Implementation of Slag Engineering Techniques at CO-Steel Raritan to Improve Melting and Refining Practices

Implementation of Slag Engineering Techniques at CO-Steel Raritan to Improve Melting and Refining Practices Don LeMar Co-Steel Raritan 225 Elm St. Perth Amboy, NJ 08862 Eugene Pretorius Baker Refractories 232 E. Market St. York, PA 17403 Key Words: Foaming, Slag, Models, Inclusions, Clogging INTRODUCTION Foamy slag principles 1 and foamy slag models were utilized to evaluate the EAF practice at Co-Steel Raritan, and to propose changes to the practice that could result in improved productivity and cost savings. Significant improvements were realized using this approach and the results are highlighted in the paper. This study also demonstrates that there are other factors besides good slag management that can have a significant influence on foaming consistency in the furnace and productivity. The importance of a predictable carbon content at melt-in is discussed, and the raw materials required to ensure predictability are highlighted. The validity of the old maxim "First make slag, then make steel" was reaffirmed by improvements that were obtained by engineering and creating ladle slags that not only improved the quality of the steel produced but also the consistency in desulfurization, castability and refractory compatibility. Different slag practices for lowcarbon and high-carbon grades were developed because of different processing and castability requirements. PLANT BACKGROUND CO-Steel Raritan is equipped with a 120 MVA, 135 ton Fuchs EBT furnace with one (1) oxy-fuel burners in the sump area, a side wall mounted oxygen lance with carbon injection, and a Fuchs door oxygen lance. The typical oxygen consumption is 89,100 scft/t per heat at a maximum injection rate of 1300 scfm/minute. Typically, 15 pounds of carbon per ton of steel is injected in the furnace per heat at a rate of 60 to 100 pounds/minute. Tap to tap time is between 60 and 72 minutes. The Arc Furnace shares a one-arm turret with the Ladle furnace. The ladle furnace feeds a five strand 130 mm Concast casting machine which typically cast at a speed of 120 inches per minute. The type of steel produced at CO-Steel Raritan is Carbon and Rod grades ranging between ASTM 1005 to ASTM 1080. Applications include, Mesh products, industrial, plating, cold heading, cold finish, welding, and high carbon wire qualities. The alloy and flux additions are added during tap and argon stirring with a purge plug is used at the ladle furnace. The LRF station is equipped to add alloys and fluxes from a hopper system into the ladle during the refining and heating period. The LRF station is also equipped to add carbon and Ca wire to the heat during refining and before sending it to the caster.

EAF PRACTICES The practices in the EAF can be separated into three distinct groups: 1) Practice A - "Old" practice with Pig Iron (PI) in the charge (before August 1999) 2) Practice B - Modified Practice with Pig Iron (PI) in the charge (August 1999 through March 2000) 3) Practice C - Modified Practice without Pig Iron (PI) April 2000 to present The typical flux additions in the EAF for the "Old" and the "Modified" practice are summarized in Table I. Practice A (Old practice) Table I. Flux and carbon additions to the EAF for different pratices Additions Practice A Practice B Practice C Lime amount (lbs/ton.) 95 30 30 Dolomitic Lime amount (lbs/ton) 0 60 60 Pig Iron 10-15%* 10-15%* 0 Charge Carbon (lbs/ton) 0 0 18-32* Injection Carbon (lbs/ton) 16** 14 14 * Pig iron adjusted by grade depending on residual allowable maximums and additional carbon content needed. ** Practice A injection material was a 50-50% blend of anthracite coal and lime stone. In practice B and C 100% anthracite coal is used as an injection carbon. Good slag foaming early on in the heat and poor foaming towards the end of the heat were typically observed when using the old practice. This resulted in poor thermal efficiency and significant arc-flare damage to the furnace roof and sidewalls at the end of refine. The furnace slags were sampled frequently and the average composition of the slag samples taken from the beginning of 1999 until August is listed in Table II. Table II. Average composition of slags samples taken during the old practice (Practice A) Average Std. Dev. Relative Dev. Min Max % MgO 3.3 0.83 25.2 1.9 5.0 % CaO 36.1 5.27 14.6 27.3 47.0 % FeO x 38.1 9.80 25.7 18.4 54.4 % Al 2 O 3 3.8 0.85 22.0 2.5 5.4 % SiO 2 8.4 3.54 40.8 3.9 17.8 % MnO 4.1 0.92 22.3 2.5 6.3 % CaO/% SiO 2 4.16 B 3 Ratio 2.88 %CaO/%MgO 10.9 Relative Deviation (Coefficient of variation) = Std. Dev./Average x 100 Only the major elements are reported in the table The standard deviations and the relative deviation values of the slags in Table II show considerable variation and it is expected that the foaming behavior in the furnace would have been very erratic as a result. The relative deviation of SiO 2 is especially high (>40%) compared to the other major components in the slag (< 26%) which might be indication of variability in the Si levels of one of the charge materials, most likely the pig iron.

The average slag composition of Practice A was evaluated using a slag foaming program and the resulting Isothermal Stability Diagram (ISD) is shown in Figure 1. Figure 1. ISD of the average slag composition of the old practice (Practice A) In this figure the x-axis labeled "FeO" actually represents FeO + MnO (this the case for all the ISD's presented in this paper). From this figure it is clear that once the combined FeO + MnO level of the slag (Practice A) exceeds about 37% at 2912 F (1600 C), the slag composition moves into the all-liquid area of the diagram and foaming behavior will be very poor. In the all-liquid area the slag is too liquid to sustain a foam. (The background and interpretation of ISD's is discussed in a previous publication 1 ). The MgO content of the slag is too low to provide any buffer against high FeO slags. Furthermore, these slags are MgO unsaturated, which can cause significant wear on the upper banks and slagline refractory materials. When slag foaming is poor, the injected carbon is not very effective to reduce the FeO from the slag and hence the FeO content of the slag increases rapidly. Not only is the foaming behavior of these slags poor but considerable amounts of Fe is lost to the slag, leading to metallic yield loss. If such a high-feo slag is not removed from the furnace it could negatively impact a number of heats following it, as the residual slag from a previous heat forms a large part (>50%) of the slag for the following heat. The evaluation of the slags with the ISD program clearly showed that the slags from the Practice A were not ideal in terms of foaming properties and refractory protection. Two scenarios were available to improve the slag composition in the furnace: Option A: Target MgO saturated slags but maintain the same slag basicity as the old practice Option B: Target lower basicity slags while maintaining MgO saturation Option A: MgO saturation and the same basicity From the Foamy Slag program and its inherent mass-balance calculations, the extra amounts of MgO were calculated to move the slag composition from its current position into the magnesio-wustite + liquid area (MW +L) of the diagram. A supersaturation of 1 1.5% MgO is normally targeted (1.5% supersaturation = 1.5% above the MgO saturation value). The new target slag is plotted in Figure 2.

Figure 2. ISD showing the new slag composition at MgO saturation and the same slag basicity (Option A) The program calculates that for an initial slag weight of 30,000 pounds, an addition of approximately 1500 lb. of magnesia would be required to achieve of 8% MgO level in the slag (1.4% supersaturation). Alternatively, the same slag composition can be attained by decreasing the lime in the charge by 2280 pounds and increasing the dolomitic lime addition by 3750 pounds (still a net increase of approximately 1500 pounds). Figure 2 shows that the foaming properties of the MgO-saturated slag to be much less sensitive to variations in FeO levels in the slag, and that the slag would be compatible with the refractories over a large composition range. The drawback of this approach is that the slag required considerable amounts of FeO to reach the optimum foaming region (> 30%), which does not improve the yield in the furnace. Furthermore, the addition of more flux materials to the furnace will also increase the energy consumption per heat. Although, this option was by far the more conservative approach, it was not followed because of the tradeoff between benefits and drawbacks. Option B: Lower basicity slags and MgO saturation From evaluating different options using the Foamy Slag program and mass-balance calculations, it was quickly realized that significant benefits could be realized by targeting less basic slags in the furnace. The benefits are: i) A decrease in the amount of basic oxides charged to the furnace ii) iii) The slag could start to foam much earlier in the heat, since much less FeO would be required to move the slag into the optimum foaming region If the foaming improved, the amount of FeO in the slag might be decreased (better carbon injection utilization) so that the overall yield might be improved. iv) If the amount of FeO in the slag decreased as expected, then the total amount of slag formed per heat could be decreased by approximately 25%, a potential energy saving. The calculated new amount of slag was still sufficient to completely cover the arcs. v) Improved energy efficiency and refractory performance because of the improved foaming behavior in the furnace (consistency and longevity) Because of the perceived benefits of the Option B approach, a number of trials were conducted aiming for MgO-saturated slags, at basicities lower than the original practice (Practice A).

Practice B (Modified practice with pig-iron in the charge) The initial flux recipe (Trial #1) aimed for slags with a basicity (B 3 = CaO/[SiO 2 +Al 2 O 3 ]) of approximately 1.6 and a MgO supersaturation of 2.0. The average composition of slag samples taken during this trial is listed in Table III and the ISD for this slag composition is shown in Figure 3. While these slags exhibited excellent foaming properties, the general feelings were that these slags coated the furnace too well and could lead to bottom build-up. It was also felt that the slags might be a little too acidic for comfort which made the practice vulnerable to changing Si levels in the charge materials. At these low basicity levels the MgO solubility varies greatly for small changes in basicity. For example, if the basicity decreased from 1.6 to 1.4, it would result in an increase in MgO solubility of about 2% 1. Table III. Average composition of slags samples taken during Trial #1 of Practice B Average Std. Dev. Relative Dev. % MgO 12.8 1.31 10.2 % CaO 31.9 2.33 7.3 % FeO x 23.6 2.71 11.5 % Al 2 O 3 5.4 0.67 12.3 % SiO 2 15.2 1.68 11.1 % MnO 5.1 0.58 11.4 % CaO/% SiO 2 2.1 B 3 Ratio 1.6 %CaO/%MgO 2.2 Figure 3. ISD showing the average slag composition of Trial #1, Practice B A second flux recipe (Trial #2) was designed to target slags with a basicity (B 3 ) of approximately 1.8 and a MgO supersaturation of 1.3. These slags also foamed very well, but had a slightly higher fluidity, which made removal from the furnace a little easier. This recipe became the standard additions for Practice B. The average composition of the slag samples taken during this trial is listed in Table IV and the ISD for this slag is shown in Figure 4.

Table IV. Average composition of slags samples taken during Trial #2 of Practice B Average Std. Dev. Relative Dev. % MgO 10.1 0.99 9.8 % CaO 32.0 1.04 3.2 % FeO x 30.3 2.06 6.8 % Al 2 O 3 5.1 0.52 10.0 % SiO 2 12.9 1.29 10.0 % MnO 4.4 0.40 9.2 % CaO/% SiO 2 2.49 B 3 Ratio 1.78 %CaO/%MgO 3.2 Figure 4. ISD showing the average slag composition of Trial #1, Practice B Not only did the new slag practice result in a significant improvement in foaming longevity and consistency, it also significantly lowered the final FeO level is the slag, as expected. The FeO level in the slag decreased from 38.1% with Practice A (Table 1) to 23.6% for Trial #1 (Table III) and 30.3% for Trial #2 (Table IV). The operational and refractory improvements that were observed by switching from Practice A to the modified practice (Practice B) are summarized in Table V. The results on the modified practice in the Table V represent plant data for the 8-month period, August 1999 to March 2000. Table V. Comparison of operational and refractory performance numbers for the old and new practice. Operational Parameter Old Practice Modified Practice Difference Tap to Tap time mins/ht 72 65-7 Total Electric Energy input (kw/h) 425 461 +31 Heats per 24 hour day 17 19.5 +2.5 Total furnace refractory cost $3.03 $2.09 -$0.94 Production rate (tons per hour) 112.5 124.6 +12.1 Heats between shell changes 250 530 +280 Oxygen consumption(scfm/ton) 750 660-90 Average tap temperature ( F) 2920 3050 +130

From the table it appears that Practice B increased the Kilowatts per hour consumption. However, the excellent foaming slag allowed the furnace to be operated at higher tap settings with more energy input and significantly higher tap temperatures. This resulted in a reduction of tap to tap time and total oxygen consumption. More energy was coming from the electrical input and conversely less from the chemical input. Additionally, as scrap prices climbed through the end of the year, Co-steel looked for alternative low residual iron sources. In late January HBI Trails were begun. Kilowatt per hour consumption increased as its usage increased. Still the slag foaming was excellent. In February, it was decided to increase furnace electrical power by dropping a reactor tap. This contributed to a record pace performance in March and a bounce of kilowatt per hour consumption. From the results of the two trials shown so far, it is clear that the slag recipe could probably still be further optimized. However, the results of the second trail recipe was so positive, it was decided not to make any more changes for the short term. Practice C (Modified practice with no pig-iron in the charge) In April 2000, the price of pig-iron became considerably more than the price of clean low residual scrap and the decision was made to remove the pig-iron from the charge mix for economic reasons. The carbon units that were originally provided by the pig-iron, now would be added in the form of charge carbon with the scrap. At the same time more HBI had to be added to the mix as a substitute for the clean iron units lost with the elimination of pig iron. The initial trials with charge carbon and increased HBI were a disaster. To complicate the problem, anthracite coal had to be delivered pneumatically to a bin on the charge floor. This limited the maximum size of the material. The quality and sizing of the charge carbon initially delivered by the vendor was poor, and contained excessive amounts of fine material. During these initial trials with charge carbon, the foaming in the furnace was very inconsistent and the same problems that were the norm with Practice A (very liquid slag), became a common occurrence with the new slag practice (Practice B). The recovery of carbon in the bath was very inconsistent which resulted in the heat being over blown and high FeO slags being generated. Furthermore, the chemical energy contribution of the Si in the pig-iron was now absent in the furnace charge which by itself resulted in higher electrical energy requirements. A C-O-FeO model is utilized to demonstrate the effect of melt-in carbon and injection carbon on the FeO content in the slag. The model simulates the change in metal and slag compositions based on the oxygen and carbon injected into the furnace and includes the effect of reduced decarburisation due to mass transfer limitations at lower carbon content. These calculations are aimed at demonstrating the principles and are not necessarily representative of CO-Steel Raritan's practices. Three general scenario's were evaluated: Case I. Ideal conditions where 13% pig-iron was part of the charge, foaming was good, and carbon injection rate was adequate to reduce the FeO that was generated by the oxygen injection. Figure 5 shows the C content in the metal and FeO content of the slag as a function of time. A carbon content of 0.6% C was assumed at the beginning of flat-bath conditions. Most of this carbon (>90%) originates from the pig-iron charge. The inflections on the FeO curve are due to changes in carbon injection rates

0.600 0.500 %C in Metal %FeO in slag 60 55 50 0.400 45 0.300 40 35 0.200 30 0.100 25 20 0.000 15 0 5 10 15 20 25 Time (minutes) Figure 5. Changes in steel and slag chemistry as a function of time. Carbon content at flat-bath is 0.6% and sufficient carbon is injected into the slag to control the FeO levels. Case II. In this example, the carbon content at flat-bath is also 0.6% but no carbon was injected into the slag to reduce the FeO that is formed during the oxygen blow period. The amount and the rate of oxygen injection is identical to Case I. Figure 6 shows that the FeO content of the slag rapidly increases at the point where the carbon removal from the bath becomes mass-transfer controlled. 0.600 0.500 0.400 %C in Metal %FeO in Slag 55 50 45 40 0.300 35 0.200 0.100 30 25 20 0.000 15 0 5 10 15 20 25 Time (minutes) Figure 6. Changes in steel and slag chemistry as a function of time. Carbon content at flat-bath is 0.6% but no carbon is injected into the slag to control the FeO levels.

Case III. In this example calculation, the amount and rate of oxygen injection is identical to the previous two calculations. The amount and the rate of carbon injection is also identical to that of Case I. The main difference in this calculation is that the carbon content of the steel at flat-bath conditions is not 0.6% C, but only 0.2% C. This simulates the conditions in Practice C where the pig-iron was removed from the charge mix and the carbon units were added in the form of charge carbon. However, actual carbon recovery from the charge carbon into the bath was much less than expected. Figure 7 shows the change in C level in the steel and FeO level in the slag, if the carbon content at flat-bath conditions is only 0.2%C instead of the 0.6%C as expected. 0.600 0.500 0.400 %C in Metal %FeO in Slag 55 50 45 40 0.300 35 0.200 0.100 30 25 20 0.000 15 0 5 10 15 20 25 Time (minutes) Figure 7. Changes in steel and slag chemistry as a function of time. Carbon content at flat-bath conditions is only 0.2%C instead of 0.6%C as expected. The above example (Case III) assumes that good foaming still occurred and that the injected carbon was still effective in reducing the FeO from the slag. However, in real operations the foaming may have been poor, which in turn renders the carbon injection ineffective so that the actual FeO levels in the slag may be much higher than calculated. From the above example it is clear that very erratic foaming and process conditions (over-blowing or underblowing) can occur if the carbon content at flat-bath conditions is not predictable and consistent. A number of slag samples were taken during the early trials with various sources of charge carbon and the results are shown in Table VI. The lime and dolomitic lime additions for Practice B and C were identical, as confirmed by similar %CaO/%MgO ratios of both sets of analyzed slags. However, the basicity of the slags of Practice C was higher than Practice B. When the pig-iron was removed from the charge in Practice C, a charge material high in Si was replaced with scrap with a lower Si content. The amounts of basic oxides added to the charge were not adjusted for the changes in charge chemistry and hence the higher basicity of these slags. It is important to note that the slags also had significantly higher FeO levels than the practice B slags, indicating possible overblowing of the heats. The foaming consistency of the initial heats was also very erratic until good quality charge carbon was acquired. The sizing and quality of the charge carbon is very important to achieve good and consistent recoveries of carbon in the steel. After numerous trials it was found that charge carbon with the sizing of 3/4 x 5/16 gave the best results. The most significant issue in the improvement of the charge carbon

practice was the control of the percent of fines in the material. There were two factors contributing to a high fine content: 1) Poor vendor quality control; and 2) pneumatic delivery piping system with numerous 90-degree bends. Vendors corrected the quality problem with pressure from the mill. The piping system was reconfigured to eliminate the 90-degree bends and minimize the 45-degree bends. Table VI. Average composition of slags samples taken during early trials with Practice C Average Std. Dev. Relative Dev. % MgO 9.3 1.0 10.4 % CaO 28.3 2.1 7.5 % FeO x 39.0 3.5 9.0 % Al 2 O 3 4.2 0.6 15.0 % SiO 2 10.2 0.8 8.3 % MnO 3.5 0.3 9.2 % CaO/% SiO 2 2.79 B 3 Ratio 1.97 %CaO/%MgO 3.1 Once the charge carbon sizing problems were eliminated, the foaming in the furnace and productivity in the shop improved as shown by the results in Table VII. However, the total electric energy input did not decrease much due to an increase in HBI usage from 10% to 15%. The statistics presented in Table VII were gathered from a representative number of heats in late spring and early summer when the changes occurred. Table VII. Comparison of operational numbers for Practice B with and without Pig Iron (PI). Operational Parameter With PI Without PI (Early Trials) Without PI (Improved Charge carbon) No. of heats in sample 483 320 564 Tap to Tap time min/ht 68.4 73.5 72.5 Tons per hour 115.8 108.4 113.4 Average heat size 132.0 132.8 137.0 HBI 7.5-10% 10% 15% Total Electric Energy input (kw/h) 461 484.4 482.4 LADLE REFINING PRACTICES When this study was initiated, the slagmaking philosophy was to aim for fairly stiff unreactive slags in order to minimize slag-metal-refractory interactions. While this approach resulted in good refractory performance, it did not always generate the best quality steel or contribute to improving the productivity in the shop. The flux additions for low-carbon and high-carbon steel were the following: Low-Carbon grades High-Carbon grades Additions During Tap 1500 lb. Lime (>96% CaO) 2000 lb. Desulf Mix (87% CaO, 5% MgO, 8% CaF 2 ) Additions at Ladle Furnace 500 lb. Lime 300 lb. Ca-Aluminate (55% Al 2 O 3, 35% CaO, 3.2% SiO 2 ) 300 lb. Lime 500 lb. Ca-Aluminate

The slags arrived at the ladle furnace very stiff and never became completely fluidized (too basic), even after extensive arcing and stirring. The slags were particularly stiff on a new tap-hole in the EAF, and gradually became more "workable" as the tap-hole aged and more slag was carried over into the ladle. A mass-balance program 2 was used to calculate the slag composition in the ladle based on the flux and alloy additions during tap, historical alloy recoveries, and estimated amounts of carryover slag from the EAF. The calculated slag compositions are summarized in Table VIII. For both the low-carbon and high-carbon grades, the slags were about 1000 pounds over-saturated with lime. These calculations certainly confirmed what was observed at the ladle furnace. The compositions of a number of slag samples taken during this practice were evaluated using a Slag Optimization Program 3 and were also found to be significantly over saturated with respect to lime (500 to 700 pounds of CaO for a 4000 lb. slag weight at 2912 F). It is probable that the analyzed slag samples were not completely representative of the bulk slag chemistry, as it is very difficult to take a representative pipe-sample when the slag is not completely liquid. Table VIII. Calculated ladle slag compositions from the old practice Low-C Grades Recommended Slag* High-C Grades Recommended Slag* % MgO 3.6 5.0 5.9 8.1 % CaO 64.6 50.9 65.9 54.8 % FeO 2.1 2.9 1.9 1.6 % Al 2 O 3 6.8 9.4 8.9 12.3 % SiO 2 21.1 29.3 11.7 16.2 % MnO 1.7 2.4 1.6 1.3 % CaF 2 4.1 5.6 Excess Lime 1095 1046 *The model option used to determine the recommended slag aims for a liquid limesaturated slag in order to determine the amount of excess lime only. The following is a summary of the problems encountered using the old slag practice: Long processing times at the ladle furnace trying to fluidize a stiff slag and desulfurizing the steel. During this practice the heats were tapped cold from the furnace (2920 F) and a temperature increase of about 70 F had to be achieved at the ladle furnace. Loss of productivity as the caster would plug off strands to pace the ladle furnace rather than the arc furnace. The furnace would wait until the caster received the heat so that the turret could be cleared to take the next tap. Poor and inconsistent desulfurization. Occasional slag build-up in the bottom of the ladle because the slag was too stiff to dump after teeming. The high-carbon heats were prone to clogging at the caster. On average about 20% of the high-carbon heats had some kind of castability issues resulting in pourbacks, scrap billets, and at a minimum, downgrade product. In light of the problems listed above and the improvements that were achieved on the furnace slags, it was decided to attempt to improve the slag practice at the LFS. The philosophy of "first make slag and then make steel" was implemented. To achieve this goal, mass-balance calculations and slag models were utilized in designing slag compositions and flux addition recipes to generate liquid slags upon arrival at the LFS. The option of just decreasing the lime addition during tap to generate more liquid slags was not appropriate because it would have resulted in a too low slag volume in the ladle. Sufficient slag volume is an important slag

requirement in the ladle to cover the arcs and protect the refractories from arc-flare damage. In order to generate a slag of sufficient volume and desired fluidity it is important to add the right balance of fluidizing oxides (SiO 2, Al 2 O 3, CaF 2 ) and refractory oxides (CaO and MgO). It is also highly recommended to add the fluidizing oxides in a blended form with the refractory oxides. The addition of a fluxing oxide, such as fluorspar, by itself (unblended) can result in extensive localized refractory erosion. The choice of fluidizing oxides is very important in terms of steel quality and castability requirements. This was clearly demonstrated in early trials with fluxing mixtures from different vendors. The use of Al 2 O 3 as a flux or Ca-aluminate in some of the flux mixtures resulted in severe clogging of the high-c grades. This will be discussed in more detail later. For economic reasons, increasing amounts of scrap with high sulfur levels are used for the low-carbon steel grades. A slag with good desulfurization properties is therefore an important requirement for these scrap mixes. However, desulfurization of these grades in practice is not an easy task because the high oxygen level in the steel (>35 ppm) is not ideal for sulfur removal. The only options left to improve desulfurization is to enhance the kinetics of the desulfurization reaction (early liquid slag formation and vigorous stirring) and designing very basic slags with good sulfide capacities. To achieve these goals, a combination of SiO 2 (8%), fluorspar (11%) and Al 2 O 3 (11.5%) the latter in the form of Ca-aluminate, are used as fluxing oxides in the material mix. Furthermore, the components are blended with the basic oxides and added as a 3000 pound super-sack to the ladle during tap. While this mixture cost more than the original recipe, the consistent desulfurization, better thermal efficiency, and improved steel quality far outweigh the added expense. The initial target composition of the slag has been designed to be slightly lime unsaturated for most conditions in the ladle (i.e., even when the tap-hole is new and the amount of EAF carryover slag is minimal). The ladle furnace operator monitors the slag throughout the heat for color and fluidity and adds the required amounts of slag deoxidants and lime to the slag, as needed. When the tap-hole is new, only 200-300 pounds of lime might be required. However, for an old tap-hole, more than 500 pounds might be required to obtain a slag with a "creamy" consistency in the ladle. The new slag practice added the task of "taking care of the slag" to the ladle operator's responsibility. However, with proper training and experiencing the benefits of the new slags, resulted in a rapid and successful transition to the new practice. The target ladle slag composition for the low-carbon grades is listed in Table IX. The initial target slag composition is a composite of 1000 pounds of carryover slag, steel deoxidation products, tap-hole sand, and 3000 pounds of a blended "desulfurization" mixture. This initial slag composition is slightly lime unsaturated to ensure rapid and complete melting. Table IX. Target slag compositions for the new ladle slag practice. Low-C Grades (Initial Slag) Low-C Grades (Final Slag) High-C Grades (Initial Slag) High-C Grades (Final Slag) % MgO 7.0 6.4 5.2 4.6 % CaO 49.5 53.8 52.4 58.1 % FeO 2.0 <2.0 0.9 <0.9 % Al 2 O 3 9.9 9.1 1.5 1.3 % SiO 2 23.7 21.7 26.1 23.0 % MnO 1.5 <1.5 1.0 <1.0 % CaF 2 7.5 6.9 12.9 11.3 % Na 2 O < 5% <5% Additional Lime Req. 430 590

The fluxing oxides used in the mixtures for the high-c grades can have a profound impact on the castability of the steel. The effect the slag composition has on the equilibrium oxygen content of the steel is very important. The oxygen content is essentially determined by the following equilibrium: [Si] + 2[O] = (SiO 2 )...(1) where [ ] denotes in the steel and ( ) in the slag The equilibrium constant for this reaction is: a ( SiO2 ) K eq =...(2) 2 a[ Si] * a [ O] This equation can be rewritten as follows by using the 1 wt% reference state in iron and the activity coefficients of Si and O: 1/ 2 a (SiO2) [%O] =...(3) 2 [%Si]* fsi * fo * K eq For high-c steel with a specific silicon specification, the equilibrium oxygen content ([%O]) will therefore be dependent on the activity of silica in the slag. The silica (SiO 2 ) in the slag originates from the SiO 2 in the carryover slag, the SiO 2 from the deoxidation products, and SiO 2 in the desulfurization mixture. If the desulfurization mixture contains insufficient levels of SiO 2, the activity of SiO 2 in the slag will be low. This could result in the oxygen content of the steel being driven down too low for good castability as shown below. Furthermore, if the desulfurization mixture contains appreciable amounts of Al 2 O 3 (in whatever form) the equilibrium reaction between Al and Al 2 O 3 also becomes important: 2[Al] + 3[O] = (Al 2 O 3 )...(4) Reactions (1) and (4) can be combined as follows: So that: [Si] + (Al 2 O 3 ) = (SiO 2 ) + 2[Al] + [O]...(5) [% Si]* a( Al * 2O3 ) Const [% Al] =...(6) a * [% O] ( SiO 2 ) From equation (6) it is clear that the amount of Al in the steel will increase as a Al 2 O 3!, a SiO 2 ", and [%O] ". When the slag contains appreciable amounts of Al 2 O 3, and its SiO 2 content is low, if given enough time, the Al content in the steel might approach levels that could lead to clogging at the caster. As the solubility product of Al 2 O 3 decreases with decreasing temperature, the formation of Al 2 O 3 tends to occur in the SEN or mould, particularly where turbulent or uneven flow conditions with large temperature drops prevail. It is therefore important to minimize the amount of Al 2 O 3 in the mixture but also ensure that it contains sufficient amounts of

SiO 2. A trial with a flux mixture containing approximately 12.8% Al 2 O 3 and only 2.2% SiO 2, proposed by a vendor for the high-carbon grades, resulted in severe clogging on the high-carbon grades. The target ladle slag composition for the high-carbon grades is listed in Table IX. The initial target slag composition is also lime deficient to ensure rapid and complete melting of the slag. The MgO levels are deliberately targeted low to minimize the potential for any spinel formation in the steel. This flux mixture (3000 pounds) contains very low levels of Al 2 O 3 (0.3%) but sufficient levels of other fluxes SiO 2 (19.2%), CaF 2 (17.5%), and Na 2 O (3.2%), to ensure rapid liquid formation in the ladle. The new slag practices listed above have been successfully implemented at CO-Steel Raritan. The following is a list of improvements that were observed with these practices: Improved surface quality on all grades of steel Improved and predictable desulfurization for the low and high-carbon grades Inclusion content reduction and improved inclusion distribution. Permanent elimination of billet grinding and conditioning facility at a saving of $1,000,000 per year with the expectation of continued participation in the cold heading market. A decrease in kwh/t at the LFS of approximately 50% The amount of Ca-wire added to the low-carbon heats was decreased by 50% 100% castability of the high-carbon grades Increased productivity because the LFS was eliminated as the bottle-neck for production throughput. SUMMARY This paper demonstrates that significant process improvements can be realized by applying sound thermodynamic and slag engineering principles. It also highlights the benefits of using slag models and massbalance calculations to evaluate existing slag practices and designing new practices. The most important factor for success is the willingness of the steelmaker to be receptive to new ideas and implement the necessary changes. ACKNOWLEDGEMENTS The authors are grateful to the CO-Steel Corp. for permission to publish this paper. We would especially like to thank Greg Greiss who initiated this project on the furnace, Dave Ehrhart who extended the project to the ladle, and Helmut Oltmann for reviewing the paper (all from Baker Refractories). REFERENCES 1. E.B. Pretorius and R.C. Carlisle, "Foamy slag fundamentals and their practical application to electric furnace steelmaking," 56 th Electric Furnace Conference, 1998. 2. E.B. Pretorius and R. Marr, "The effect of slag modeling to improve steelmaking processes," 53 rd Electric Furnace Conference, 1995. 3. E.B. Pretorius and R. Marr, "Computer modeling of Refractory/Slag/Metal interactions," 5 th International Conference on Molten Slags, Fluxes and Salts. Australia, 1997.