High-performance Forehearth Coloring using Lorentz Forces Torres J.O. 1, Halbedel B. 1, Weber C. 2, Reche. R. 3 1 Technische Universität Ilmenau, Ilemanu, Germany 2 Ferro GmbH, Frankfurt am Main, Germany 3 O-I GLASSPACK GmbH & Co. KG, Rinteln, Germany Abstract It was observed that the frits used to coloring the molten glass tend to sink, degrading the homogenization done by the stirring battery that needs a certain frit level, and contaminating the glass when they sediment. A Lorentz force density distribution is created against the glass flow in melt zone of the forehearth to prevent the sinking process by a push-up effect and therefore to increase the frit levels. Key words: Forehearth, glass coloring, Lorentz force, heat transfer, frits, numerical modeling 1. INTRODUCTION In the production of coloring glasses, molten glass which flows out of the glass tank is colored in the melt zone of the forehearth (cf. Fig. 1) by means of color concentrates (frits). These frits are added to the molten glass through feed tubes, melting and diffusing progressively. This diffusion is not enough to homogenize the color, therefore a battery of stirrers (cf. Fig. 1b, No. 7) are located at the end of the melt zone. It was observed that the frits tend to sink, degrading the glass coloring because the stirring battery needs a certain frit levels to do a good homogenization, and contaminating the lower part of the molten glass when they sediment. A suitable Lorentz force distribution generated in the molten glass and against its flow direction is used to prevent the sinking process with a push up effect and thus to improve the input flow conditions of the stirring battery. This force distribution is created with an arrangement of three molybdenum electrodes, located in the molten glass, and two water cooled copper coils, located in the insulating material (cf. Fig. 1a). Both systems are energized with a single-phase alternating current (50 Hz). The electric current density that flows in the molten glass from the molybdenum electrodes [2] interacts with the magnetic flux density created by the electric current in the coils and generates the Lorentz force density distribution. It is included in the Navier- Stokes equation (Eq. 1) to compute its influence on the velocity distribution of the molten glass. Also, the same electric current density produces a Joule heating density in the melt which is considered into the energy equation (Eq. 2) to compute the temperature distributions within the molten glass as well as in the forehearth. The heat transfer by radiation between the crown (walls surrounding the atmosphere), the flames and the molten glass surface as well as the radiation inside the molten glass was done with the discrete ordinates radiation model [6].
( ) ( ) ( ) ( ) (1) ( ) ( ( )) ( ) (2) The numerical simulations were performed using the commercial computational fluid dynamics software ANSYS/ FLUENT with a non-conformal mesh to isolate the different requirements of mesh resolution - more in the molten glass and in the atmosphere and less in the refractory and in the insulating materials. Finally, the selected frits trajectories in the molten glass flow were modeled with the trajectories of passive particles computed with the discrete phase model. The construction of the 3D numerical model where the simulations were performed was done with the geometrical and physical properties of the forehearth as well as the physical properties of the molten glass used at the installation of O-I GLASSPACK GmbH & Co. KG in Rinteln/ Germany. 2. DESCRIPTION OF THE MELT ZONE OF THE FOREHEARTH 2.1 Geometry The Fig. 1 shows the schematics of the melt zone which includes also the electrode and coil arrangements to generate the Lorentz force. The region in which the Lorentz forces shall operate is located between the feed tubes and the stirring battery (Fig. 1b). Figure 1: Schematics of the melt zone of the forehearth, a) cross section (x,775,z) mm, b) length section (0,y,z) mm. 1 Refractory material, 2 Electrodes, 3 Coils, 4 Insulating material, 5 feed tubes, 6 flames, 7 stirring battery. D C Diameter of the coils, L E Position of the electrodes in the y-axis, L C - Position of the coils in the y-axis.
2.2 Physical Properties of the Glass Melt The thermal conductivity [3], the specific heat, the emissivity [4], and the refraction index [5] were taken from the literature and later tuned with the temperature measurements on the forehearth. The density, viscosity and electrical conductivity of the molten glass are considered dependent on the temperature. The dependency of the density is equal to the DGG standard soda lime glass Ia, the dependency of the viscosity was selfmeasured with a rotational viscometer VIS 403/ NETZSCH at the TU Ilmenau and the dependency of the electrical conductivity was measured by JSJ Jodeit GmbH in Jena-Maua. The results and used dependencies are shown in table 1. Table 1. Used physical properties of the molten glass in the forehearth Density [Kg/m 3 ] (T) = 2508.43-0.15T Dynamic viscosity [Pa s] log( (T)) = -4.10 + 7213.38T -1 Electrical conductivity [S/m] log( (T)) = 3.55 2863.26T -1 2.3 Operating parameters The amount of glass pull rate 80 t/d introduces a total thermal power in the forehearth of about 1.35 MW. A part of it is lost through the glass melt surface and through the forehearth surface which is compensated with a burner system that produces a power about 38 kw [1]. 3. 3D NUMERICAL MODEL OF THE MELT ZONE OF THE FOREHEARTH 3.1 Validation of the 3D model with temperature measurements The temperatures were measured on selected points of the forehearth surface and in the melt with a platinum rhodium thermocouple sensor when green glass was running in the melt zone of the forehearth and the feed tubes as well as the stirring battery was in non-operating state. The signals were measured by the meter VOLTCRAFT R 306 data logger thermometer with an estimated total measurement error of ± 5 C. The comparison of the simulated temperature distributions of the forehearth surface as well as inside the molten glass with the measured temperatures are shown in the Fig.2 and Fig.3. The deviations are less than ± 5 C which are in the range of the measurement error. Only the simulated temperature distribution in the molten glass on the side region of the outlet (under the first line of the stirring battery Fig. 1b) was underestimated (approx. - 10 K). The tuning of the measured surface temperatures of the forehearth with the calculated temperatures was done with a linear variation of the heat transfer coefficient with a mean value of 20 W/m 2 K. The numerical model predicted total heat losses of 37.5 KW which are compensated by the burner flames included into the model as a normal temperature distribution.
Figure 2: a) Measured (dots with the error range of ± 5 C) and simulated (line) temperature distribution inside the molten glass along the z-axis of the forehearth at the inlet middle (0,0,z) mm. b) Computation of the temperature in each face of the forehearth surface (red color) and temperature measurements in each face (black color). Figure 3: Measured (dots with the error range of ± 5 C) and simulated (line) temperature distribution inside the molten glass along the z-axis of the forehearth a) at the inlet side (420,0,z) mm under the feed tube position and b) at the outlet side (420,1549,z) mm under the first line of the stirring battery. 3.2 Push up effect The executed numerical simulations show that to cover the large width of the forehearth an arrangement of three electrodes with diameters of D E = 60 mm for the side ones and D E = 80 mm for the middle ones (cf. Fig. 1a) are necessary to reduce the Joule heating density around the electrodes and the resultant additional negative velocities. For such electrode system, two coils with external diameter of D C = 200 mm must be located in the insulating material between the electrodes. The electric current on the electrodes flows from the left and the right electrode to the middle electrode with a total electric current of I E = 217 A (cf. Fig. 4a) using a voltage of U E = 20V. The electric current on each coil I C = 1500 A circulates in opposite direction to obtain a magnetic flux density B in only one sense (cf. Fig. 4b) and thus to have the Lorentz force density f L, aligned opposite to the molten glass flow over the total cross
section (cf. Fig. 4c and 4d). The trajectories of the passive particles with starting position at (0,0,82) mm, (40,0,82) mm and (300,0,82) mm (equal to the positions of the right feed tubes due to the symmetry of the forehearth) have a final level bigger than the configuration without electrodes in almost 20 mm, 2 mm and 1 mm respectively (Fig. 5a) when the electrodes and coils are located at the middle of the melt zone, L E = L C = 775 mm (cf. Fig. 1b). Due to the positive slop of the trajectories is possible to increase the final level of them locating the electrodes closer to the feed tube (L E = L C =388 mm). It was found that the final level was enhanced up to 25mm, 3mm and 2mm for starting position at (0,0,82) mm, (40,0,82) mm and (300,0,82) mm respectively (cf. Fig. 5b). Figure 4: a) Distribution of the electric current density magnitude J at the plane (x,775,z) mm, b) vector field of the magnetic flux density B colored by its magnitude at the plane (x,775,z) mm, c) distribution of the Lorentz force density f L magnitude in a cross section plane (x,775,z) mm, d) vector field of the Lorentz force density f L colored by its magnitude. Figure 5: Particle trajectories with I E = 217 A and I C = 1500 A as well as without electrodes with initial positions located at (0,0,82) mm, (40,0,82) mm and (300,0,82) mm, a) electrodes and coils located at y=775 mm (L E = L C =775 mm), b) electrodes and coils located at y = 388 mm (L E = L C =388 mm).
4. CONCLUSIONS The investigations show that on the basis of a verified numerical 3D model which included the boundary conditions, process parameters and temperature depending material properties, it is possible to obtain a sufficiently accurate temperature distribution within the errors of ± 5 K on the forehearth surface as well as inside the molten glass. Only the calculated temperatures in the side of the outlet are in the range of ± 10 K due to the simplification that the molten glass is gray and opaque with its optical properties independent of the temperature. In order to cover the large width of the O-I forehearth, it was determined an arrangement of three electrodes and two coils to obtain enough push up effect over the total cross section. To avoid the regions that have additional negative velocities due to the joule heating, the positions of the feed tubes in the x-axis, used as starting position for the passive particles, corresponds to the regions where the push up effect is more intense. Furthermore, it is expected that the push up effect in white glass will be stronger than in green glass at the same electric currents I E and I C because: The Joule heating will have a lower impact in white glass than in green glass due to the better internal heat transfer. Consequently, the bottom temperature of the white glass will be several degrees larger than within green glass. Therefore, the white glass will have lower viscosity and higher electric conductivity resulting in a stronger Lorentz force density with a larger push effect. 5. ACKNOWLEDGEMENT This research project was developed between the Technische Universität Ilmenau, FERRO GmbH Frankfurt/ France S.a.r.l Saint Dezier and O-I GLASSPACK GmbH & Co. KG Rinteln. The authors are grateful to the FERRO France for financial support and to the O-I GLASSPACK GmbH & Co. KG for the provision of the geometrical and physical data of the forehearth as well as the support at the necessary temperature measurements. 6. REFERENCES [1] E. C. JR. BAUKAL, Heat transfer in industrial combustion, Florida, CRC Press LLC 2000. [2] PLANSEE SE, http://www.plansee.com, 2014. [3] S. GOPALAKRISHNAN, A. THESS, G. WEIDMANN, U. LANGE, Chaotic mixing in a Joule-heated glass melt. Physics of fluids, 22 (1), 013101, 2010. [4] LAWRENCE BERKELEY NATIONAL LABORATORY, http://www.lbl.gov, 2014. [5] J. PISCHKE, Wärmestrommessungen an opaken Quarzglas-Flanchen. Experimentelle Studienarbeit, Technische Universität Claushal, 1994. [6] G.D.RAITHBY AND E.H. CHUI. A Finite-Volume Method for Predicting a Radiant Heat Transfer in Enclosures with Participating Media. J. Heat Transfer, 112, 415 423, 1990.