Viscosity Measurements of Some Mould Flux Slags

Transcription

1 , pp Viscosity Measurements of Some Mould Flux Slags Mikael PERSSON, 1) Marten GÖRNERUP 2) and Seshadri SEETHARAMAN 1) 1) Department of Materials Science and Engineering, Royal Institute of Technology, Brinellvägen 23, Stockholm, Sweden. 2) Metsol AB, Karlav.2, Stockholm, Sweden. (Received on March 15, 2007; accepted on June 11, 2007) Continuous casting has been the dominating process for steel casting over the past decades. During the process, mould fluxes are added to enable a smooth functioning of the process, enabling better process performance and products with less defects. The viscosity of the mould flux slag is a key parameter determining the optimum casting conditions. Several experimental studies have earlier been carried out in order to determine viscosity data for mould flux slags, both industrial ones as well as synthetic slags with compositions close to industrial mould fluxes. However, the continuous evolving of new steel grades, casting dimensions and product quality in the steel industry also demands better control and development of the mould fluxes. In industrial practice for clean steel production, the Al 2 O 3 pick up has generally been observed to be about 2 4%. In view of this, the present study was initiated to experimentally investigate the viscosity of mould fluxes used in Swedish steel industry and the effect of dissolution of alumina in the same. The industrial implications of the slag viscosities measured in the present work are discussed. Viscosities of mould fluxes for continuous casting in steel production have been measured by the rotating cylinder method. Seven industrial mould fluxes, with different compositions, used were included in the study. The effect of the Al 2 O 3 content in the mould fluxes was also investigated. Even relatively small additions of Al 2 O 3 show a significant increase in viscosity. The measurements were carried out in the temperature range of to K. KEY WORDS: viscosity; mould flux; continuous casting; slags. I. Introduction Continuous casting has been the dominating process for steel casting over the past decades. During the process, mould fluxes are added to enable a smooth functioning of the process. Mould fluxes play an important role in continuous casting, enabling better process performance and products with less defects. The mould flux has mainly four functions 1,2) ; protect the steel from oxidation, lubrication at the mould/metal interface, provide the optimum level of horizontal heat transfer and absorb inclusions from the steel. The viscosity of the mould flux slag is a key parameter determining the optimum casting conditions. An empirical rule usually used to predict the cast conditions is based on casting speed and viscosity. Several experimental studies have earlier been carried out in order to determine viscosity data for mould flux slags, 3 12) both industrial ones as well as synthetic slags with compositions close to industrial mould fluxes. Almost all mould powders contain CaF 2 as it lowers the viscosity and melting point of the fluxes. The fluorine content varies normally between 4 and 10 wt%, but some powders can contain as high as 20 wt% of fluoride. The high volatility of the fluorine species is a problem both from process as well as environmental considerations. The fluoride evaporation from mould flux slags could also be a problem in the case of viscosity measurements, since the flux composition may vary with time during the measurements due to the formation of gaseous fluorine compounds. The fluoride vapor loss from mould fluxes has earlier been investigated in the Division of Material Process Science. 13) However, the fluorine volatilization is not the only factor affecting the flux composition with time. It is well known liquid mould slag layer between the steel and the mould flux powder bed, is a receptor of oxide inclusions. Along with the evaporation of fluorine species, absorbtion of inclusions is also likely to affect the chemical composition of the solidifying slag film, formed between the solidified steel shell and copper mould, thereby changing the physical properties influencing the lubrication capability during the casting. These are important aspects to consider when modeling the casting process. The continuous evolving of new steel grades, casting dimensions and product quality in the steel industry also demands better control and development of the mould fluxes. Here, the physical properties are of great importance. In view of this, the present study was initiated to experimentally investigate the viscosity of seven mould fluxes used in Swedish steel industry and the effect of dissolution of alumina in the same ISIJ

2 2. Experimental 2.1. Materials and Sample Preparation The mould powders were decarburized by heating in a muffle furnace at K for 48 h. The decarburized powders were premelted in an induction furnace. A graphite crucible was heated to K before approximately 300 g of mould powder was added and kept in the temperature range of K for min to ensure a completely molten slag. The melted slag was poured in to an iron crucible used in the viscosity measurement. Alumina was added to two of the mould powders. The Al 2 O 3 used was of 99.7% purity supplied by Sigma Aldrich Chemie. The mould powder and Al 2 O 3 powder were thoroughly mixed before pre-melting. The crucible dimensions are shown in Table 1. Argon gas was used both during the melting of the mould flux slag and subsequent solidification. The argon gas used during viscosity measurements and premelting was % pure and supplied by AGA Gas (Stockholm, Sweden). Table 1. Dimensions of iron crucible and spindles Experimental Apparatus and Procedure The viscosity measurements were carried out using the rotating cylinder method, with a Brookfield digital viscometer (Model RVDV-III, full-scale torque N m). The viscometer was rigidly mounted on a movable platform above the furnace so that it can be aligned to the crucible. A metal flange-teflon bellows coupling was used to close the gap between viscometer and alumina reaction chamber. A high temperature furnace system, supplied by Thermal Technology Inc (Laboratory Furnace Group 1000 series) with a maximum temperature limit of K was used for this study. The furnace was controlled using an Eurotherm (Model 818) controller as well as an optical pyrometer. The temperature was monitored by high temperature thermocouple, Pt 30%Rh/Pt 6%Rh, during the experiments, which was positioned just below the crucible. The experimental setup for viscosity measurements is shown in Fig. 1. The crucible containing the premelted slag was placed in the even temperature zone of the furnace. The spindle was then mounted on the iron transducer and lowered into alumina tube. The iron spindle dimensions are given in Table 1. To avoid oxidation of iron components and contamination of the sample, the system was purged with argon gas, which was rigorously purified by means of a gas-purification train for approximately 1 h before starting the heating procedure. As shown in Fig. 2, the gas was passed through silica gel and Mg(ClO 4 ) 2 for the removal of moisture. The CO 2 content in the argon gas is lowered by passing the gas through a column of ascarite. Oxygen impurity in the argon gas was removed by using columns of copper turnings kept at 973 K and magnesium turnings at 773 K. The heating rate employed during all the experiments was 5 K/min and the gas flow was kept at 0.2 NL/min. The furnace was heated up to K and left for 1 h to completely melt the slag and stabilize the temperature. The spindle was then carefully lowered into the slag, the tip being kept about 2 cm above the crucible base and about 1 cm of the shaft immersed in the slag. For evaluation of the slag viscosity, five rotation rates were used at each temperature and the Fig. 2. Fig. 1. Experimental setup for viscosity measurements. Illustration of the gas cleaning system connected to the furnace ISIJ 1534

3 Table 2. Mould Flux composition provided by supplier, wt%. time to reach equilibrium for viscosity measurements at each speed was estimated to be 120 s. This ensured that the molten slag was Newtonian. The rotating speeds used in the measurements varied between 8 to 70 rpm. To obtain thermal equilibrium at each measuring point, the temperature was kept constant for 30 min before each measurement. The major part of the measurements was carried out during cooling cycles, but to verify reproducibility, some measurements were also performed during heating sequences. Two of the slags were also selected for a second viscosity evaluation in new crucibles, in order to confirm the reproducibility of the results. The viscometer was calibrated using several kinds of mineral oil standards with viscosities in the range of Pa s at 298 K. In view of the possibility of the fluoride volatilization during the experiments, samples were taken at two different stages of the experimental sequence, viz. after premelting as well as after the viscosity measurements. In order to ensure that the sampling procedure was reliable and that the slags were homogenous without concentration gradients along the height of the crucible, especially with reference to fluoride, three samples were taken from the solidified slag mass after furnace cooling, viz. one sample 1 cm below the surface, one in the middle, and one 1 cm above the bottom of the crucible. All samples were subjected to chemical analysis after the experiments. The oxide contents were analyzed using X-ray fluorescence spectroscopy and the fluorine contents by electrode spectrometry after dissolving in NaOH solution. 3. Results In the present work, viscosity measurements were carried out in the case of seven proprietary mould flux slags used in different Swedish steel industries. The analyses of these fluxes provided by the suppliers are presented in Table 2. The listed values are averages of the maximum and minimum compositions guaranteed by the supplier. The span of this composition range can be relatively large for certain components. To two of these fluxes, viz. slags A and D, Al 2 O 3 was added and the corresponding viscosities were measured. The results of the chemical analysis of the premelted slags and the post measurement samples taken at different levels in the crucible are presented in Table 3. In general, the weight loss of the samples during the experiments varied between 0.1 to 0.5%. A comparison between supplier s analyses and the analyses after sample preparation, decarburization and premelting, does not show any obvious change in composition, apart from the carbon loss. It is seen in Table 3 that there is a small increase of FeO content in the slag during eight of the measurements. This is thought to be contributed by iron components. It should also be mentioned, a relatively large increase was discovered in three of the experiments. The cause of this is unclear. All the experiments were conducted under identical conditions. The crucibles were inspected after the measurements and no sign of oxidation was apparent on the crucible walls. The possibility of the solubility of metallic iron in the molten slag was considered in the present work. In view of the presence of iron only in the case of specific slag compositions, there is a possibility that the solubility of iron could be higher in the case of these compositions. Further work is being carried out in this regard. The chemical composition carried out post measurement is considered to be in better agreement with the slag composition during the viscosity measurement then the chemical analysis obtained previous to the measurements. The results of the viscosity measurements are presented in Table 4. The viscosity data in this table are the average values obtained from the results obtained using five different rotation speeds. The deviation of the experimental data from the mean value is less then 1% for all the measurements. The highest temperature of the measurements was ISIJ

4 Table 3. Mould Flux composition obtained in the present work, wt%. limited by the use of iron components, while the lower temperature limit for measurements is determined by the appearance of the two-phase region as marked by the drastic increase in the viscosity values. The lower temperature limit is almost the same for all of the mould fluxes in the range of to K except in the case of slag G, for which the lower temperature limit was around K. Repetition of the viscosity measurements of some of the slags, slag A and E, showed a maximum deviation of 3% and often less. This deviation is relatively small, considering the experimental uncertainties usually associated with viscosity measurements. It should also be pointed out that the second measurements in the case of the two slags were carried out with new slag samples and new crucibles. The samples from repetition experiments were not subjected to chemical analysis. The viscosities of mould flux slags for continuous casting have been studied by a number of researchers. 3 8) The viscosity values reported in these references are in most cases considerably higher than those found in the present study. This diversity is attributed to the differences in the chemical compositions. The earlier mould fluxes reported in literature had greater amounts of SiO 2 and Al 2 O 3 ; and lesser amounts of alkali oxides and CaF 2. It is well-known that Al 2 O 3 and SiO 2 have an increasing effect on slag viscosities, due to the tendency to promote the formation of the silicate network. On the contrary, CaF 2 and alkali oxides have the opposite effect on the viscosities due to the breaking of the silicate network. The measured viscosities show for most slags large deviations from the supplier s viscosity data. The supplier s viscosity data are generally estimated by the equation suggested by Riboud et al. 14) at one or several fixed temperature, which is only able to predict the slag viscosities approximately. No measured information is given on the viscosity temperature dependency or crystallization temperature. In the present work, special attention was paid to the 2007 ISIJ 1536

5 Table 4. Measured viscosity values of various slags. vapor loss of fluoride. The chemical analyses, in Table 3, show that the difference in the fluorine contents in the preand post measurement samples was negligible. This is further supported by the small weight loss during the measurements. The samples could safely be considered as homogenous as the chemical analyses of the samples taken at various levels showed negligible differences. 4. Discussion Mould flux slag basicity is usually in the range of 0.7 and ) Higher basicity leads to a higher degree of crystallinity of the mould flux slag. The degree of crystallinity, in turn, will affect the heat flux through the slag film. Higher degree of crystallinity will lead to lesser heat transport, while the glassy phase will result in higher thermal conductivity. Further, viscosity will have a strong impact on the thickness of the slag layer between the mould and the steel. Mould fluxes C, D and E are used in the casting of low alloying steels, while flux A and F in the casting of high-alloyed steel grades. The primary criterion for the selection of the type of mould flux depends upon the steel grade, whether the steel will undergo the peritectic phase transformation, L d g, associated with a volume change, or not. Steel grades subjected to a strong peritectic reaction are ISIJ

6 Table 5. The liquidus temperatures of various slags as estimated by the second derivative method, and measured using DSC compared with the supplier data. Fig. 3. Viscosties of seven slags as function of temperature. Fig. 4. Viscosities measured at K plotted as function of (NBO/T) ratio. usually related to greater risks of irregular surface quality and depressions. These steel grades require good heat flux control equivalent with insulating properties of the mould flux such as high melting temperature and high degree of crystallinity. The thickness of the slag layer, affected by the viscosity, is also of importance controlling the heat flux. In the case of steel grades with insignificant impact of the peritectic reaction, the mould flux friction to mould and steel is normally the controlling parameter. A less crystalline slag, with lower basicity, and lower melting temperature would be more suitable in these cases. Slag B is a mould flux particularly used in the start up phase of the casting process. As seen in Table 5, it has a significant higher content of iron oxide compared to the others and a high content of F, which gives it the features of good wetting properties respectively low melting temperature and low viscosity. High contents of iron oxide are, however, not preferable to use with alumina killed steels. It is easily reduced and the mould flux properties may be altered, thus only preferable in the early stages of the casting. Figure 3 graphically represents the variation of viscosities with respect to temperature of the seven mould fluxes investigated. The variation in measured viscosities of slags can usually be attributed to the basicity of the slag; higher basicity gives a lower viscosity. However, the CaO/SiO 2 ratios in the case of mould fluxes will not give a satisfying explanation of the variation in measured viscosity. The complex composition of mould fluxes is probably the reason for this. The basicity of slags can also be described by the ratio of (CaO MgO)/(SiO 2 Al 2 O 3 ) and these ratios in better agreement with the obtained viscosity values, but still not good enough. Another common way to estimate the degree of depolymerisation in slags is the (NBO/T) ratio, non-bridging oxygens per tetrahedrally-coordinated atom. 15) The (NBO/T) ratio is given by the following equation: (NBO/T) 2[ x( CaO) x(mgo) x(feo) x(mno) x(na2o) x(k2o)] 61 ( f ) x( Fe2O3) 2x( Al2O 3) 2x(Fe2O3) x( SiO ) 2x( Al O ) 2x(Fe O ) x(tio ) 2x(P O ) where x is the mole fraction and f the fraction of Fe 3 with IV coordination. The Fe 2 O 3 was not chemically analyzed and in the calculation of the (NBO/T) ratio was all iron oxides assumed to be present as FeO. This assumption should be fairly accurate since the experiments were carried out in an atmosphere with low partial pressure of oxygen. The average slag compositions of the three different parts analyzed post measurements were used to calculate the (NBO/ T) ratio. The ratio of (NBO/T) is in good agreement with the measured viscosity, as seen in Fig. 4 where the viscosity measured at K is plotted as function of (NBO/T). A clear trend seen as the viscosity is decreasing with increasing ratio of (NBO/T), despite one exception found in case of slag B. Slag B should have a higher viscosity according to the NBO/T ratio. However, the fluoride content is not taken in account in the (NBO/T) and slag B contains a relatively high amount of CaF 2, which explains the low viscosity of slag B. Slag G was found to have a much higher viscosity. In this case, viscosity measurements were not possible below K in contrast to the other slags. Slag G has a signifi ISIJ 1538

7 Fig. 5. Viscosties of slag A with Al 2 O 3 additions as function of temperature. Fig. 7. The second derivative of the activation energy. 2 (Q/R)/ T 2, as function of temperature for slag A. Fig. 6. Viscosties of slag D with Al 2 O 3 additions as function of temperature. Fig. 8. The second derivative of the activation energy. 2 (Q/R)/ T 2, as function of temperature for slag A 2.5 wt% Al 2 O 3. cantly lesser amount of fluoride and Na 2 O, as well as the highest amount of Al 2 O 3. These compositional differences would contribute to higher viscosities. In the case of Figs. 5 and 6, the effect of addition of alumina on the viscosity values in the case of slags A and D has been presented. As expected, the viscosity values were found to increase with the decrease in temperature. In industrial practice for clean steel production, the Al 2 O 3 pick up has generally been observed to be about 2 4%. 16) In the present measurements, 2.5 and 5.0 wt% alumina were added to slags A and D (chemical analysis shows 2.5 and 4.6 wt% as well as 2.2 and 4.2 wt% respectively) and the viscosities were measured. The results presented in Figs. 5 and 6 show that viscosity increases significantly with increasing alumina content, the phenomenon being more prominent at lower temperatures. Such a relatively large increase in viscosity would affect the slag film thickness, hence the heat transfer, and the friction force acting on the solidified shell. 17) The present results agree well with a previous study on the effect of Al 2 O 3 addition to mould flux slags. 13) It is well-known that alumina contributes to the silicate network thereby causing an increase in viscosities. From the measured viscosity values, the Arrhenius activation energies for viscous flow for various temperature intervals have been evaluated. The Arrhenius activation energy at higher temperatures, where viscosity variation with temperature is not drastic could be approximated to a constant value, these are presented in Table 5. However, the activation energy is found to increase with decreasing temperature as the liquidus temperature was approached. Similar observations have been reported by Seetharaman et al. 18) These authors have reported that the second derivative of Fig. 9. The second derivative of the activation energy 2 (Q/R)/ T 2, as function of temperature for slag E. the activation energy for viscous flow with respect to temperature shows a break around the liquidus temperature of the slag. The second derivatives of the activation energies for viscous flow for some of mould flux slags employed in the present work are plotted as functions of temperature in Figs It is seen that, in the case of slags A, D and E, the second derivatives show the break points around K for the first two mentioned and somewhat higher for slag G. The liquids temperatures obtained from the second derivative method for all slags, are listed in Table 5, along with the DSC measurements, carried out at KTH, and supplier data. It should be noted that DSC measurements were not employed for all fluxes. The application of the second derivative method for the estimation of the liquidus temperatures depends upon the number of viscosity values available at different temperature intervals. Further, the scatter in the data would result in uncertainties. On the other hand, the ISIJ

8 measurements of liquidus temperatures of slags by DSC method are often very difficult and unreliable in view of the significant supercooling that the slags undergo, as indicated by lower values. DSC values, taken during heating cycles can show higher values due to compositional inhomogeneities. In this respect, the second derivative method appears to give somewhat reliable values of liquidus temperatures. The higher liquidus temperature of slag G compared to the other slags is supported by the second derivative method and the supplier data, in contrast to the DSC measurement. As pointed out, the viscosity of mould fluxes is an important parameter in the continuous casting process. However, in the continuing development of mould fluxes, the determination of other properties is also very important. Measurements of thermal diffusivities as well as rupture strength of the same slags have been carried out with the objective to develop a model capable of predicting mould flux properties and modeling of mould flux behavior in the mould. 19,20) A similar dependency between these investigated properties and the slag structure has been found, indicating a possibility of a connection between thermophysical and mechanical properties of slags. 5. Conclusion The rotating cylinder method was used in order to determine the viscosity of seven slags from the corresponding proprietary mould fluxes. Four additional measurements were also carried out with increased alumina contents. Small additions of alumina resulted in a relatively large increase in the viscosities of mould flux slags. The chemical analyses established that fluoride losses during the experiments were insignificant. From the experimental results, the Arrhenius activation energies for viscous flow at various temperature intervals were calculated. The activation energies were found to increase with decreasing temperature, an observation in conformity with an earlier work in this division. The second derivative of the activation energies for viscous flow, when plotted as functions of temperature showed break points corresponding to the liquidus temperatures of the slags. The industrial implications of the slag viscosities measured in the present work are discussed. Acknowledgements The financial support from Swedish Steel Producers Association (Jernkontoret) is gratefully acknowledged. REFERENCES 1) K. C. Mills and A. B. Fox: High Temp. Mater. Process., 22 (2003), ) K. C. Mills and A. B. Fox: ISIJ Inter., 43 (2003), ) T. Mukongo, P. C. Pistorius and A. M. Garbers-Craig: Ironmaking Steelmaking, 31 (2004), ) A. B. Fox, K. C. Mills, D. Lever, C. Bezerra, C. Valadares, I. Unamuno, J. J. Laraudogoitia and J. Gisby: ISIJ Int., 45 (2005), No. 7, ) M. D. Lanyi and C. J. Rosa: Metall. Trans., 12 (1981), ) K. C. Mills and B. J. Keene: Int. Met. Rev., 1, (1981), 21. 7) W. L. McCauley and D. Apelian: Can. Metall. Q, 20 (1981), ) K. C. Mills, A. Olusanya, R. Brooks, R. Morell and S. Bagha: Ironmaking Steelmaking, 15 (1988), ) H. Y. Chang, T. F. Lee and Y. C. Ko: ISS Trans, 4 (1984), ) H. Y. Chang, J. F. Lee and T. Ejima: Trans Iron Steel Inst. Jpn., 27 (1987), ) F. Shahbazian, Du. Sichen, K. C. Mills and S. Seetharaman: Ironmaking Steelmaking, 26 (1999), ) F. Shahbazian, Du. Sichen and S. Seetharaman: ISIJ Int., 39 (1999), ) F. Shahbazian: Ph.D. thesis, Royal Institute of Technology, Stockholm, (2001). 14) P. V. Riboud, Y. Roux, L. D. Lucas and H. Gaye: Fachber. Hüttenpraxis Metallweiterver, 19 (1981), ) B. O. Mysen, D. Virgo and C. M. Scarfe: Am. Mineral., 65 (1980), ) Report No. TO24-163, Swedish Steel Producers Association (Jernkontoret), (2005). 17) A. Yamauchi, T. Emi and S. Seetharaman: ISIJ Int., 42 (2002), ) S. Seetharaman, S. Sridhar, Du. Sichen and K. C. Mills: Metall. Mater. Trans. B, 31 (2000), ) M. Hayashi, R. A. Abas and S. Seetharaman: ISIJ Int., 44 (2004), ) Y. Umezawa, T. Matsushita and S. Seetharaman: ICS 2005 The 3rd Int. Cong. on the Science and Technology of Steelmaking, AIST, Warrendale, PA, USA, (2005), ISIJ 1540