During the past decade, several novel melting and casting techniques including vacuum arc melting, skull casting, Hopkins’ Process or electroslag melting, and electron-beam melting have been developed on an industrial scale. All of these techniques employ a water-cooled crucible. Heat transfer during vacuum-arc melting with water-cooled crucibles has been studied on a theoretical bases, but limited experimental data are available. It is the purpose of this report to detail the results of several experiments conducted with vacuum-arc melting equipment in which heat transfer from the molten charge to the cooling medium was studied.

Kroll has attributed the discovery of the cold-crucible principle to Von Bolton who melted tantalum on a water-cooled copper hearth with an arc. Further development of this device is highlighted in the work of Moore, Kroll and Parke, and in the intensive expansion of commercial arc melting of titanium and zirconium.

The application of consumable electrode vacuum arc melting has increased rapidly during recent years, particularly in the melting of special alloy steels, reactive, and refractory metals. The process consists of melting a consumable electrode of the material by means of a high current electric arc maintained between the lower end of the electrode and a pool of molten metal contained in a water-cooled copper crucible. Melting is normally conducted in vacuum or in an atmosphere of an inert gas such as helium or argon. The ingot which forms as the electrode is consumed, solidifies in the water-cooled copper crucible. The crucible is usually of circular cross section.

Cooling of the crucible is characterized by high heat fluxes, and the use of water represents a potential safety hazard in that the crucible is occasionally ruptured during melting with the resultant mixing of molten metal with water. In rare instances, particularly during the melting of reactive metals, this mixing of water and molten metal has resulted in serious explosions in which a number of fatalities have occurred.

As a safeguard against this hazard, practically all large furnaces for melting reactive metals are now installed in vented explosion vaults and are extensively equipped with safety interlocks. Alternate coolants have also been investigated, but water has remained the primary coolant.

It is generally agreed that local overheating of the crucible causes crucible rupture, but the reasons for the local overheating are not fully understood. Suggested reasons include defects in the crucible wall, defects in the cooling system, localized wetting or alloying of the ingot and crucible wall, and irregularities in the arc discharge.

Little information is available on rates of heat transfer, crucible wall temperatures, and other criteria useful in designing cooling systems for the crucible. Most cooling-system design is based on theoretical assumptions regarding the mode of heat transfer to the crucible plus experience gained through years of melting. Theoretical aspects of cooling of the crucible during arc melting are contained in the work of Rossin, and some of the work in the field of continuous casting is closely related. A more recent publication was based on the solution of a mathematical model of heat flow by means of a computer during the consumable electrode arc melting of molybdenum.

In this report, experimental values of heat flux from the water-cooled copper crucible are presented, and the effects of some of the variables associated with the process are shown. The data presented were obtained during the consumable electrode arc melting of zirconium, titanium, and steel ingots up to 8 inches in diameter using straight polarity (electrode negative) direct current.

Initial studies were conducted during nonconsumable-electrode melting of 2-5/16-inch-diameter ingots. These heats were conducted in a six-piece crucible in which average values of heat flux over short increments of crucible length were determined. Small-scale, consumable electrode heats were conducted in a two-piece, 2-5/16-inch-diameter crucible, and the technique developed was then applied to consumable electrode melting in a two-piece, 8-inch- diameter crucible. Data were also obtained on crucible wall temperatures and on the effectiveness of water jackets with very narrow annular spacing.

The information obtained in this study is considered representative of conditions existing during consumable electrode arc melting by industrial units, particularly for the reactive metals. The values of heat flux given are not intended to represent conditions which exist during the starting of a melting operation or when the ingot is very short. Determinations of heat flux during this portion of the melting cycle would be a study in itself. The values of heat flux given are intended to represent conditions after the ingot becomes at least one diameter in length. After this length has been achieved, melting has become relatively stable and the effect of heat transfer to the crucible bottom has diminished.

Theoretical Pattern of Heat Transfer from Ingot to Cooling Water

Figure 1 represents the cross section of an ingot during consumable electrode arc melting in a water-cooled copper crucible. During melting, the electrode (1) is consumed by heat of the direct current arc (2), which is maintained between the electrode and the molten pool (5). Normally, this molten pool extends nearly to the wall of the copper crucible (3), which is cooled by the water flow in the annular water jacket (4). The size and shape of the molten pool will depend on the conditions prevailing during melting, particularly on arc current and arc potential. The molten metal solidifies where it contacts the cold wall of the crucible and shrinks sufficiently to leave a gap (6) between the ingot (7) and the crucible wall.

The rate of heat transfer along the length of the crucible varies over a wide range of values with high values of heat flux near the top of the ingot where molten metal from the pool splashes against the crucible wall. High heat transfer rates also occur in an area just above the top of the ingot where molten spatter from the electrode strikes the wall to form a crown of metal. Bureau research on the effect of extremely long arcs indicated the existence of two areas of high wall temperatures above the ingot top. These two areas of high temperature coincided with the excessively high crowns formed during melting under these arc conditions.

The rate of heat transfer is greatly reduced along the lower portion of the ingot where solidification and cooling of the ingot result in shrinkage of the ingot away from the crucible walls. Heat transfer rates are also relatively low a few inches above the ingot top where radiation from the hot electrode, the arc, and the molten pool contribute heat to the crucible wall. Heat transfer rates through the bottom of the ingot are greatly reduced for the same reason heat transfer rates on the lower ingot sides are low. Solidification and cooling of the ingot result in shrinkage and poor contact with the bottom. One of the best evidences of this fact is the thin layer of solidified metal which remains in a casting ladle following a consumable electrode casting heat. Such a skull of metal is shown in figure 2 and is typical of skulls obtained. On occasions, intimate contact occurs at a spot where the skull becomes welded to the ladle wall. When this happens, the skull thickens markedly at the point of good thermal content. In general, however, the thermal resistance between the ingot and crucible is relatively high along the lower surfaces of the wall and along the bottom.

In the experimental work conducted, it was assumed that uniform arc conditions existed throughout each run conducted. In addition, it was assumed that after the ingot had developed to a length equal to one diameter, the heat flux profile did not change with increasing ingot length. Further discussion of the experimental work and presentation of experimental results to justify these assumptions are included in the appendix. A mathematical study of the distribution of temperature and heat flow in arc-melted ingots is included in reference.

Experimental Procedure and Results

The studies conducted were divided into three major areas: (1) Nonconsumable electrode arc melting in a six-piece crucible, (2) consumable electrode arc melting in a two-piece crucible, and (3) the effect of- narrow annular spacing for water jackets. The work with the two-piece crucibles received the greatest attention since consumable electrode arc melting is of more commercial importance than is nonconsumable electrode arc melting. Melting studies were conducted in two-piece crucibles 2-5/16 inches in diameter and 8 inches in diameter. The experiments on nonconsumable electrode melting and the study of water jackets with narrow annular spacing were conducted in 2-5/16-inch~ diameter crucibles.

Nonconsumable Electrode Arc Melting Studies

First attempts to measure the distribution of heat-transfer rates along the length of the crucible were made with a nonconsumable tungsten electrode and a crucible divided into six increments of length. A diagram of the crucible is shown in figure 3. The crucible was divided into one 3-inch length at the bottom; four ½-inch lengths, and one 1-inch length at the top. These sections were numbered 1 through 6 from bottom to top. The number 3 section was equipped with two iron-constantan thermocouples, installed as shown in the insert of figure 3. Detail of the mica insulation between sections is also shown in the insert of figure 3. Each section was cooled by an individual water-cooling circuit. Water inlet and exhaust was by 3/8-inch-OD copper tubing which connected to a circumferential water jacket machined in each section.

The rate of heat transfer to each section was determined by measuring coolant flow rates and the temperature change of water for each section during melting of a zirconium ingot. For each run, the ingot was positioned with the top of the ingot at a selected level in the crucible, and an arc was maintained

between a ½-inch-diameter tungsten electrode and the ingot. Recordings were made of arc current and arc potential for runs conducted at arc currents ranging from 750 amperes to 1,200 amperes. As soon as the run had reached equilibrium conditions, coolant flow rates and temperature changes were obtained for each of the crucible sections. Coolant flow rates were determined by weighing samples of coolant taken over a 1 to 2 minute interval during the run. Inlet and outlet temperatures of the coolant were measured by means of mercury thermometers. From these data, the heat transfer rates from the ingot to each of the crucible sections were determined.

Figure 4 summarizes data for a series of runs at various power levels from 18 kw to 36 kw. For these runs, the arc gap was 0.4 inch, the furnace pressure 400 torr of helium, and the ingot top was at the midpoint of section 3 of the crucible. The values plotted in figure 4 represent the average value of heat flux for each individual crucible section. Maximum heat flux occurred in section 3 for all power levels and was greatly reduced both above and below this section.

High values of heat flux occurred in section 3 because of direct radiation from the arc to the crucible and because of the proximity of the molten pool to the crucible wall. At the upper surface of the ingot, only a thin skin of solid metal separated the crucible wall from the molten pool, and portions of this skin were continually melted and reformed as the arc moved about the surface of the pool. Further down from the top of the ingot, heat transfer rates decreased as a result of shrinkage of the ingot from the walls of the crucible.

With increased power input, the band of high heat transfer at the top of the ingot was increased in width. Heat flux to section 2 of the crucible increased markedly as the power was increased from 27 kw to 36 kw. Very little increase in heat flux to section 2 was noted with the power increase from 18 kw to 27 kw because the band of high heat transfer had not been widened sufficiently to affect section 2.

In addition to the information on the distribution of heat flux along the length of the ingot, data were obtained on crucible wall temperatures during nonconsumable-electrode melting. For these tests, two iron-constantan thermocouples were imbedded in the wall of section 3 of the segmented crucible. These thermocouples were located at the midpoint of the ½-inch length of this section of the crucible as shown in the insert of figure 3, and provided a measurement of the temperature gradient through a portion of the crucible wall. The outputs from these thermocouples were recorded during nonconsumable-electrode arc melting of a zirconium ingot, the top of which was at the same level as the thermocouples.

The radial position of each of the two thermocouples was measured optically on a movable-stage metallographic microscope after sectioning the crucible at the thermocouple junctions. The temperature of the inside surface

of the crucible wall was then calculated from the relationship

where ti was the temperature of the inside surface of the crucible wall, t1 was the temperature indicated by the thermocouple nearest the inside surface of the crucible wall, and Δt was the temperature difference between the two thermocouples. The radial distance to the inside surface of the crucible, ri, was 1.156 inches; to the inside thermocouple, r1, was 1.235 inches; and to the outer thermocouple, r2, was 1.352 inches. Substituting these values in equation 1 yields the relationship:

ti = t1 + 0.725 Δt

Table 1 lists data obtained at various power levels, and figure 5 is a plot of the calculated temperatures of the inside surface of the 2-5/16-inch-diameter crucible as a function of power input. Both the maximum and minimum observed values are included in table 1 and in figure 5 to provide an indication of the variation in the wall temperature noted for any given power level.. This variation was caused by movement of the arc and the subsequent nonsymmetry of the molten pool. At lower power levels, the differences between maximum and minimum temperatures were relatively small, but the differences became larger as the power input increased. The maximum values occurred when the arc was directed toward the side of the crucible containing the thermocouples. Table 1 also includes values of heat flux calculated from the temperature gradient within the crucible wall.

The trends noted from these data are considered of greater significance than the actual numerical values reported. Precise determination of the short distance between the two thermocouples was impossible because of the uncertainty of the exact location of the junction within the thermocouple bead. The effect of removal of a portion of the wall for the thermocouple installation was an additional unknown which contributed to the uncertainty of the calculations. For these reasons, these data are presented to indicate trends in wall temperature and radial heat flux rather than precise values. Values of heat flux calculated from cooling-water data are considered more precise.

The experimental work conducted with nonconsumable electrode melting in the six-piece crucible provided background information for the more important studies of heat transfer from the ingot to the crucible during consumable electrode arc melting. Consumable electrode studies were not conducted in the six-piece crucible because of the difficulty of instrumenting six separate cooling circuits to yield meaningful data and because of the limited cooling capacity of certain of the individual crucible sections.

Consumable Electrode Arc Melting Studies

The study of heat transfer during consumable electrode arc melting was difficult because the length of the ingot, and consequently the position of the top of the ingot, was continually changing. The dynamic nature of the process, both with respect to this change in length of the ingot and with respect to the characteristics of an arc, precludes any attempt at steady state conditions. One possible solution to this problem would be to use a bottom withdrawal mechanism with which the ingot top could be maintained at a constant level throughout the run. However, this approach was eliminated in favor of the one used because of expected difficulties in the mechanics involved in maintaining a constant ingot position.

Experiments with the 2-5/16-Inch-Diameter Crucible

Figure 6 shows a cross section of the crucible which was designed to provide the data required. The 2-5/16-inch-diameter crucible was divided into two sections, each 4 inches long. Each of the two sections had its own integral water jacket. Water entered through a ¾-inch copper pipe at the bottom of the crucible, flowed upward along the walls of the lower section, over the lip of the cylindrical baffle, and out the exhaust of the lower

section. The water was carried to the upper section through pipe. In the upper section an elliptical baffle directed the water downward, under the lower lip of the cylindrical baffle, and the water then flowed upward along the wall of the upper section and exhausted through the outlet pipe.

Thermistor probes were installed in the inlet and outlet pipes of the upper section. Signals from these thermistor probes were recorded as a measure of the outlet water temperature of the bottom section and the inlet and outlet water temperature of the upper section. The inlet temperature for the bottom section remained constant throughout the run and was measured by means of a mercury thermometer installed in the water supply line. An O-ring seal between the two sections made the crucible vacuum tight and maintained a slight spacing between the two sections to reduce heat transfer from one section to the other.

Four iron-constantan thermocouples were connected in parallel and imbedded in the crucible wall three-eighths of an inch above the joint between the upper and lower sections, and four more were similarly installed three-fourths of an inch above the joint. Each of the four thermocouples of the lower set were 90° apart, and the four thermocouples of the upper set were evenly spaced between them. These thermocouples yielded data on crucible wall temperatures at two levels three-eighths of an inch apart. They were radially located within the crucible wall to yield mean wall temperatures which were used to estimate the heat involved in any temperature change of the crucible itself.

Heat transfer data were obtained during consumable electrode arc melting of zirconium and titanium at various power levels and at furnace pressures ranging- from 50 to 400 torr. The diameters of the consumable electrodes used were 1-inch for titanium and 1-3/8-inch for zirconium. Power levels were varied from run to run as were relative values for arc current and arc potential.

The melting procedure for these runs was as follows : (1) The furnace was loaded with the consumable electrode of titanium or zirconium attached to the negative furnace electrode. A starting pad of sponge metal or, in some cases, a short length of ingot was placed in the bottom of the crucible which was the positive furnace electrode. (2) The furnace was closed and evacuated to the ultimate pressure of the system (10 to 15 millitorr). The desired furnace pressure was obtained by backfilling with helium. (3) The electrode was lowered until an arc was initiated between the lower end of the electrode and the base. Until a full pool of molten metal was formed in the crucible, arc current and potential were maintained at values which would cause melting of the base but at which no melting of the electrode took place. (4) The power level was then rapidly increased to the value at which the melt was to be conducted and was maintained at that level until a predetermined length of electrode had been melted. (5) Power was terminated and the ingot was allowed to cool at the pressure maintained during melting.

During the run, recordings were obtained of arc current, arc potential, temperature of the cooling water at the inlet and outlet of the upper crucible section, crucible wall temperatures, and electrode consumption rate. Direct

readings of the inlet water temperature were taken, thus providing inlet and outlet water temperatures for both upper and lower sections of the crucible. Coolant flow rate was determined by taking periodic manometer readings of the pressure drop across a sharp-edged orifice.

Data from the two sets of thermocouples imbedded in the crucible wall are presented in figure 7. In this figure, temperature from both sets of thermocouples is plotted as a function of time. The dotted line at 200 seconds represents the time when the ingot length was equal to the length of the lower crucible section or the time when the level of the ingot top moved past the division between the upper and lower crucible sections. Since the two sets of thermocouples are three-eighths and three-fourths of an inch above the division between the two crucible sections, they would be expected to reach their maximum values about the time the ingot was 4-3/8 inches and 4-¾ inches long. These data indicate that the wall temperature drops off rapidly on either side of the ingot top. Unfortunately, the two sets of thermocouples were not far enough apart to provide information on the change in temperature profiles for ingots much different in length.

Data from these thermocouples indicated that the amount of heat involved during changes in temperature of the mass of copper making up the crucible wall was only 0.9 percent of the total input. This figure was obtained by calculating the changes in heat content of 1-inch increments of crucible length based on temperature changes obtained from the thermocouples. The maximum rate of change of the total heat content of the crucible was 970 Btu/hr when the ingot length changed from 4 inches to 5 inches.

These figures were based on the supposition that the temperature profile was of the shape shown in figure 7 regardless of the length of the ingot. This supposition would not hold if the ingot were very short, and somewhat larger error would be expected during the first part of any melting operation.

Data on coolant flow rate and the temperature change of the coolant were used to calculate the heat absorbed by the cooling water to the upper and lower sections of the crucible as a function of time. Figure 8 is a typical

plot of the rate of heat transfer to the water in the upper crucible section during the consumable-electrode arc melting of a 1-3/8-inch-diameter zirconium electrode. This particular run was made at a furnace pressure of 300 torr of helium, an arc current of 950 amperes, and an arc potential of 34 volts. The ingot formed was 6-¾ inches long and weighed 6.8 pounds.

Measurements of the rate of electrode movement during these runs indicated that the electrode consumption, and consequently ingot buildup, was constant throughout the run. Distance along the abscissa thus is proportional to the length of the ingot at any time during the run. In the run represented in figure 8, 1 inch of ingot formed every 50 seconds. The dotted line at 200 seconds represents the time at which the ingot was 4 inches long, and the top was even with the division between the upper and lower crucible sections.

The rate of heat transfer to the upper crucible section was small during the early part of the run and increased gradually until the top of the ingot approached the division between the two crucible sections. The rate of heat transfer increased rapidly as the zone of high heat transfer near the top of the ingot moved into the upper crucible section. During the latter portions of the run, the rate of heat transfer increased only slightly with increasing ingot length.

The change in the rate of heat transfer observed during any interval of time during the run resulted from an increase in length of the ingot and movement of the heat flux profile relative to the upper crucible section. With each incremental increase in ingot length, the top crucible section was exposed to an additional increment of the heat flux profile. For example during the time the ingot length changed from 4 inches to 4½ inches, the observed increase in the rate of heat transfer to the upper section represented the heat transfer from the upper one half inch of ingot. The average heat flux in Btu/ft² hr for this one-half inch of ingot was determined by dividing the change in the rate of heat transfer by the sidewall area of the one half inch of ingot. A similar calculation of heat flux can be made for any increment of ingot length. Since the melting rate was uniform throughout the run the slope of the curve in figure 8 is proportional to the change in the rate of heat transfer to the upper crucible section divided by the change in ingot length. The slope of this curve is therefore proportional to the heat flux.

Figure 9 is the graph of the slope of the curve in figure. 8. The ordinate of figure 9 represents the distance in inches above and below the top of the ingot, and the abscissa is proportional to the slope of the curve in figure 8 and is expressed in Btu per hour per square foot of ingot surface. In figure 8, the slope of the curve to the right of the 200 second line represents the heat flux for increments of ingot length below the ingot top, and slopes to the left of the 200 second line represent the heat flux above the ingot top. Figure 9 thus represents the distribution of heat flux to the crucible during consumable electrode arc melting.

Data similar to that contained in figure 9 were obtained during the melting of zirconium and titanium under a variety of conditions; however; these data were greatly affected by the inherent instability of the arc during small-scale consumable electrode melting. As a result, the effects of many of the variables investigated were masked by variations in the arc characteristics caused by operational difficulties associated with small-scale melting. These difficulties were greatly increased at low furnace pressures, and, consequently, no attempt was made to obtain data during melting at furnace pressures below 50 torr. Because of these limitations imposed by the small-scale equipment; similar experiments were conducted using an 8-inch-diameter crucible.

Experiments with the 8-Inch-Diameter Crucible

The crucible for this study was of the same basic design as that shown in figure 6 and was installed in place of the ladle in an over-the-lip vacuum arc casting furnace. Figure 10 shows the crucible installed in the furnace, and figure 11 shows the crucible with the lower section separated from the upper section following a run.

The crucible was designed to produce an 8-inch-diameter ingot, 16 to 20 inches long. The inside length of the lower crucible section was 8 inches, and that of the upper section was 14 inches. The coolant flow pattern was the same as for the 2-5/16-inch crucible; that is, water entered at the bottom of the lower section, cooled the lower section from bottom to top and then was directed to the upper section where it cooled the wall from bottom to top.

Thermistor probes were positioned at the inlet and outlet of the upper section and thus provided the outlet temperature for the bottom section and inlet and outlet temperatures for the upper section. The inlet temperature for the bottom section was taken as the temperature of the water supply and remained constant throughout the run.

The coolant flow rate was determined by weighing four to six 1-minute samples taken during the run. The coolant flow rate was relatively constant throughout each run and was maintained at approximately the same level for all runs conducted.

Nine heats were conducted in this crucible of which six were zirconium and three were steel. The electrodes for three of the zirconium heats were 6 inches in diameter, and three were 4-½ inches in diameter. All three steel electrodes were 4-½ inches in diameter. All of the ingots produced were 14 to 20 inches long, and all the runs were conducted at furnace pressures in the range of 10-² to 10-¹ torr. Time for melting the individual heats varied from 13 to 22 minutes depending on power input and the material being melted. Table 2 summarizes the conditions which prevailed for the various heats.

The distribution of heat flux along the length of the ingot was determined by the same method outlined for the 2-5/16-inch ingots. A significant difference in the maximum heat flux was noted for these 8-inch runs as compared to that noted for the small-scale runs. For the 8-inch runs the maximum heat flux was the range 0.4 to 0.5 x 10 6 Btu/ft² hr compared to 1.4 x 10 6 Btu/ft² hr for the 2-5/16-inch runs described previously. This marked

difference is not surprising. Experience indicates that the frequency of crucible burn-through is much higher for small diameter melts (less than 4-inch diameter) than for larger melts.

Electrode diameter had a significant effect on the distribution of heat flux along the length of the crucible and on the maximum heat flux. Figure 12 shows the distribution of heat flux for two runs made under approximately the same operating conditions except for electrode diameter. For the 6-inch-diameter electrode, a higher maximum heat flux occurred, and the heat flux was concentrated over a shorter length of crucible. The maximum heat flux for the larger diameter electrodes also occurred at a higher level than for the smaller electrodes.

These results are substantiated by previous Bureau work (1) related to the effect of electrode diameter on the yield of metal (percent of charge poured) during consumable electrode arc melting and casting operations. This previous research indicated that, while the electrode diameter did not affect the yield of metal, the shape of the molten pool changed significantly with electrode diameter. Small diameter electrodes yielded skulls which were thin on the bottom and lower sides but which were thickened at the top. Large diameter electrodes yielded skulls which were thick at the bottom and lower sides but which were thin at the top The shapes of these skulls indicated that for the larger electrodes the heat in the molten pool was concentrated near the top of the pool, and for the small electrodes the heat was concentrated lower in the pool. Figure 13 represents the effect of electrode diameter on the shape of the molten pool during consumable-electrode arc melting in which other furnace parameters were equal. The patterns of heat flux shown in figure 12 indicate the same distribution of heat within the molten pool.

Figure 14 shows a comparison of a 4-½-inch-zirconium and a 4-½-inch steel electrode melted at approximately the same arc current and arc potential. In general, all of the steel melts yielded maximum heat flux at levels above similar runs with zirconium. The effect is believed to be the result of the manner in which the steel electrodes melt because of the gas content of the

metal. Steels which have not been vacuum melted previously usually melt with very unstable arc characteristics because of the gas content of the electrodes. The electrodes of the steel runs were made from 4-inch by 61 inch bars of type 4340 steel which were forged to 4-½-inch-diameter round electrodes. Compared to mild steel, this material melted with relatively stable arc characteristics, but compared to the vacuum melted zirconium electrodes, the material melted with a considerably less stable arc.

The patterns of the heat flux for the steel runs indicate differences which can occur during the melting of dissimilar materials, but these runs would have to be supplemented by additional data obtained during the melting of a variety of materials before specific conclusions could be stated. Such data would be of particular interest for the more refractory materials such as molybdenum, tantalum, columbium, and tungsten. When melting these materials, cooling of the cold-mold crucible is often marginal, which indicates the possibility of heat fluxes much greater than those calculated for the zirconium and steel heats conducted.

Materials other than steel and zirconium were not melted in the 8-inch crucible; however, both titanium and zirconium were melted in the 2-5/16-inch crucible The results of these small-scale runs indicated little difference in the maximum heat flux observed for the two metals, but the maximum heat flux observed for the titanium occurred further below the ingot top than for the zirconium,. The use of smaller diameter electrodes for the titanium runs would account for this shift in the location of the maximum heat flux.

The effects of arc potential; arc current, and power input were studied for a relatively narrow range of values at which normal operation of the arc occurred. The effects of abnormally high arc potentials or extremely high or low arc currents were not observed. Consequently the effects of these parameters were not as clearly shown as would be desired. Figures 15 and 16 include data from the zirconium and steel runs with 4-½-inch-diameter electrodes.

Figure 15 is a comparison of two runs at approximately equal power input but with different values of arc current and arc potential. The heat flux profile for the run made at the higher arc potential was shifted upward; that is, more heat was transferred to the wall above the ingot top with increased arc potential. Melting experience substantiates this effect, in fact; one way of maintaining a full molten pool is to increase the arc potential. This practice is especially effective in the researcher’s experience when melting and casting tungsten. Higher heat losses by radiation from the molten pool of

tungsten often cause a skin of metal to solidify from the sides inward across the top of the pool. This skin can be eliminated by operating at higher arc potentials.

Similar evidence of the effect of high arc potential is shown by the heat flux profiles for runs No. 7 and 8 in figure 16. These two runs were made at

approximately the same arc current but with different values of arc potential. A much larger percentage of the heat transfer for the run at higher voltage occurred above the top of the ingot.

When both arc current and arc potential were increased as in run 9 of figure 16, the peak value of the heat flux was increased as would be expected, but the heat flux profile was almost symmetrical with respect to the top of the ingot. Increased heat flux below the top of the ingot would be expected because of the deeper pool formed with increased power input. The higher arc potential and greater arc length would also cause high values of heat flux above the ingot top.

Run 6, which was not included in figure 15, was run at 7,200 amperes and 32 volts, and the curve for heat flux for this run falls between the curves for runs 4 and 5, as would be expected. The data presented in figures 15 and 16 are in agreement with the experience gained during melting. That is, for a given power input, a high arc potential and low arc current will yield a wide shallow pool while a low arc potential and a high arc current will yield a deeper, narrower pool.

The effects of arc potential were masked by the effect of electrode diameter for the 6-inch-diameter electrodes shown in figure 17, and the results of

run 1 of this series are not in agreement with the results expected. Maximum heat flux for run 1, which was made at 31 volts and only 5,700 amperes, exceeds the maximum heat flux for other runs conducted at both higher arc potential and arc current. The pattern of heat flux for this run does indicate an extremely shallow pool, however, and fits the general trends outlined in this respect. An additional study of the effect of extreme variations in arc current should be conducted to determine whether high heat flux can result under certain conditions at low power input,

Effect of Narrow Annular Spacing for Water Jackets

Experiments were conducted to determine the effectiveness of water jackets with a narrow annular spacing during nonconsumable electrode melting of zirconium in a 2-5/16-inch-diameter crucible. The outside diameter of this crucible was 2.750 inches, and the inside diameter of the water jacket used varied from 2.770 inches to 2.850 inches. The size of the annular spacing of the water jackets studied was expressed in terms of equivalent diameter which, for an annulus, is defined as the difference between the inside and outside diameters of the annulus. Tests were conducted for five different equivalent diameters of from 0.020 inch to 0.100 inch. Coolant flow rate was varied between 2,000 and 6,000 pounds per hour for each equivalent diameter, and the temperature of the crucible wall was recorded as a function of power input.

The primary purpose of these tests was to determine whether water jackets with very narrow annular spacings could be used for arc melting, and if difficulties would be experienced with crucible expansion or the formation of steam pockets. No such difficulties were encountered during tests conducted.

Figure 18 includes data obtained for 0.020 and 0.100-inch equivalent diameters at a coolant flow rate of 4,000 pounds per hour. Wall temperatures were obtained with an iron constantan thermocouple imbedded in the crucible

wall approximately ½ inch below the top of a zirconium ingot. The radial position of the thermocouple within the wall was not precisely determined; therefore, the temperatures are considered only a qualitative measure of the effectiveness of the particular water jacket being studied.

The data in figure 18 substantiate the fact that higher coolant velocities provide increased rates of heat transfer. These data, plus experience gained through research on the melting and casting of refractory metals at the Bureau, indicate that more consideration should be given to the design of water jackets which will provide high coolant velocities. These design considerations must be tempered by consideration of allowable pressure drops across the cooling jacket and mechanical problems of crucible and water jacket alinement.

Discussion

The studies conducted with the two-piece crucible yielded values of heat flux from the ingot to the crucible wall which are believed to be representative of values encountered during consumable electrode arc melting of iron and steel and the reactive metals. These data, plus experience gained in the melting and casting of refractory metals, provide guidelines for the design of water-cooled copper crucibles and their water jackets.

Crucible Wall Thickness

One of the design considerations on which considerable disagreement exists is the optimum thickness for the crucible wall. For an assumed heat flux of 10 6 Btu/ft² hr and temperature drop through the crucible wall of 700° F, the wall of a copper crucible could be more than 1.5 inches thick before conductance through the copper became a limiting factor. These assumed values of heat flux and temperature drop are conservative estimates based on the data obtained from these studies. It is believed that, up to 1.5 inches, wall thickness is not a critical factor from the standpoint of heat transfer. Heavy walls offer the advantage of greater longitudinal heat transfer along the wall of the crucible and increased mechanical strength. It is recommended that crucible wall thickness be determined from consideration of mechanical strength. If wall thicknesses exceed 1.5 inches, consideration must be made for heat conductance. Crucible walls less than one-fourth inch thick are not recommended.

Water Jacket Dimensions

An increase in the mass velocity, in pounds per hour per square foot of cross section, of a coolant will result in an increase in the coefficient of heat transfer from the crucible to the coolant. Thus, water jackets should be designed to provide optimum mass velocity of the coolant based on consideration of allowable pressure drop. Small-scale experimental work with very narrow annular water jackets yielded lower wall temperatures with increasing mass velocity, and no difficulty was encountered due to expansion of the crucible or the formation of steam pockets.

During visits at several industrial vacuum arc melting installations, it was noted that the use of a common water jacket with several different sizes of crucibles is a common practice. The use of a small diameter crucible in a water jacket designed for large crucibles results in a decrease in the mass velocity of the coolant for a fixed mass rate of flow. This practice evidently does not cause any difficulty, probably because of low rates of heat transfer involved. However, this practice may well prove unsatisfactory for melting more refractory metals when higher rates of heat transfer will be involved. In such cases, provisions should be made to provide increased coolant mass velocity by means of increased mass rate of flow, special water jackets, or inserts designed to reduce the annulus cross section.

The value of high coolant mass velocity has been demonstrated in vacuum arc melting and casting of tungsten by the Bureau. In this work, tungsten was melted in a 5-inch-diameter ladle using an arc current of 13,000 amperes at an arc potential of 40 to 42 volts. The annulus of the water jacket for this crucible had an inside diameter of 5-½ inches and an outside diameter of 6 inches. Coolant flow was maintained at 835 pounds per minute which was equivalent to a mass velocity of 1.6 x 10 6 pounds per hour per square foot. The wall thickness of the ladle was approximately 5/16 inch. No measure of the heat flux during tungsten melting has been attempted, but it is assumed to be much higher than for melting studies included in this report.

These few aspects of equipment design are not intended to be all inclusive but represent areas of importance about which questions have been raised. Such details of design as methods of grounding the crucible and the effects of nonsymmetrical magnetic fields set up by the arc have a great effect on the performance of vacuum arc furnaces but were not studied in this work.

Conclusions

During vacuum-arc melting in water-cooled copper crucibles, heat flux in excess of 10 6 Btu/ft² hr can occur. The effects of arc current, arc potential, ratio of electrode diameter-to-crucible diameter, crucible diameter, and electrode composition were studied and the following results were obtained:

1. The effect of arc current and arc potential were related. In general, increasing arc current increased the maximum heat flux. Increasing arc potential increased the maximum heat flux and shifted the location of the maximum heat flux nearer the top of the ingot. For a given power input, higher maximum heat flux will occur at high arc potential and low arc current.
2. For a given crucible diameter, higher heat flux will occur with larger diameter electrodes, and an increase in electrode diameter will shift the peak heat flux upward.
3. Melting on a small scale yields higher values of maximum heat flux because of instability of the arc during small-scale melting.
4. Electrode composition will affect values of heat flux because of differences in the thermal properties of the material melted and also because of gas content of the metal. An increase in gas content will shift the peak flux upward and increase the maximum heat flux.
5. Maximum values of heat flux observed during consumable-electrode arc melting of steel and zirconium electrodes were in the range of 0.2 x 10 6 to 0.5 x 10 6 Btu/ft² hr for 8-inch-diameter ingots. Values up to 1.4 x 10 6 Btu/ft² hr were observed during small-scale melting of zirconium and titanium.

Studies on narrow annular water jackets indicated that the high mass velocity achieved with narrow annuli yield lower crucible wall temperatures.

Additional data should be obtained, particularly with respect to the effect of the metal being melted and the effect of furnace pressure. In the case of the effect of the metals being melted, it would be of interest to conduct similar experiments with the refractory metals, particularly molybdenum, tantalum, and tungsten. Much higher values of maximum heat flux would be expected with these metals, and valuable information could be obtained which is needed for designing furnaces for melting these metals.

The effects of furnace pressure were not studied in detail although the small-scale runs were conducted at furnace pressures ranging from 50 to 400 torr, and the 8-inch-diameter runs were conducted at a furnace pressure of approximately 5 x 10-² torr. It is believed that it would be necessary to conduct melting studies at furnace pressures between 10-² torr and 50 torr to cover the range in which the greatest changes take place.

The studies were designed to measure the values of heat flux during normal melting conditions. It would be equally interesting to measure values of heat flux during abnormal arc conditions when maximum values of heat flux are undoubtedly very different.

A similar study should be conducted during electroslag melting. This method of melting is increasing in importance, and, based on examination of the ingots produced by this method, the patterns of heat transfer differ considerably from those during vacuum arc melting.