Table of Contents
Among the most important processes being developed for beneficiating nonmagnetic taconites is the reduction roasting-magnetic separation process. This process consists of first converting the nonmagnetic iron oxides of the ore to a magnetic form at an elevated temperature in a reducing atmosphere, after which the ore is concentrated by conventional magnetic separation methods. Of particular interest to the overall economics of this process is the reduction in costs that accompanies the grinding of the reduction roasted and quenched ore. The objective of this investigation was to study further this particular facet of the roasting-magnetic separation process.
Several previous research papers have discussed the effects of thermal treatment and cooling rate on the subsequent grindability of brittle materials. In general, it has been shown that heat treatment and rapid cooling can cause a marked improvement in ore friability. In addition, Person and Mitchell have studied a combination of heating, cooling, and reduction treatments on a nonmagnetic taconite. In their paper, the authors have shown the general effects that these variables have on the grindability of the ore in that heat treatment, rapid cooling, and roasting all improve the grindability of the ore; however, no attempt has been made to refine these effects. The present investigation was designed to determine and isolate more completely the effects of these variables. Through the use of a statistical analysis method, the order of variable influence and the presence or absence of variable interaction were established.
The as-received ore, a low-grade Upper Michigan hematite material, was initially closed-circuit cone crushed to minus 3/8 inch. Subsequently, the ore was stage crushed to a grindability test feed size of 3 mesh in a rolls crusher. The prepared feed material, containing about 36 percent iron, consisted principally of fine-grained hematite (mostly as martite) and chert with minor quantities of magnetite, quartz, and goethite. A photomicrograph of the untreated ore is shown in Figure 1, and a sieve analysis of the grindability test feed is presented in Table I. This particular ore was selected for testing because of its uniform mineralogy and refractory nature. The natural grain size of this ore occurs between 270 and 400 mesh. It was anticipated that an ore with these properties would produce wide variations in the comminution response when subjected to roasting treatments.
Heat treatment of the ore was accomplished in an electrically heated horizontal tube furnace using 5-pound charges (see equipment photo, Figure 2). As a means of minimizing variations in the ore, charges to be roasted were picked at random from a large group of blended and cut samples. Heating was done by placing the charged roasting tube into the furnace that had been preheated to 232° C. To provide a neutral atmosphere, a continuous 780 cm³/min flow of nitrogen gas was passed through the tube, revolving at 3 rpm, during the heating period. Rate of heating was held at approximately 13° C/min for all runs, while total heat-up time depended on the test temperature desired. Roasting temperatures were set at 450° C, 650° C, and 850° C. The choice of temperatures was such that the 575° C alpha to beta quartz transformation temperature was bracketed. Temperature within the reactor tube during roasting was controlled to within ± 6° C.
The chemical condition of the ore, one of the variables of the experimental design, was determined by the atmosphere in which the ore was roasted, either neutral or reducing. Time at roasting temperature was held constant at 1 hour for all roasting tests. Roasting of the ore in a neutral atmosphere was accomplished by continuing the flow of nitrogen at 780 cm³/min throughout the roasting period. Reduction roasting of the ore was done with a hydrogen-water vapor mixture. The flow rate of hydrogen was held constant at 1,920 cm³/min for all tests, while the percent moisture was varied, as a means of controlling the hematite to magnetite conversion reaction and to avoid unwanted production of wustite. The desired hydrogen-water vapor mixture was produced by percolating hydrogen through a column of water heated to a predetermined constant temperature. The conditions for generating the H2-H2O gas mixtures are detailed in Table II.
The roasted ore was cooled to ambient temperatures at two wide-spread rates, one accomplished by quenching and the other by slow cooling. To quench the heat treated material, the tube containing the ore at the test temperature was removed from the furnace and emptied into a container containing about 6 gallons of 24 ± 3° C tap water. Water temperatures were not increased by over 6° C. during quenching. Following quenching, the water was decanted and the samples were dried at 105° C.
Slow cooling of the roasted ore was accomplished, by controlling the heat loss of the cooling tube and ore. A procedure was developed so that a rate of 7 ± 1° C/min could be maintained during the full cooling period of the ore from roasting temperature to 204° C. Slow cooling of all samples was done in an inert atmosphere using a 780 cm³/min flow of nitrogen gas to eliminate possible oxidation of the magnetite.
Bond ball mill grindability and work indices were selected as parameters for comparing the condition of the treated and untreated ore. The standard, dry Bond ball mill grindability test has been used quite successfully in predicting commercial-size ball mill power requirements. The test procedure simulates a closed-circuit grinding operation with a circulating load of 250 percent, although it is actually run as a series of batch tests. Prior to performing grinding test, a 35-pound blended composite was made up from the individually roasted 5-pound samples.
In this investigation, grindability tests were made at three different size levels; 48, 100, and 400 mesh. The ball mill grindability index (Gbp) was obtained by averaging the net grams of sieve undersize product per mill revolution of the last three cycles of each individual grindability test after equilibrium was reached.
Work index (Wi) is a parameter which expresses the resistance of a material to comminution; numerically it is the kilowatt-hours/short ton required to reduce the material from theoretically infinite feed size to 80 percent passing 100 microns. The work indices were calculated from the ore grindability index, and the product size data were derived from the sieve analysis of ball mill feed and product of the last cycle. The equation’used in. these calculations was the 1960 empirical Bond formula (8);
Wi = 44.5/(P1)0.23 x (Gbp)0.82(10/√P – 10/√F)
where P1 is the micron size opening of the sieve size at which the test is being performed, Gbp is the ball mill grindability index in grams/revolution, and F and P refer to the 80 percent passing microns sizes of feed and last cycle product of the grindability test. During the full test series, the ball charge was maintained at a constant weight (about 22,590 grams), ball number (285), and surface area (5,600 sq cm).
A procedure consisting of a combination of dry and wet-dry sieving was developed to permit maximum reproducibility in material sizing. Ground products were first dry sieved through 200 mesh, allowing no more that 20 grams to accumulate on any one sieve. Where required, all minus 200 mesh material was wet screened and subsequently resieved dry. Again no more than 20 grams was allowed to accumulate on each sieve finer than 200 mesh. Sieves were frequently cleaned, with a wet ultrasonic cleaner.
A factorial test design was chosen for examining the influence of three variables on the ore grindability and work index. The test variables were:
- Roasting temperature: 450° C, 650° C, and 850° C.
- Roasted ore cooling rate: quench and slow cool.
- Ore chemical condition: reduced and unreduced.
Data from the investigation were analyzed using a three-way analysis of variance (ANOVA) technique where roasting temperature, ore chemical condition, and cooling rate were varied while holding the mesh size of the grindability test, 48, 100, and 400 mesh, constant. Prior to actual test work, the sequence of material roasting and grindability determinations was randomized to minimize the effect of testing sequence and technique change. A detailed description of the statistical methods used to evaluate the data can be found in Davies.
Table III presents both grindability index and work index data generated in the roasting test-series; however, only work index will be discussed in detail. The work index is considered to be more indicative of material hardness as it takes into consideration both the grindability index and the slope of the particle size distribution of the last cycle, sieve-undersize product. Repetitive runs at the 48 and 100 mesh level on the untreated material indicated that the work indices are reproducible to within ± 2 percent. Work indices generated from roasted material can be expected to be reproducible to within ± 5 percent.
Work index data from Table III have been extracted and plotted in the series of three figures using as a constant the mesh size of the grindability determinations. Figures 3, 4, and 5 are plots of work indices for 48, 100, and 400 mesh, respectively. Within each figure, four curves are plotted with roasting temperature as the variable using the four combinations of ore chemical condition and cooling rate.
Figure 6 shows the effect that ore chemical condition and cooling rate have on the Bond ball mill work index values at 48, 100, and 400 mesh while holding roasting temperature constant at 650° C.
Discussion of Results
The general improvement in grindability with increased heat treatment temperature is caused by thermal expansion and contraction of the constituent minerals. Several important forces are operative in determining the overall effect of the change in treatment temperature. In the case of uniform heating and cooling, the situation that existed for the heating and slow cooling cycles of this investigation, there was little difference in temperature between the exterior surface and center of the ore particles. Therefore, only minimal mechanical stresses resulting from differential thermal expansion between the exterior and inside of the particles were involved, whereas the thermal expansion forces between particles had a maximum effect.
The expansion which takes place in a crystal when its temperature is increased depends on the direction in which the expansion is measured relative to the crystal axis. As an example, quartz an anisotropic mineral, has a 7.97 x 10 -6 linear coefficient of thermal expansion parallel to its axis and 13.37 x 10 -6 expansion coefficient perpendicular to the axis.
The thermal expansion of a quartz material has been found to vary significantly between 0 and 1,000° C. Quartz increases its expansion rate linearly in the temperature range of 0 to 550° C. As the alpha to beta phase transformation temperature of 573° C is approached, the expansion rate undergoes a rather rapid increase. After reaching a maximum at the transformation temperature, the expansion rate decreases to a very low value at higher temperatures. Because the ore used in this study is about 50 percent quartz, the rapid improvement in the grindability between 450° C and 650° C for each chemical condition-cooling rate combination can be attributed largely to thermal expansion of quartz in this temperature region.
In a heterogeneous solid, the situation in this case with an iron oxide and quartz present, the effects of thermal expansion are magnified by the difference in the coefficients of the ore components and the individual crystal orientations. On heating and cooling of the ore, localized areas of high stresses and strains will be set up at the grain boundaries which can both cause cracks or become loci for crack initiation.
A decrease in grindability is evident in most cases when comparing the untreated ore work indices to the work indices obtained from unreduced ore at 450° C for the three grindability levels. This phenomenon is thought to be caused by the stress relaxation and annealing of cracks originally present in the feed material. This effect is rapidly overcome at higher roasting temperatures.
The effect of thermal shock in the form of quench cooling is evident in examining Figures 3, 4, and 5. In comparing the quenched to slow cooled materials for both the reduced and unreduced conditions, thermal shock has had a positive effect on grindability. Advantage was taken of nonuniform rapid cooling in quenching. When quenching into water, the outer layer of the ore particles rapidly contracts, causing the appearance of tensile stresses on the surface while the particles in the interior remain under compression. Minerals exhibit their maximum strength under compression and minimum strength under tension. Therefore, tensile cracks were initiated at the particle surface and propagated inward, causing both transgranular and intergranular breakage to occur as the particles rapidly cooled.
Improvements in grindability where ore was uniformly heated and cooled were due principally to intergranular breakage and were limited by the natural grain size of the material. Quenching, because of both the intergranular and transgranular nature of the breakage, can significantly improve the grindability past the natural grain size limit. This is shown by comparing the unreduced-slow cool to unreduced-quench data in Figures 3, 4, and 5. Additionally, the divergence of the curves at higher temperatures is thought to be caused by the increased tensile forces that accompany quench cooling at the higher treatment temperatures. The magnitude of the tensile stress at the particle surface when quench cooling can be shown to be proportional to the difference between the initial particle temperature and the temperature to which the particle surface is cooled.
Ore Chemical Condition A major improvement in ore grindability resulted from the hydrogen reduction of hematite to magnetite at all temperatures investigated. The overall reduction reaction can be summarized by the following equation where hexagonal hematite is converted to isometric magnetite by removing part of the oxygen from the hematite structure:
3Fe2O3 + H2 → 2FeO·Fe2O3 + H2O
In this conversion reaction, a rather large volume increase takes place as the hematite is converted to magnetite and the resulting material is left in a more porous condition. The major positive influence of the reduction reaction of grindability is illustrated by comparing both the quench and slow cool curves of the reduced material to those of the unreduced materials at the three grindability levels.
The work index curves for the reduced ore are very similar in shape to those obtained for the unreduced material between 450° C and 650° C, indicating that reduction has had a rather uniform effect in this temperature range. For the reduced materials, a minimum work index value is determined to be near 625° C. This minimum becomes less pronounced as the fineness of the grindability test level is increased. At the 400-mesh level a distinct minimum no longer exists, indicating that this phenomenon is probably associated with the natural grain size of the material. The natural grain size of this ore occurs between 270 and 400 mesh.
In comparing the results of the two cooling rate treatments, it can be noted that the effect is more pronounced on unreduced than on reduced ore. It is very probable that the large effect that reduction, has had on the ore has masked the cooling rate effect.
The effects that cooling rate and chemical condition have on the ore are graphically illustrated when comparing the photomicrographs of the treated and untreated materials in Figure 7. Figure 7A shows ore in an untreated, condition. No fracturing is present within the particle interiors. The effects of quench-cooling following a neutral atmosphere roast at 650° C are shown in Figure 7B. A significant amount of fracturing has taken place along quartz-iron oxide grain boundaries with additional fracturing along both the quartz-quartz and iron oxide-iron oxide grain boundaries.
The effects of optimum roasting and cooling treatments are shown in Figure 7C in which the ore has been reduction roasted at 650° C and quench-cooled. The treated material is highly fractured both along the grain boundaries and across the individual grains. The material softness is evident in that many of the individual grains were dislodged during the polishing of the briquet surface. The softness of the reduced material was also apparent in that ore particles could be easily broken between the fingers, with the broken surface having a sugary appearance.
Using optimum roasting and cooling conditions, reduction roasting at 650° C, and quench-cooling, the grindability tests show that substantial reduction in grinding power requirements can be realized. At 48 mesh, the work index of the treated ore has been reduced to 1.66, about 13 percent of the original untreated ore work index. At the 100- and 400-mesh levels, work indices were reduced to 2.18 and 7.36, respectively, about 20 and 60 percent, respectively, of the untreated material work indices.
An interesting phenomenon is observed when comparing the work indices of the treated material as the roasting temperature, atmosphere, and cooling rate are held constant while varying the grindability mesh level, see Figure 6. The work index determinations between 48 and 100 mesh are of similar magnitude; however, a substantial increase takes place between 100 and 400 mesh. A plausible explanation would be that in grinding to the 400-mesh level, the product size is finer than the natural grain size of the ore. Higher power inputs would be needed to fulfill the increasing transgranular breakage requirements as the grindability level approached and exceeded the natural grain size of the material.
Statistical Analysis of Data
One of the more important objectives of this investigation was to determine the order of importance of the variables and if any interactions existed between the variables. All analysis of variance (ANOVA) technique was used to examine the data in which a level of 5 percent or greater was’ ‘chosen as the limiting value for significance. As explained in a preceding section, this investigation was set up as a 3 x 2 x 2 x 3 factorial experiment, using as variables roasting temperature, ore chemical condition, cooling rate, and grindability mesh level, respectively. Analysis of the data, however, was made using a three-way ANOVA method of analysis in which grindability mesh level was held constant while varying the remaining three variables.
No attempt was made to replicate any of the data points within the factorial design. An assumption was therefore made that the second order of three-way interaction, temperature x chemical condition x cooling rate, was small or nonexistent, and could be included in the residual along with the error. Using these assumptions, it was concluded that all two-way interactions, temperature x chemical condition, temperature x cooling rate, and chemical condition x cooling rate, were not significant at any of the three grindability mesh levels. There is some evidence on close inspection of Figures 3, 4, and 5, however, that indicates one or more interactions may be important. To test for this situation, it would have been necessary to get an estimate of error independent of the interaction in which case replication of the experiment would be required.
In comparing the order of the main effects, chemical condition of the ore was found to be by far the most significant of the three variables. Treatment temperature was found to be the next most significant, while cooling rate was found to have the least effect on ore grindability. The order of variable influence was found to be consistent for the three grindability levels; however, it was noted that as the grindability mesh level was successively varied from 48 to 400 mesh, the influence of cooling rate diminished. At 400 mesh, the influence of cooling rate had decreased to the point where it was only marginally significant at 5 percent.
A factorially designed experiment has been used to determine and clarify the effects of roasting temperature, ore chemical condition, cooling rate, and grindability mesh level on the grindability of a low-grade hematitic ore. A summary of the salient conclusions for the investigation is as follows:
- Under optimum heat treatment and cooling conditions, grinding power requirements are substantially reduced. At 48 mesh, the work index was reduced to about 13 percent of that obtained from the untreated ore, while at 100 mesh and 400 mesh, the work indices had been reduced to about 20 and 60 percent, respectively, compared with those of the untreated ore.
- The sieve size at which grindability is determined has an important influence on work index values. For determinations made on similarly conditioned ore at sieve openings larger than the natural grain size, work index values appear to be rather constant; however, when making determinations at sieve openings approaching or smaller than the natural grain size, a substantial increase in the work index was found. This is attributed mainly to the increased power required to accomplish transgranular breakage.
- Optimum heat treatment conditions are found to be near the quartz alpha to beta phase transformation temperature of 575° C.
- Application of thermal shock in the form of quench cooling is found to improve ore grindability under all conditions. The effect of thermal shock on unreduced ore diminishes as the grindability mesh size became finer and is less pronounced on reduced as compared to unreduced ore.
- Of the three treatment factors evaluated, reduction roasting of the ore causes by far the largest improvement in grindability. Roasting temperature has the next most important positive effect, while cooling rate has the least effect of the three factors.
- Using an analysis of variance technique and assuming the three- way interaction, temperature x ore chemical condition x cooling rate, is small or nonexistent, all two-way interactions are found to be nonsignificant at a 5-percent level.