Electromagnetic Separation of Nonmetallic Inclusion from Liquid Metal by Imposition of High Frequency Magnetic Field

Transcription

1 , pp Electromagnetic Separation of Nonmetallic Inclusion from Liquid Metal by Imposition of High Frequency Magnetic Field Koichi TAKAHASHI and Shoji TANIGUCHI 1) Doctor Course Student, Graduate School of Engineering, Tohoku University, Aoba-yama, Aoba-ku, Sendai Japan. 1) Graduate School of Environmental Studies, Tohoku University, Aoba-yama, Aoba-ku, Sendai Japan. (Received on November 0, 00; accepted in final form on February 18, 003 ) In this study, a method is investigated for separating inclusion particles from liquid metal by the electromagnetic force generated by an alternating magnetic field. To make clear the characteristics of the inclusion separation, a model experiment has been performed by the use of a SiC-liquid aluminum system in a high frequency induction furnace. The thickness of particle-accumulated layer formed on the side wall and its area fraction of particles are measured from the micrographs of the cross section of solidified aluminum sample. It is found that the growth of the particle-accumulated layer is completed in a short time (ten and several seconds), and that the thickness of the layer becomes smaller and the particle fraction in the layer becomes larger with increasing coil current. Mechanical stirring is found to retard particle separation in the case of small coil current. The maximum thickness of the particle-accumulated layer obtained for the 3 mass% addition of particle is almost same as the skin depth. To investigate the electromagnetic separation in the present system, a complete mixing model is made, which takes into account the effect of the particle-accumulated layer on the electromagnetic force distribution. This model is based on the Lavers theory with a simple modification of the change in apparent electric conductivity in the particle-accumulated layer. The estimated results of the change in particle concentration and the growth of particle-accumulated layer are in good agreement with the observed results. KEY WORDS: electromagnetic separation; liquid aluminum; clean metal; recycling; induction coil; removal of inclusion. 1. Introduction Nonmetallic inclusions in steel products cause various defects such as slivers in steel plates, rupture of thin wire during drawing, and short lifetime of bearing steel. The problems of inclusions are also seen in various metal products including aluminum, magnesium, copper, and their alloys. It is well known that the energy consumption for producing recycled aluminum is only 3 % compared with that for virgin aluminum. However, the wrought materials, which comprise 60 % of all aluminum products in Japan, need over 80 % virgin aluminum to dilute impurities and inclusions from recycled materials to avoid the harm in drawability. The amount of aluminum scrap becomes larger year by year and fine inclusions tend to be introduced from painted materials, and as a result, the difficulty of aluminum recycling increases. Various methods have been proposed for separating inclusion particles from liquid metal until now: flotation or sedimentation; ceramic filter; bubbling; and centrifugal separation. However, these methods are not effective for the case of inclusions such as very fine size and/or small difference in density between particle and metal. Innovative method for inclusion separation is strongly desired. Several researchers including one of the present authors proposed the application of electromagnetic force to inclusion separation. The methods of the imposition of the electromagnetic force are classified into the following six types: (a) simultaneous imposition of direct current and stationary magnetic field, 1 3) (b) imposition of alternating current, 4) (c) imposition of alternating magnetic field, 5 8) (d) imposition of traveling magnetic field, 9) (e) simultaneous imposition of alternating current and alternating magnetic field, 10,11) (f) imposition of stationary magnetic field. 1) In addition to these studies, Shu et al. 13) and Makarov et al. 14) recently investigated the availability of the electromagnetic separation theoretically. Among these methods, type (c) seems to be one of the most realizable methods because it does not need complicated equipments. However, it seems necessary to investigate the effect of fluid flow on the separation efficiency for practical application. The flow of liquid metal may play two important roles: (1) convey inclusions from bulk to the place where electromagnetic force is acting (positive role); () disturb separation by some fluid-dynamic forces (negative role). The purpose of the present study is to make clear the separation efficiencies of inclusion particles from liquid metal by the use of an alternating magnetic field considering the effect of fluid flow. In the experiments, silicon carbide (SiC) particles were electromagnetically separated from mechanically agitated liquid aluminum. 003 ISIJ 80

2 . Principle of Electromagnetic Separation Figure 1 shows the principle of electromagnetic separation. When a uniform electromagnetic force is applied to a liquid metal, the metal is compressed by the electromagnetic force and a pressure gradient is generated in the metal. The nonconductive particle suspended in the liquid metal receives only the pressure force because it does not experience the electromagnetic body force. As a result, the particle is forced to move in the opposite direction of electromagnetic force. Based on the Leenov Kolin s theory, 15) a nonconductive spherical particle receives a force in a conductive fluid with the imposition of a uniform electromagnetic force. This force is expressed by d FP 3 π 4 6 P 3 F...(1) Fig. 1. Principle of electromagnetic separation. where F P is the force acting on a particle, F imposed electromagnetic force, d P particle diameter. As this force acts inversely against the electromagnetic force, it can be applied for the separation of nonconductive particles from liquid metal. Figure shows the schematic diagram of the electromagnetic separation of nonconductive particles from a cylindrical liquid metal by using an induction coil, which is the system investigated in this study. A high frequency-alternating current supplied to a cylindrical coil generates an alternating magnetic field and induces an eddy current in the cylindrical liquid metal. The interaction between the induced current and the imposed magnetic field generates an electromagnetic force which always directs to the centerline of the liquid metal. The nonconductive particle in liquid metal receives a force in the opposite direction to the imposed electromagnetic force as shown in Fig. 1. As a result, the particle moves to the side face of the liquid metal and therefore they can be separated. Fig.. Electromagnetic separation of nonconductive particles from a cylindrical liquid metal in an induction furnace. Fig. 3. Experimental apparatus used for the electromagnetic separation of SiC particles from liquid Al under mechanical agitation. 3. Experiment 3.1. Experimental Procedure Figure 3 indicates the schematic of experimental apparatus used for the electromagnetic separation of SiC particles from liquid aluminum. The apparatus is composed of a high frequency generator with 30 khz in frequency, a 15 turn induction coil with 94 mm diameter, a silica (SiO ) crucible with 40 mm inside diameter, a mechanical agitator of a flat plate with 10 mm width, 50 mm length, and 3 mm thickness, and eight spray nozzles for water cooling aligned around the crucible. The experimental procedures are as follows kg of aluminum was inductively melted in the crucible. The power of the generator was turned off when the melt temperature reached K. A piece of aluminum alloy containing SiC particles was put into the melt and agitated during about 30 s. The power was then turned on again, and after a prefixed time the melt was cooled rapidly by the spray-cooling unit. For the case with mechanical agitation, the melt was agitated by rotating flat plate with constant speed during the electromagnetic separation. The solidified aluminum was cut and polished, and observed by an optical microscope for the distribution of SiC particles. In some conditions, temperature was measured by a thermocouple at the center of the melt in order to know the solidification time. The experimental conditions are shown in Table 1. The size distribution of SiC particle was measured for the particles extracted from aluminum alloy by sodium hydroxide solution. The mean diameter obtained from the size distrib ISIJ

3 Table 1. Experimental conditions. Fig. 5. Photograph of horizontal cross section of solidified aluminum (I rms 163 A, C mass% t h 16 s). Fig. 4. Change in melt temperature with time at different coil currents (t s 10 s). ution was 0.5 mm in the volume-mean diameter. 3.. Estimation of Solidification Time In order to know the exact period effective for the electromagnetic separation, it is needed to estimate the solidification process. In this study, a one-dimensional heat conduction equation was used for the estimation. ρc P T 1 T λr Q...() t r r r where C P is specific heat, T temperature, l thermal conductivity, Q Joule heat. The heat of solidification of aluminum was included in the specific heat. Boundary conditions were set at the axis, inner and outer surfaces of the crucible. Equation () was solved by using finite difference method. The heat transfer coefficient for spray cooling was obtained by adjusting the calculated cooling curve with the observed one. Figure 4 shows the change in the metal temperature with time from the beginning of the spray cooling at two different coil currents. Although the initial temperature of the melt is K, it decreased to 933 K due to the addition of low temperature aluminum alloy and 30 s agitation. Melt temperature is kept constant during 0 30 s, then dropped suddenly after the completion of solidification. The solidification time is longer for I rms 163 A than for I max 0A because of the imposition of Joule heat. The calculated curves agree well with the observed curves. Based on these calculations, the holding time for the electromagnetic separation was estimated as the time that the solidified shell reached the skin depth (d 1/ pm 0s f 1.49 mm, where m 0 is magnetic permeability of vacuum, s electrical conductivity, f frequency). Estimated holding times, t h, were 6 8 s for t s s, s for t s 10 s, and s for t s 30 s. The difference in holding time for the same t s is due to the difference Fig. 6. Optical micrographs of vertical cross section of solidified Al near the crucible wall at the middle position (I rms 163 A, C mass%). in the coil current. The separation procedure of the SiC particles from liquid aluminum will be analyzed based on the holding time as the accurate effective time for the electromagnetic separation in each experimental condition. 4. Experimental Results Figure 5 shows the optical micrograph of the horizontal cross section of solidified aluminum cylinder. It is seen from the figure that the thickness of the particle accumulated layer is uniform along the circumference. Figure 6 shows the optical micrographs of the vertical cross section near the crucible wall at the middle height of the cylinder at different holding times. It is seen from the figure that a considerable amount of SiC particles are collected at wall even in 8 s. The average thickness of particle-accumulated layer along vertical distance, d p, is found to be saturated at 16 s as shown in Fig. 6. Figure 7 shows the magnified optical micrographs of the particle-accumulated layer and the inner region of solidified aluminum at different coil currents, 163 and 85 A. Particle-accumulated layer at higher coil current 003 ISIJ 8

4 Fig. 7. Magnified micrographs of the particle-accumulated layer and inner region of solidified aluminum for different coil currents (t h s, C mass%). Fig. 9. Change in mean thickness of particle-accumulated layer, d p and area fraction of SiC particle, e p with agitation speed for (a) I rms 85 A and (b) I rms 163 A (t h s, C mass%). Fig. 8. Change in mean thickness of particle-accumulated layer, d p and area fraction of SiC particle, e p with time for (a) I rms 85 A and (b) I rms 163 A (C mass%). is denser than that at smaller coil current. On the other hand, it can be seen from Figs. 7(c) and 7(d) that there are few particles in the inner region of aluminum cylinder for both currents. To evaluate the characteristic of separation, the area fraction of particles, e p, and the average thickness, d p, of the particle-accumulated layer, were measured from the photographs of cross section. Figure 8 indicates the change in d p and e P with time at two different coil currents. It is found from the results of d p that the electromagnetic separation is saturated after ten and a few seconds for both coil currents. The saturated value of d p is small and the value of e P is large for large coil current. From these results, it is concluded that a strong electromagnetic force compresses the particle-accumulated layer, which results in a high efficiency for the removal of inclusions. However, it should be noted that the strong liquid metal flow is generated at high coil current, and it may prevent the electromagnetic separation, which will be discussed in Chap. 6. Figure 9 shows the effect of the liquid flow on the electromagnetic separation. In the case with agitation at weak current, there can be seen a marked decrease in d p with increasing agitation speed (Fig. 9(a)). However, for strong current, d p does not change with agitation speed (Fig. 9(b)). It is understood from the above results that the electromagnetic separation is retarded by fluid flow if the separation force is weak. Figure 10 gives the optical micrographs of the cross section near the crucible wall for different particle concentrations. In the figure, values of the mean thickness of the particle-accumulated layer and the scale of the skin depth (d 1.49 mm) are indicated. It is noteworthy that the particle-accumulated layer can be locally formed a little over the skin depth at C mass%. At the same particle concentration, a lot of free particles are observed in the inner region of aluminum melt. Considering these facts, the limiting thickness of the particle-accumulated layer seems to be up to the skin depth ISIJ

5 Fig. 11. Schematic of complete mixing model. Fig. 10. Optical micrographs of vertical cross section of solidified Al near the crucible wall at the middle position for different particle concentrations (I rms 163 A, t h 36 s). 5. Mathematical Models 5.1. Outline of Complete Mixing Model Considering strong electromagnetic stirring, it seems reasonable to apply a complete mixing model to predict the rate of the inclusion separation. The separation rate is obtained easily as a product of the particle concentration (uniform) and the migration velocity due to electromagnetic force perpendicular to the side wall and the bottom. The electromagnetic migration velocity of particle is obtained by equating the migration force expressed by Eq. (1) to the Stokes viscous force. For example, the radial migration velocity is expressed by dp ue Fr...(3) 4µ where m is viscosity of melt, F r radial component of electromagnetic force. As SiC particle has larger density than liquid aluminum, it goes down in the melt with the Stokes terminal velocity. Fig. 1. Size distribution of SiC particles extracted from aluminum. d P ( ρp ρ) g v t...(4) 18µ where r P and r are the density of particle and melt, respectively. Figure 11 shows the schematic diagram of the complete mixing model. The conservation equation of particle in molten aluminum is given by V dc [ AWue AB( ve vt)] C...(5) dt where V is volume of melt, C particle concentration, A W area of side wall, A B area of bottom, ū e averaged radial velocity at side wall, and v e averaged vertical velocity at bottom. Due to the skin effect, the electromagnetic force decreases rapidly with the distance from wall surface at high frequencies. Accordingly, the radial migration velocity should be taken at the front of the particle-accumulated layer. So, ū e in Eq. (5) is replaced by ū e,d p that is the migration velocity at the front of particle-accumulated layer. The thickness of the layer, d p, was obtained from the total mass of particles transferred into the layer. In the calculation, the particle size distribution was divided into 1-size groups from 0.5 to 51 mm as shown in Fig. 1 and the change in the concentration of each size group was estimated by solving Eq. (5) numerically. Furthermore, the thickness of the layer was estimated by using the observed particle-area fraction in the particle-accumulated layer. The electromagnetic force, F, in the melt was calculated by the modified Lavers theory 11) to take into account of the existence of the particle-accumulated layer. In the original theory, it is difficult to consider the local change in electrical conductivity like the particle-accumulated layer of which conductivity is reduced by the existence of nonconductive particles. The change in distribution of electromagnetic force in the vicinity of the particle-accumulated layer is shown schematically in Fig. 13. The electrical conductivity of particle-accumulated layer, s Al SiC, was assumed to be proportional to a volume fraction of liquid, (1 e p ). Neglecting the electrical conductivity of SiC particle, the following equation is obtained. s Al SiC (1 e p )s Al...(6) The modification of the Lavers theory and calculation procedure are shown in Appendix. 003 ISIJ 84

6 Fig. 13. Radial distribution of electromagnetic force taking into account of the particle-accumulated layer. 5.. Fluid Flow Analysis (without Mechanical Stirring) In order to investigate the profile of particle accumulated layer, flow field of liquid aluminum in an induction furnace was estimated numerically by using the electromagnetic force field calculated by the Laver s theory. In the fluid flow calculation, the existence of particles as well as the particleaccumulated layer was not taken into consideration. The continuity equation, Navier-Stokes equation for time-averaged velocities and k e equations were solved numerically. The method of the calculations was same as that adopted in the previous study. 17) Fig. 14. Comparison between calculated and observed results of C and d p for (a) I rms 85 A and (b) I rms 163 A. Fig. 15. Comparison between three calculated conditions: (a) considering both of the change in electromagnetic force and the distribution of particle-size, (b) neglecting the change in electromagnetic force, (c) neglecting the distribution of particle-size. 6. Discussion Figures 14(a) and 14(b) show the calculated results of the change in d p and C with time for I rms 85 and 163 A, respectively, for the case without mechanical agitation. There can be seen reasonable agreement between observed and calculated results. This agreement indicates that liquid metal is actually well mixed like a complete mixing by the electromagnetic stirring. The flow of liquid metal carries particles from the bulk liquid to the region near the wall where the electromagnetic force is effective, and as a result particle separation is promoted. There can be seen a saturation in each particle-separation curve. There are two reasons for the saturation. One is that the electromagnetic force at the front of particle-accumulated layer becomes small as the thickness of the layer increases. The second is that large particles are separated in the early stage but fine particles are remained in bulk liquid because of weak separation forces. To examine the major reason of the saturation, calculations were made for the following three conditions: (a) considering both items, (b) neglecting the change in electromagnetic force, (c) neglecting the distribution of particle-size, that is, using a mean diameter (0.5 mm). Figure 15 gives the results of these three calculations. From the figure, the particle-size distribution is found to be the major reason of the saturation of particle separation. The reason of the small influence of the existence of the particle-accumulated layer on the separation rate is attributed to the small difference between the mean electrical conductivity of the layer and liquid aluminum. Figures 16(a) and 16(b) indicate the calculated removal efficiency of each particle-size group at different holding times for I rms 85 and 163 A, respectively. The removal efficiency of each size group, h i, is defined by the following equation. Ci0 Ci η i...(7) Ci0 It is found from the figure that the removal efficiencies of larger particles reach 100 % in a short time. On the contrary, the efficiencies are small for smaller particles even in a long holding time. In the case of high current imposition, the time for getting large efficiency becomes short. In order to remove small particles like several microns, it seems effective to enlarge particle size by enhancing coagulation ISIJ

7 Fig. 17. Efficiency map of electromagnetic separation of inclusion particle obtained from the experimental results (t s 10 s). Fig. 16. Predicted removal efficiency as a function of particle diameter at different holding times for (a) I rms 85 A and (b) I rms 163 A. with a strong agitation, 18) and then separate electromagnetically. For the case with mechanical agitation, the disturbing effect of agitation on separation was investigated under various conditions of agitation speeds and imposed currents. The results are summarized in Fig. 17 that indicates the efficiency map of electromagnetic separation of inclusion particle. The removal efficiency, h, was obtained from the ratio of the amount of separated particles calculated from the values of d p and e p to the total amount of particles put in the melt. The figure clearly indicates that h increases with increasing coil current and decreases with increasing agitation speed. There can be seen a critical condition between separation and non-separation, which seems useful not only for designing the inclusion separator but also for producing particle-reinforced metal-matrix composites by the use of induction furnace. Furthermore, this result is suggestive to avoid the inclusion accumulation in the initially solidified shell for the electromagnetic casting of steel (EMCS) in which electromagnetic force is applied to support solidified shell in the continuous casting mold and reduce the friction between the mold and the shell. 19) Figure 18 shows the whole photographs of particle-accumulated layer near the crucible wall at high particle concentration (a), and the flow pattern calculated by the k e model (b). Comparing Fig. 18(a) with 18(b), it is found that the thickness of particle-accumulated layer becomes thin with closing to top and bottom where liquid velocities are high. Fig. 18. Comparison between experimental result and calculated flow pattern (C 0.1 mass%, I rms 163 A, t h 36 s). And also, there can be seen a re-entrainment of separated particles from wall region to bulk liquid at the upper part of the area A. These observations combined with simulated results reveal that the strong flow locally retards the particle separation. However, the mechanism of the retardation of particle separation is not clear at present. Although the dynamic pressure of turbulent fluctuation or the lift force due to velocity gradient is conceivable, further studies are needed to make clear the mechanism of retardation. 003 ISIJ 86

8 7. Conclusion The electromagnetic separation of SiC particles (0.5 mm in mean diameter) from liquid aluminum has been investigated to obtain detailed aspects for electromagnetic separation in an alternating magnetic field. As a result, the electromagnetic separation proceeded very quickly and electromagnetic force formed a particle-accumulated layer adjacent to crucible wall. The particle-accumulated layer was denser as coil current was larger. Strong electromagnetic force could achieve the highly efficient removal of inclusion. In the case that the melt was mechanically agitated, the particle separation was retarded, which seemed to be applicable to avoid inclusion accumulation in an initially solidified shell in EMCS. The characteristics of particle separation were reproduced well by the mathematical model for the case without the melt agitation if the particlesize distribution was taken into account. The effect of flow on the retardation of inclusion separation should be elucidated in the future study. Acknowledgements Authors are grateful to Mr. Kjell Odland, Hydro Inc., Norway for offering SiC-included aluminum alloy. This study was partly supported by the Ministry of Education, Science, Sports and Culture, Grant-in-Aid for Scientific Research on Priority Areas ( ). REFERENCES 1) P. Marty and A. Alemany: Proc. of Symp. of IUTAM, The Metals Society, London, (1984), 45. ) J. Park, J. Morihira, K. Sassa and S. Asai: Tetsu-to-Hagané, 80 (1994), ) E.-P. Yoon, J.-H. Kim, J.-P. Choi and H.-R. Kwon: J. Mater. Sci. Lett., 1 (00), ) S. Taniguchi and J. K. Brimacombe: ISIJ Int., 34 (1994), 7. 5) N. El-Kaddah, A. D. Patel and T. T. Natarajan: JOM, 47 (1995), 46. 6) F. Yamao, K. Sassa, K. Iwai and S. Asai: Tetsu-to-Hagané, 83 (1997), 30. 7) D. Shu, B. Sun, K. Li, T. Li, Z. Xu and Y. Zhou: Mater. Lett., 55 (00), 3. 8) K. Li, J. Wang, D. Shu, T. X. Li, B. D. Sun and Y. H. Zhou: Mater. Lett., 56 (00), 15. 9) Y. Tanaka, K. Sassa, K. Iwai and S. Asai: Tetsu-to-Hagané, 81 (1995), ) A. Alemany, J. P. Argous, J. Barbet, M. Ivanes, R. Moreau and S. Poinsot: French patent No , (1980). 11) S. Taniguchi and A. Kikuchi: Proc. 3rd Int. Symp. on EPM, EPM000, ISIJ, Tokyo, (000), ) B. Pillin and P. Gillon: Proc. nd Int. Symp. on EPM, EPM 97, Vol. 1, Paris, (1997), ) D. Shu, B. D. Sun, J. Wang, T. X. Li and Y. H. Zhou: Metall. Mater. Trans., 30A (1999), ) S. Makarov, R. Ludwig and D. Apelian: IEEE Trans. Magn., 36 (000), ) D. Leenov and A. Kolin: J. Chem. Phys., (1954), ) J. D. Lavers and P. P. Biringer: Elektrowärme, 9 (1970), 3. 17) S. Taniguchi, A. Kikuchi and S. Kobayashi: Mater. Trans., JIM, 37 (1996), ) T. Nakaoka, S. Taniguchi, K. Matsumoto and S. T. Johansen: ISIJ Int., 41 (001), ) K. Ayata, K. Miyazawa, E. Takeuchi, N. Bessho, H. Mori and H. Tozawa: The 3rd Int. Symp. on Electromagnetic Processing of Materials, EPM 000, ISIJ, Tokyo, (000), 376. Appendix. Approximate Treatment of Particle-accumulated Layer It is impossible to deal with the change in apparent electric conductivity due to the particle-accumulated layer in the original Lavers theory. In the present study, an approximate treatment was performed based on a one dimensional model expressed by Eqs. (A-1) to (A-3): B B jωb 1 1 r 1...(A-1) µσ 0 r r r µσ 0 x x 0: B B 0...(A-) B B x δ p : B B, (A-3) µσ 0 1 x µσ 0 x where x r 1 r, B is the z-component of magnetic flux density, suffix 1 and represent liquid Al and particle-accumulated layer, respectively. The values of magnetic flux density, B, induced current, J, and electromagnetic force, F, are obtained by the following equations by solving the above equations and Maxwell equations. x d p : B 1 C 1 exp( h 1 x)...(a-4) 1 db1 C1η 1 J1 exp( η 1x)...(A-5) µ e dx µ e 1 F1 Re{ J1 B 1 }...(A-6) x d p : B C exp(h x) C 3 exp( h x)...(a-7) 1 db Cη C3η J exp( ηx) exp( η x) µ e dx µ e µ e...(a-8) 1 F Re{ J B }...(A-9) where the symbols in the equations are shown as follows: δ1 ; δ ωµ 0σ1 ωµ 0σ ; h 1 (1 j)/d 1 ; h (1 j)/d...(a-10) ηδ 1 p Be 0 σδ 1 1 C1 ηδ p ηδ p ηδ p ηδ p σδ 1 1( e e ) σδ ( e e ) ηδ p Be 0 ( σδ σδ 1 1) C ηδ p ηδ p ηδ p ηδ p σδ 1 1( e e ) σδ ( e e ) ηδ p Be 0 ( σδ σδ 1 1) C3 ηδ p ηδ p ηδ p ηδ p σδ 1 1( e e ) σδ ( e e )...(A-11)...(A-1)...(A-13) The value of s is obtained from s 1 by the following equation (same as Eq. (6)). s (1 e p )s 1...(A-14) The electromagnetic force at the wall obtained by the above equations was adjusted with the average force on the side wall obtained by the Lavers theory. The existence of particle layer on the bottom was not taken into account because of its small amount ISIJ