J. I Phys. D Appl. Phys. 24 (1991) 658-1563, Printed in the UK The induction melting of glass Robert Ducharmet, Fred Scarfez, Phiroze Kapadiat and John Dowdensll t Department of Physics, University of Essex, Colchester CO4 3SQ UK $ Electroglass Ltd, 4 Brunel Road, Benfleet, Essex SS7 4PS, U K 5 Department of Mathematics. University of Essex, Colchester CO4 3SQ UK Received 21 May 1990, in final form 26 October 1990 Abstract. The possibility of using a coupled system of an induction coil and a metal susceptor for the purposes of melting glass is investigated. The magnetic field strength and the current density in the different regions of the furnace are calculated. Detailed consideration is then given to the influence of the glass on the electrical Characteristics of the furnace and to the problem of magnetic forces on the susceotor 1. Introduction The ohmic heating of glass furnaces by the passage of an electric current between carefully placed electrodes in the melt is a well established and growing technology in the glass industry (Scarfe 1980). Electroheat can also he generated w i n g an induction mi! and this is a we!! established method used by other industries. This paper examines the possibility that if the large metal susceptor is incorporated into the design of a glass furnace, then an alternating magnetic field may be used to raise its temperature and so melt t h e charge. One important advantage of using a susceptor coupled to an induction coil to melt glass, is that the furnace can be started up from cold. This method is therefore tailor made for applications where furnaces are not run on a continuous basis. Other advantages of electric over fuel-fired furnaces are that they are easily sealed to prevent the leakage of volatile emissions from the melt into the environment, are relatively maintenance free, and offer more control over the convection currents in the melt. A general mathematical formulation of the induction problem for a cylindrical tube in an axial magnetic field has been given by McLachlan and Meyers (1935). Their analysis has been developed further for particular application to glass furnaces. An additional problem considered here is the presence of a glass layer, which serves to protect the outer surface of the susceptor from oxidation. This layer is electrically conducting and the eddy currents generated inside it must therefore be taken into account. Maxwell’s equations are taken as the starting point and it is assumed that the thickness of the susceptor is at least several times larger than the depth of penetration of the coil‘s magnetic field into the susceptor. The magnetic field, the current density 1 To whom all correspondence should be addressed. 0022.3727191 1050658 + 06 503.50 0 1991 IOP Publishing Ltd and the power dissipation in the susceptor are calculated. Consideration is then given to the electrical characteristics of the furnace and to the problem of ’ magnetic forces on the susceptor. The problem of the eddy current heating of a composite cylindrical system has arisen before in connection with the manufacture of thermionic valves and has received consideration from Wright (1Y37). There are, however, two important differences between that problem and the one treated here. Firstly, a complex time dependence is introduced, and this results in considerable simplification of the analysis. Secondly, the glass laycr in an induction furnace is relatively much thicker than the outer envelope in a thermionic valve, and the ‘thin cylinder’ approximation used by Wright cannot, therefore, be invoked. 2. Mathematical formulation of !he preblen! Suppose a molybdenum susceptor, in the shape of a uniform coaxial cylindrical shell heated by an induction coil, is to be used for the purposes of melting glass in a cylindrical crucible. This is shown schematically in figure 1. The electric and magnetic fields generated inside the furnace by the coil are described by Maxwell’s equations: V 2 H = up,(aH/al) where the electric field strength (in Vm-’) will be denoted by E , H is the magnetic field strength (A m-I), J the current density (Am-*), pn the permeability of free space ( = 4n x lo-’ H m-’), U the electrical conductivity (Q-I m-l) and f the time ( s ) . The induction melting of glass J , and K O are complex and it is therefore useful to represent them in the form Ju(kri312)= F,(kr) = ber(kr) ibei(kr) + and + I Ko(kri’lZ) = F2(kr) = ker(kr) i kei(kr). The first derivatives of these functions can similarly be expressed as i3/*J;(kri3/*) = F ; (kr) = ber‘(kr)+i bei’(kr) and i’/2Kt)(kri”2)= F ; ( k r ) = ker’(kr)+ikei’(kr) . . . CloY sectlo“ figure 1. Cross section of an induction-heated glass furnace. Equation (1) can also be written as VZH= iwup,H when a separable solution of the form H = ffs(r, 8, z ) exp(iwf) , , . --- is assumed; (r, 0, z ) are cylindrical polar coordinates and w denotes the angular frequency of the coil. If the ratio of the length of the crucible wall to the inner radius c is a large number (of the order of IO), then a long cylinder approximation can be made and it may be assumed that the magnetic field lines run parallel to the axis. Under these conditions equation (3) can be expressed in the form d2H dr2 1dH . + -=I U ~ ~ W H r dr - (4) This is a special form of Bessel’s equation and its general solution is H = AJ,(kri’i2) + BKo(kri’/2) The theory of Bessel functions has been worked out in considerable detail (Watson 1966) and their values are therefore readily accessible from standard tables. In order to solve equation (4) for a composite cylindrical system consisting of N layers, a solution of the form of expression ( 5 ) must in general be written for each layer. The 2N arbitrary coefficients can then be evaluated from the conditions that the electric and magnetic fields are continuous across the N boundaries (McLachlan 1935, Wright 1937). If, however, the thickness b - a of the susceptor is large compared with the skin depth of molybdenum (as must be the case to ensure the best possible coupling between the coil and the susceptor) then the magnetic field strength in the region r < a will be negligible. The conduction currents in the glass inside the susceptor can therefore he ignored. The approximations dZH H, I d H Hb and d r 2 ?i2 r dr 6b (5) where J, and KOdenote Bessel functions of the first and second kind (McLachlan 1961, ch.7), the parameter k is defined by k = g2/2/6and where Hb= H ( b ) , show that for large values of the parameter b / 6 , the contribution of the middle term in expression (4) can be neglected. It is shown below that this condition is generally satisfied in the susceptor, for all coil frequencies of practical interest. Use of the coordinate transformation x = b - r thus results in the simplified equation describing the magnetic field in the susceptor. This has the general solution H = Cexp[-(I-i)x/6]+Dexp[(l-i)x/6] (7) where C and D are arbitrary complex coefficients. However, the term exp[(l - i)+/6] does not attenuate with depth. Therefore, since it has already been argued that H is negligible where r a , it is clear that the coefficient D must be set equal to zero. The foregoing considerations suggest the following solution to expression (4) for the induction furnace H= is called the penetration or skin depth. The values of ---- { HO r,>r>c A F , ( k r )+ B F 2 ( k r ) car> b (8) r< b C exp[ - (1 - i)x/S,] where 6, denotes the skin depth in the susceptor. 659 R Ducharme et a/ In order to complete the solution to Maxwell's equations for this problem, the current density and the electric field in both the glass layer and the susceptor must now be calculated. The only non-zero component of the current density J runs parallel to the 0 direction. Equation (2) can therefore be expressed in the much simplified form dH dr J = -- (9) On substituting expression (6) into this equation and recalling x = 6 - r, it can be seen that J = - rk&W+Bk&i(kr) czr>b C[(1- i)/~3~]exp[ -(1 -i)x/6$] IC' rsb where the subscript G denotes the glass layer. The corresponding electric field can be obtained by a simple application of Ohm's law and this gives .=-[ [AF;(kr) +BFi(kr)]kG/oG czr>b [C(1-i)/oS6,]exp[ - (1 -i)x/aS] r66. (10) The values of the unknown coefficients A, B and C may now be calculated from the boundary conditions on the electric and magnetic fields at the surfaces r = b and r = c. Clearly three conditions are required and ihcse may 'De cxprcssed as Ioliows. (i) The magnetic field H i s continuous at the boundary r = c H, = AF,(kc) + BF,(kc). (ii) The magnetic field H is continuous at the boundary r = 6 C = AFl(k6) + BF,(kb). (11) (iii) The electric field E is continuous a t the boundary r = 6 C5-I = 4 F ; ( k b ) + BF; ( k b ) where 5 = (1 + i)(o,/20c)"Z. (12) The solution to this set of equations for the coefficients A and B is given by and 'g 05 k 0 660 O lcml b 6 lrml Figure 2. The magnetic and electric fields (in dimensionless units) as a function of distance from the boundafy of the glass layer for different values of the phase: (a) w t = 0; (b) ut = in; (c) wf = in. with time. These graphs are based on the following values for each of the relevant furnace parameters. 6=0.35m ~G = c=0.4m w=2nx106s-' , i(jc2-:m-: a,=i.5xi06R ' m , L . A coil frequency of 1 MHz has been selected in order to ensure that the variations in E and H are as marked as possible. The graph illustrates that even with this high coil frequency value, the variation in the coil's magnetic field across the glass layer is quite small compared with its maximum value H,,, ( A H / & < 0.15). This result follows as a consequence of the fact that the thickness of the glass layer is several times smaller than the penetration depth of the coil's magnetic field in the glass. It is clear from the second set of curves, however, that a large variation in the electric field does occur. Indeed. the value of E can be up to eighty times larger in the glass than its maximum value in the metal. This indicates the existence of significant eddy currents in the melt and shows why the conductivity of the glass cannot be ignored. 3. The magnetic flux inside the furnace The magnetic flux QF can be calculated from the magnetic field strength H , using the integral expression Q = p, The value of the coefficient C may thus be obtained by substituting these results into expression (11). Figure 2 shows the development of the real parts of the electric and magnetic fields in the glass layer ''Oh 0. . H dS (13) where S denotes the surface area penetrated by the flux. The total flux will be the sum of its components inside the crucible, the glass layer and in the susceptor. Given that H = H o across the air gap separating the coil from the crucible and inside the crucible itself, The induction melting of glass expression (13) can be evaluated throughout this region and gives (PA = n(rz ~ - c2)pnHn where z ( r : - 6’) is the area of the air gap and the crucible perpendicular to the H field. The flux QG in the glass can be found by substituting expression (8) into (13) and integrating between the limits r = b and r = c . Thus QG= 2 n p n lbc [AF,(kr)+BF,(kr)]rdr. This expression is then evaluated to show that Im operators take the real and imaginary parts of whatever follows. Thus, on substituting expression (14) into equations (15) and (16) the following set of results is obtained. (i) The effective resistance of the glass layer where G = ( A / H o ) [ c F (; k c ) - bF; ( k b ) ] + ( B / H o ) [ c F i ( k c )- b F i ( k b ) ] . (ii) The effective resistance of the susceptor A similar procedure shows that the flux in the susceptor is n p n w N 2 ( d- c’) I (iv) The inductive reactance of the glass layer XA where the integration in this case was between the limits r = b and r = a , and it has been assumed that H ( a ) 0. [(L) 2npowN2b Re (1 + i)] kJd2 Hn (iii) The combined inductive reactance of the air gap and the crucible R, = +B[cF;(kc)-bF;(kb)]). (17) = - 4. The electrical characteristics of the furnace The magnetic field H o induced inside the walls of a glass furnace by a coil of length I, having N turns and carrying a current I, is given by the expression Ho = N I $ . This equation is not exact, but does follow as an automatic consequence of the long cylinder approximation discussed earlier. The coil current I, is alternating and therefore the magnetic flux through the furnace QF = @ A + Q G + Q s (14) (v) The inductive reactance of the susceptor The back EMF generated by the furnace does not equal the total voltage drop across the coil, because there is t h e coil’s own resistance and inductive reactance to be taken into account. Coils in this application usually consist of long lengths of copper tubing wound into the form of cylinders. It is found in practice that the resistance and inductive reactance of a coil for a given frequency of alternating current can be calculated approximately from the formulae is also time dependent. Subsequently a back EMF - V, is generated across the coil and this is related to the rate of change of flux linkage by Faraday’s law aQF V, = N - . at On multiplying the right-hand side of this expression by the normalizing factor (NIc/lHo) and making the standard replacement aiai-. iw, the transformed equation can then be expressed as (see Paschkis and Persson 1960, p 487). The formulae have the same mathematical form as the equivalent expressions for the resistance and inductive reactance of the susceptor (see equations (17) and (U)), although one difference is that a correction factor n(=1.1) has been included in equation (19) to take account of the spaces between the windings. VB = (RE + ~ X E ) I , where 5. The power dissipation is the effective resistance of the glass layer and the susceptor referred to the coil, and The instantaneous power dissipation in the susceptor can be evaluated by integrating the Poynting vector over the susceptor’s outer surface S.This integral may be expressed as P= is the inductive reactance of the furnace. The Re and I, [(Re E) X . (Re H ) ] d S (20) where it is necessary to take real parts because two 661 R Ducharme et a1 Table 1. The real and imaginary parts of C/Hofor several differentfrequencies.The assumed values for the parameters b, c, U,, uo are those quoted for figure2. ~ Frequency (MHz) 0.1 0.5 0.25 1 .o 0.75 ~~ Re(C/Ho) Im(C/Ho) 0.999 0.999 -1.92 x 10-2 -4.66 x 0.998 -4.26 X 0.995 -7.57 x 10-2 10.' 0.988 -0.117 the power dissipation in the glass substantially exceeds that in the susceptor. It is notable that this effect would be even more pronounced if the glass layer were thicker or characterized by a higher conductivity. The power dissipation in the glass is important because it influences both the electrical characteristics of the furnace and the convection currents in the melt. 6. The force on the susceptor Cal frequency IMHzl A current element J d r flowing in a magnetic field H will experience a force dF given by the Lorentz force law as Figure 3. The power ratio R,/R. as a function of frequency. . L'LI!CIK!S~ ,I ce-pirx .... quanr!i!rs I o r c e ~ rmixed (!errain 1970, p 665). Given that E and H are constant over the surface r = b, expression (20) is easily integrated to give P = Re H(b) Re E(b)2nbl. (21) The values of R e H(b) and R e E(b) may always be calculated directly from expressions (8) and (10). If, however, the penetration depth of the glass is at least four or five times larger than (c - b), it is reasonable to make the approximation C/H,- 1 (see table 1). Thus, recalling that Ho varies as exp iwr, it can be seen that Re H ( b j = H , cos mr. A similar calculation based on expression (10) shows that Re E(b) = ( H m k , / o , ~ 2 ) ( c o wr-sin s or) .. where ni :s d F = [(ReJ) x (ReH)]p,, d r _. differenri-i vciuz-e &:e=:. :.?e h c e on the susceptor for the problem depicted in figure 1 is directed radially inwards towards the cylinder axis. The magnitude of this force can be found by integrating over the volume of the susceptor. Consequently B b F=npol/ d (-(ReH)')rdr dr (23) where equation (9) has been used to eliminate J . The penetration depth in molybdenum for all coil frequencies greater than 5 kHz can be shown to be less than 4 mm. This implies 6 Q b and, on the strength of this result, it would seem a good approximation to set r = b in expression (23). Evaluation of this integral, subject to the approximation C/H, = 1, then gives F = n b l p u H i cos2 mt (24) where in accordance with expression (22) Integration of expression (21) over one complete time cycle shows that the time-averaged power dissipation (P)has the value The ratio of the power dissipation in the glass layer IZRG to the power dissipation in the susceptor IZR, may be expressed as RG/R,. A graph of this power ratio as a function of coil frequency is shown in figure 3. This curve is based on the same characteristic furnace parameters as were used to obtain figure 2 . The striking features of the graph is that at high frequencies, 662 It is known from experience in working with rod shaped electrodes of the type used in conventional electric glass furnaces, that molybdenum is an extremely brittle metal and great care must be taken in its handling. It has happened on occasion, for example, that a molybdenum rod 1m long and 4 cm in diameter has fractured inside an electrode holder as a result of stresses produced only by. its own weight (120N). Since forces must be scaled in proportion to cross sectional area for shear stresses to be held The induction melting Of glass ./loo ‘YO0 kW I Fwquerry IkHzl Figure 4. Dependence of the magnetic force (F,) experienced by the susceptor on frequency. constant, it is therefore to he expected that a molybdenum susceptor, with a cross sectional area fifty times greater than that of a rod shaped electrode, is at some risk from forces of the order of, or exceeding, 6000 N.It should he noted that damaging stresses are most likely to arise in the vicinity of the bolts and brackets which fix the susceptor in place inside the crucible. The peak force on the susceptor is given by and this result follows from expression (24). A plot of F, against coil frequency f( = w / 2 n ) , for three different values of (P)is shown in figure 4. It can he seen from the graph that the magnitude of F,,, can exceed 6000 N at low coil frequencies. Under these conditions there is therefore the danger that the time varying forces on the susceptor will cause it to fatigue and ultimately to crack. 7. Summary ~ An analytical solution to the problem of eddy current heating of a composite cylindrical system has been provided in the case when the outer cylinder is glass and the inner cylinder is molybdenum. The metal susceptor is in fact hollow, but this is not important to the solution since the glass inside is well shielded from the coil’s magnetic field by the metal. The coil induces eddy currents in the unshielded glass and in the susceptor. These currents in turn produce fields outside themselves and therefore react hack on both the outer cylinder and on the coil. This back induction has been calculated and used to determine the electrical characteristics of the complete system. The flow of conduction currents in the glass layer becomes most significant when the coil frequency is high. Under these circumstances it has been shown that the greater proportion of the power dissipation in the furnace takes place in the glass and not in the susceptor. The problem here is that this only becomes true when the glass is heated and ceases to he an insulator. The melting of the glass is therefore accompanied by a sudden change in the hack potential generated by the furnace across the coil. Since power supplies are necessarily limited in the range of operating conditions to which they can adapt, it is clear that the wrong choice of equipment may result in a furnace becoming inefficient or even inoperable at high temperatures. A susceptor in an induction coil will experience a force owing to the action of the coil’s magnetic field on the current flowing inside it. An expression for this force has been obtained and its magnitude calculated under a variety of operating conditions. This is an important consideration because molybdenum, like most refractory metals, becomes very brittle at the high temperatures encountered in glass furnaces. There is, therefore, the danger that the susceptor will sustain damage if the forces on it are too large. It has been demonstrated that this is most likely to he a problem at very low coil frequencies (of the order of magnitude of 1OOOHz or less), when large, time varying forces u p to several thousand Newtons could cause fatigue damage in the vicinity of the bolts and brackets that fix the susceptor in place inside the crucible. Acknowledgment One of us (R Ducharme) is indebted to the Science and Engineering Research Council for the award of a Case Studentship. References Lorrain P 1970 Electromagnetic Fields and Waues (San Francisco: Freeman) McLachlan N W 1961 Bessel Functions for Engineers (Oxford: Oxford University Press) McLachlan N W and Meyers A A 1935 Eddy current loss in a tube in axial alternating magnetic field Phil. Mag. 29 846 Paschkis V and Persson J 1960 Industrial Elecrric Furnaces and Appliances (New York: Wiley Interscience) Scarfe F 1980 Electrical melting, a review Glass Technol. 21 (1) 37-50 Watson G N 1966 A Treatire on [he Theory of Bessel Functions (Cambridge: Cambridge University Press) Wright D A 1937 Eddy current heating of composite cylindrical systems Phil. Mag. 24 1 663
© Copyright 2024 ExpyDoc