industrial sampling The plant data collected from the industry is given in Table IV. Table IV: The results from the industrial sampling. Heat HYDRIS H (ppm) T (K) DV52373 BV* min x AV x DV52574 BV* min* AV DV52510 BV* min* AV * DV52557 BV* min* AV* DV52574 BV* min* AV * In the cases when HYDRIS malfunctioned, samples were taken manually and analysed in the laboratory. Figure 5 illustrates the rate of dehydrogenation as a function of vacuum treatment time. In the figure it can be seen that hydrogen is removed quite fast up to 10 minutes of treatment time. After 10 minutes the dehydrogenation time is slowed down. 20

29 Chapter 4 ppm H min 6 min 9 min 15 min 16 min Vacuum treatment time (minutes) Figure 5: Dehydrogenation during vacuum treatment. Table V shows the hydrogen concentration in the slag from heat DV Table V: The results from the slag analyses. Process time (min) Hydrogen concentration in the steel (ppm) Hydrogen concentration in the slag (ppm) Partition coefficient From table V it can be seen that the hydrogen concentration in the steel decreases faster than the hydrogen content in the slag phase. This is in accordance with a previous study Modeling work Hydroxyl capacity model Only binary interactions are considered for the model optimization. The optimized results are summarized in eq. (20). R T ln C X 3 y y 3 Al y H O 2 SiO Ca y y Al 2 = R T ln Ca y Mg Si 3 y y y y X y y y y X 3 ( y Al y Ca ) y Al y Mg ( y Al y Mg ) 3 ( y y ) y y ( y y ) Al Al 3 Si 4 Ca Ca Si 3 3 Al O Ca 2 3 Al 3 Mg Mg Mg Ca 3 Si CaO 3 Mg y Al y S i X MgO (20) 21

30 Results The calculated iso-lines of water capacities at 1873 K in the sections of 5 mass% MgO and 10 mass% MgO are presented in Figure 6a and 6b, respectively. The experimental data available in these two sections are also included in the figures. Data of the slags containing 4 and 7 mass% MgO are also included in Figure 6a. It is seen that the model predictions are in reasonable agreement with most of the experimental data. Only a few values differ somewhat from the model predictions. SiO 2 0 CaO x x x M J I G H E F N L D/K O x x x x x10 3 B C A A 3 1.0x10 3 B 3 1.1x10 3 C 3 1.1x10 3 D 3 3.0x10 3 (7% MgO) E 3 2.1x10 3 (7% MgO) F 3 1.9x10 3 (7% MgO) G 3 1.8x10 3 (4% MgO) H 3 1.5x10 3 (4% MgO) I 3 1.3x10 3 (4% MgO) J 3 1.1x10 3 (4% MgO) K 2.9x10 3 (7% MgO) L 1.7x10 3 M 1.1x10 3 N 3.8x10 3 O 2.9x Al 2 O 3 Figure 6a: Iso-lines of water capacity at 1873 K in the section of 5 mass% MgO. The experimental results available in the section are also included. SiO 2 CaO P Q R S/V 2.2x x x x x10 3 X Z UT Y x x x x x P 3 1.0x10 3 Q 3 1.0x10 3 R 3 1.2x10 3 S 3 0.9x10 3 T 1.9x10 3 U 2.0x10 3 V 1.0x10 3 X 1.2x10 3 Y 1.7x10 3 Z 2.2x Al 2 O 3 Figure 6b: Iso-lines of water capacity at 1873 K in the section of 10 mass% MgO. The experimental results available in the section are also included. 22

31 Chapter 4 Process model Figure 7 a-e presents the calculated hydrogen concentrations for the different heats as functions of the total process time. The experimentally determined hydrogen concentrations of the different heats are also presented. calculated value * measured value calculated value * measured value a b calculated value * measured value calculated value * measured value c d calculated value * measured value e Figure 7 a-e: Calculated and measured hydrogen concentration as a function of the total vacuum treatment time 23

32 Results From the figures it can be seen that the predicted hydrogen concentrations are in acceptable agreement with the experimentally determined concentrations. In all heats, the hydrogen concentrations decreased fast in the initial stages of the vacuum degassing operation. This is well described also by the model prediction. Thereafter, the hydrogen refinement proceeded at a low rate until termination of the process. The final predicted hydrogen concentrations are somewhat lower than the experimental values. 24

33 Chapter 5 DISCUSSION 5.1 Laboratory work Characterization of melting of slags A comparison of the present results with the phase diagrams suggested by Slag Atlas 18 shows that there is considerable disagreement. As an example, Figure 8 compares the present results with the section having 10 % MgO reproduced from Slag Atlas. 18 According to the phase diagram all of the investigated samples should have been melted at the experimental temperature. However, the quenched samples 5, 15, 16 and 17 are all multi-phase mixtures. Likewise, samples 2 and should have been melted according to the phase diagram in Slag Atlas 18, but according to the present investigation these samples consist of multi-phase mixtures. 25

34 Discussion Figure 8: Investigated compositions in the quaternary phase diagram Al 2 O 3 -CaO-MgO-SiO 2 with 10 mass% MgO. The closed squares represent compositions of single liquids and the open ones represent the compositions of multi-phase mixtures. Water vapour solubility The experimental results for the binary CaO-SiO 2 system are included in Figure 9 for comparison with literature data. It is seen in the figure that the present results agree well with the data of Ende et. al. 7 Note that even most of the data reported by Iguchi 8 follow the same trends as that of Ende et. al. 7 and the present study. On the other hand, the values by Wahlster et. al. 6 is in controversy with the rest of the other researchers. 7-8 Hence, the present experimental results along with values reported by Ende et. al 7 was employed in the optimization of the model parameters. 26

35 Chapter 5 C H2 O (ppm H2 O /atm1/2 ) Wahlster et al. [6] Ende et al. [7] Iguchi et al. [8] Present work (1853 K) Present work (1823 K) n CaO Figure 9: Experimental data for the CaO-SiO 2 system. The only experimental data for the water solubilities of CaO-Al 2 O 3 -MgO-SiO 2 quaternary slags were reported by Jo and Kim 3. They determined the water solubilities of a number of slags at 1873 K using an inert gas fusion technique with thermal conductivity detection. The slag compositions of these authors are mostly in the region of relatively high silica contents. Two of the slag compositions of the present work are comparable with the slags studied by Jo and Kim 3. A comparison between the water capacities reported by Jo and Kim 3 and the present work in the sections of 5 mass% MgO and 10 mass% MgO are presented in Figure 10a and 10b along with the present experimental values. Even the experimental results in the section of 7 mass% MgO are included in Figure 10a for comparison. The values obtained by the two research groups appear to follow the same trends. Figures 10a and 10b also show that water capacity does not vary substantially with composition, while an appreciable increase is noticed when the 3 3 slag composition approaches CaO saturation. A value of = 1 10 to 2 C 2 H O 10 prevails in the majority of the composition range. On the other hand, CH 2 O can 3 reach a value higher than 3 10, when the slag compositions are close to CaO saturation. 27

36 Discussion SiO Present work ( K) x x x x x I II V IV III /VI CaO Al 2 O Jo and Kim (1847 K) I 1.0x10 3 II 1.1x10 3 III 1.1x10 3 IV 1.9x10 3 V 2.1x10 3 VI 3.0x10 3 Figure 10a: Comparison of the experimental results between the present study and the study carried out by Jo and Kim 3 in the section of 10 mass% MgO. The values presented are the water capacities (10-3 ). Present work ( K) 1A 1.9x10 3 4B 1.0x x x10 3 9A 1.7x x SiO VII IIX 4B/X IX A 9A CaO Al 2 O 3 1A Jo and Kim (1873 K) VII 1.0x10 3 IIX 1.0x10 3 IX 1.2x10 3 X 0.9x10 3 Figure 10b: Comparison of the experimental results between the present study and the study carried out by Jo and Kim 3 in the section of 10 mass% MgO. The values presented are the water capacities (10-3 ). In many ladle refining processes, the industries operate using synthetic slags close to CaO saturation The addition of CaO containing raw materials would introduce considerable amount moisture into the furnace, especially when the humidity level is high. The higher water level introduced and higher CH 2 O of the ladle slag would result in high initial hydrogen content in ladle slag. According to Dor et. Al. 2, the water content in the slag after vacuum treatment could be

37 Chapter 5 times higher than the value predicted from the metal-gas-slag equilibrium. This high water content could be attributed to the slow dehydrogenation kinetics in the slag. 3 The high initial water content in the slag and the high capacity of the slag would make the ladle slag as a resource of hydrogen and therefore having great impact on the final hydrogen concentration of the liquid steel. In order to optimize the dehydrogenation process, the hydrogen transfer from slag to metal and therefore the water capacity of the ladle slag must be taken into consideration. 5.2 Industrial work The hydrogen content in the steel decreases fast in the initial stage of hydrogen removal, but slows down at the end of the treatment. The reason for this could be that there is a strong driving force for mass transport in the steel in the beginning of the dehydrogenation since there is a big difference of concentration. In the end, as the differences decreases, the driving force is small and the dehydrogenation rate is slowed down. If it could be acceptable to increase the hydrogen concentration with less than 1 ppm after the vacuum treatment the degassing time could be reduced to half. 5.3 Modeling work Hydroxyl capacity model Some of the experimental values differ somewhat from the model predictions. While the experimental uncertainty caused by the presence of small amount of solid phase could be the reason of the discrepancy, the limitation of the model when applied to the region very close to CaO saturation could be another plausible explanation. Nevertheless, the model prediction could be considered satisfactory. The water capacity increases with the increase of the CaO content irrespective of the MgO concentration. However, the increase of the water capacity with CaO content is not very substantial, only a factor of about ten, across the whole composition region of the liquid. This implies that a variation of the slag composition during ladle treatment would not vary the water capacity substantially. This finding would be useful for the optimization of the slag composition in ladle treatment. Still, high initial water content in the slag and a relatively high water capacity of the slag would make the ladle slag as a source of hydrogen and slow down the dehydrogenation process. The model would be a useful aid to the optimization of ladle slag. As an example of the application, Figure 11 presents the variation of water capacity at 1773 and 1873 K when Al 2 O 3 is replaced by SiO 2 addition, assuming that the slag contains 50 29

38 Discussion mass% CaO and 10 mass% MgO. The plot shows that the ability of the slag to pick up water increases as SiO 2 is replaced by Al 2 O 3. The figure also indicates that the temperature has little effect on the water capacity. Calculated C H2 O (ppmh 2 O/atm 1/2 ) K 1873 K Mass% Al 2 O 3 Figure 11: Variation of water capacity at 1773 and 1873 K when Al 2 O 3 is replaced by SiO 2 with 50 mass% CaO and 10 mass% MgO. Process model As can be seen in Figure 7 a-e the model predictions are somewhat lower than the experimental values in the final stage of the vacuum treatment. In the model calculation an attempt was made to take care of the slag acting as a source of hydrogen that can be transferred back into the liquid steel. However, assuming equilibrium between slag and metal can be one explanation why the final values of the model predictions are lower than the real data. Reaction (14) might be slow to reach equilibrium instantly. Moreover, the mass transfer in the slag, which is not considered in the present model, could also be slow. Another explanation for the difference could be the uncertainties involved in measurement of the initial content of OH - in the slag. The lower predicted hydrogen concentrations could also be due to the uncertainties in the thermodynamic data. Also, many input data are simplified for this first model version and could also have an impact on the final result. Nevertheless, the result of prediction is satisfactory for this preliminary model. The reasonable model prediction shows that a dynamic process model could be satisfactorily constructed using the results of CFD calculation. Since both steel and slag is protected from the surrounding atmosphere during vacuum degassing, there is no water transferring from the atmosphere to the slag. 30