Shell

Mill shells are designed to sustain impact and heavy loading, and are constructed from rolled mild steel plates connected together. Holes are drilled to take the bolts for holding the liners. For attachment of the trunnion heads, heavy flanges of fabricated or cast steel are usually welded or bolted to the ends of the plate shells, planed with parallel faces that are grooved to receive a corresponding spigot on the head, and drilled for bolting to the head.

Mill Ends

Also known as trunnion heads, they may be of nodular or gray cast iron for diameters less than about 1 m; larger heads are constructed from cast steel, which is relatively light and can be welded. The heads are ribbed for reinforcement and may be flat, slightly conical, or dished. They are machined and drilled to fit shell flanges (Figure 7.7).

Trunnions and Bearings

The trunnions are made from cast iron or steel and are spigoted and bolted to the end plates, although in small mills they may be integral with the end plates (see also Anon., 1990). They are highly polished to reduce bearing friction. Most trunnion bearings are rigid high-grade iron castings with 120–180° lining of white metal in the bearing area, surrounded by a fabricated mild steel housing, which is bolted into the concrete foundations (Figure 7.8).

The bearings in smaller mills may be grease lubricated, but oil lubrication is favored in large mills, via motor-driven oil pumps. The effectiveness of normal lubrication protection is reduced when the mill is shut down for any length of time, and many mills are fitted with manually operated hydraulic starting lubricators, which force oil between the trunnion and trunnion bearing, preventing friction damage to the bearing surface on starting by re-establishing the protecting film of oil (Figure 7.9).

Drive

Tumbling mills are most commonly rotated by a pinion meshing with a girth ring bolted to one end of the machine (Figure 7.10). The pinion shaft is either coupled directly or via a clutch to the output shaft of a slow-speed synchronous motor, or to the output shaft of a motor-driven helical or double helical gear reducer. In some mills, electrical thyristors and/or DC motors are used to give variable speed control. Very large mills driven by girth gears require two motors, each driving separate pinions, with a complex load sharing system balancing the torque generated by the two motors. (See also Knecht, 1990.)

The larger the mill, the greater are the stresses between the shells and heads and the trunnions and heads. In the early 1970s, maintenance problems related to the application of gear and pinion and large speed reducer drives on dry grinding cement mills of long length drove operators to seek an alternative drive design. As a result, a number of gearless drive (ring motor) cement mills were installed and the technology became relatively common in the European cement industry.

The gearless drive design features motor rotor elements bolted to a mill shell, a stationary stator assembly surrounding the rotor elements, and electronics converting the incoming current from 50/60 Hz to about 1 Hz. The mill shell actually becomes the rotating element of a large low speed synchronous motor. Mill speed is varied by changing the frequency of the current to the motor, allowing adjustments to the mill power draw as ore grindability changes.

The gearless drive design was not applied to the mills in the mineral industry until 1981 when the then-world’s largest ball mill, 6.5 m diameter and 9.65 m long driven by a 8.1 MW motor, was installed at Sydvaranger in Norway (Meintrup and Kleiner, 1982). A gearless drive SAG mill, 12 m diameter and 6.1 m length (belly inside liners) with a motor power of more than 20 MW, went into operation at Newcrest Mining’s Cadia Hill gold and copper mine in Australia, with a throughput of over 2,000 t h⁻¹ (Dunne et al., 2001). Motor designs capable of 35 MW have been reported (van de Vijfeijken, 2010).

The major advantages of the gearless drive include: variable speed capacity, removal of limits of design power, high drive efficiency, low maintenance requirements, and less floor space for installation.

Liners

The internal working faces of mills consist of renewable liners, which must withstand impact, be wear-resistant, and promote the most favorable motion of the charge. Rod mill ends have plain flat liners, slightly coned to encourage the self-centering and straight-line action of rods. They are made usually from manganese or chrome-molybdenum steels, having high impact strength. Ball mill ends usually have ribs to lift the charge with the mill rotation. The ribs prevent excessive slipping and increase liner life. They can be made from white cast iron, alloyed with nickel (Ni-hard), other wear-resistant materials, and rubber (Durman, 1988). Trunnion liners are designed for each application and can be conical, plain, with advancing or retarding spirals. They are manufactured from hard cast iron or cast alloy steel, a rubber lining often being bonded to the inner surface for increased life.

Shell liners have an endless variety of lifter shapes. Smooth linings result in much abrasion, and hence a fine grind, but with associated high metal wear. The liners are therefore generally shaped to provide lifting action and to add impact and crushing. From a survey (Wei and Craig, 2009), the most common shapes were wave, rib, step, and Osborn (Figure 7.11). The liners are attached to the mill shell and ends by forged steel countersunk liner bolts.

Rod mill liners are generally of alloyed steel or cast iron, and of the wave type, although Ni-hard step liners may be used with rods up to 4 cm in diameter. Lorain liners consist of high carbon rolled steel plates held in place by manganese or hard alloy steel lifter bars. Ball mill liners may be made of hard cast iron when balls of up to 5 cm in diameter are used, but otherwise cast manganese steel, cast chromium steel, or Ni-hard are used.

Efforts to prolong liner life are constantly being made. With the lost production cost associated with shut-downs for replacing liners, the trend is toward selecting liners that have the best service life and least relining down-time (Orford and Strah, 2006).

Rubber liners and lifters have supplanted steel at some operations, particularly in ball mills. They have been found to be longer lasting, easier and faster to install, and their use results in a significant reduction of noise level. In primary grinding applications with severe grinding forces, the higher wear rate of rubber tends to inhibit its use. Rubber lining may have drawbacks in processes requiring the addition of flotation reagents directly into the mill, or temperatures exceeding 80°C. They are also thicker than their steel counterparts, which reduces mill capacity, a potentially important factor in small mills. There are also important differences in design aspects between steel and rubber linings (Moller and Brough, 1989). A combination of rubber lifter bars with steel inserts embedded in the face, the steel providing the wear resistance and the rubber backing cushioning the impacts, is a compromise design (Moller, 1990).

A different concept is the magnetic liner. Magnets keep the lining in contact with the steel shell and the end plates without using bolts, while the ball “scats” in the charge and any magnetic minerals are attracted to the liner to form a 30–40 mm protective layer, which is continuously renewed as it wears.

Mill Feeders

The type of feeding arrangement used on the mill depends on whether the grinding is done in open or closed circuit and whether it is done wet or dry. The size and rate of feed are also considerations. Dry mills are usually fed by some sort of vibratory feeder. Three types of feeder are in use in wet-grinding mills. The simplest form is the spout feeder (Figure 7.12), consisting of a cylindrical or elliptical chute supported independently of the mill, and projecting directly into the trunnion liner. Material is fed by gravity through the spout to the mills. They are often used for feeding rod mills operating in open circuit or mills in closed circuit with hydrocyclone classifiers.

An alternative to a spout feeder, drum feeders (Figure 7.13) may be used when headroom is limited. The mill feed enters the drum via a chute or spout and an internal spiral carries it into the trunnion liner. The drum also provides a convenient method of adding grinding balls to a mill.

Combination drum-scoop feeders (Figure 7.14, see also Figure 7.2) are generally used for wet grinding in closed circuit with a spiral or rake classifier. New material is fed directly into the drum, while the scoop picks up the classifier sands for regrinding. Either a single or a double scoop can be used, the latter providing an increased feed rate and more uniform flow of material into the mill; the counter-balancing effect of the double-scoop design serves to smooth out power fluctuation and it is normally incorporated in large-diameter mills. Scoop feeders are sometimes used in place of the drum-scoop combination when mill feed is in the fine-size range.

Rod Mills

These mills can be considered as either fine crushers or coarse grinding machines. They are capable of taking feed as large as 50 mm and making a product as fine as 300 µm, reduction ratios (F₈₀:P₈₀) normally being in the range 15:1–20:1. They are often preferred to fine crushing machines when the ore is “clayey” or damp, thus tending to choke crushers. They are still to be found in older plants but are often replaced upon modernization (e.g., Leung et al., 1992) and are now relatively rare (Wei and Craig, 2009).

The distinctive feature of a rod mill is that the length of the cylindrical shell is between 1.5 and 2.5 times its diameter (Figure 7.15). This ratio is important because the rods, which are only a few centimeters shorter than the length of the shell, must be prevented from turning so that they become wedged across the diameter of the cylinder (a rod tangle). The ratio must not, however, be so large for the maximum diameter of the shell in use that the rods deform and break. Since rods longer than about 6 m will bend, this establishes the maximum length of the mill. Thus, with a mill 6.4 m long the diameter should not be over 4.57 m. Rod mills of this size have seen use, driven by 1,640 kW motors (Lewis et al., 1976).

Rod mills are classed according to the nature of the discharge. A general statement can be made that the closer the discharge is to the periphery of the shell, the quicker the material will pass through and less overgrinding will take place.

Center peripheral discharge mills (Figure 7.16) are fed at both ends through trunnions and discharge the ground product through circumferential ports at the center of the shell. The short path and steep gradient give a coarse grind with a minimum of fines, but the reduction ratio is limited. This mill can be used for wet or dry grinding and has found its greatest use in the preparation of specification sands, where high tonnage rates and an extremely coarse product are required.

End peripheral discharge mills (Figure 7.17) are fed at one end through the trunnion, discharging the ground product from the other end of the mill by means of several peripheral apertures into a close-fitting circumferential chute. This type of mill is used mainly for dry and damp grinding, where moderately coarse products are involved.

The most widely used type of rod mill in the mining industry is the trunnion overflow (Figure 7.18), in which the feed is introduced through one trunnion and discharges through the other. This type of mill is used only for wet grinding and its principal function is to convert crushing-plant product into ball mill feed. A flow gradient is provided by making the discharge (overflow) trunnion diameter 10–20 cm larger than that of the feed opening. The discharge trunnion is often fitted with a spiral (trommel) screen to remove tramp material.

Rod mills are charged initially with a selection of rods of assorted diameters, the proportion of each size being calculated to approximate to a seasoned or equilibrium charge. A seasoned charge will contain rods of varying diameters, ranging from fresh replacements to those that have worn down to such a size as to warrant removal. Actual diameters in use range from 25 to 150 mm. The largest diameter should be no greater than that required to break the largest particle in the feed. A coarse feed or product normally requires larger rods. Provided the large feed particles are broken, the smaller the rods the larger is the total surface area and hence the greater is the size reduction. Generally, rods should be removed when they are worn down to about 25 mm in diameter or less, depending on the application, as small ones tend to bend or break, filling the mill with “scats”. High carbon steel rods are used as they are hard, and break rather than warp when worn, and so do not entangle with other rods.

Optimum grinding rates are obtained with about 45% charge filling (see Section 7.2.2). Overcharging results in inefficient grinding and increased liner and rod consumption; reduced rod consumption may indicate an economic optimum at less than 45% filling. Rod consumption varies widely with the characteristics of the mill feed, mill speed, rod length, and product size; it is normally in the range 0.1–1.0 kg of steel per ton of ore for wet grinding, being less for dry grinding.

Rod mills are normally run at between 50% and 65% of the critical speed, so that the rods cascade rather than cataract; many operating mills have been sped up to close to 80% of critical speed without any reports of excessive wear (McIvor and Finch, 1986). The feed pulp density is usually between 65% and 85% solids by weight, finer feeds requiring lower pulp densities. The grinding action results from line contact of the rods on the ore particles; the rods tumble in essentially a parallel alignment, and also spin, thus acting rather like a series of crushing rolls. The coarse feed tends to spread the rods at the feed end, so producing a wedge- or cone-shaped array. This increases the tendency for grinding to take place preferentially on the larger particles, thereby producing a minimum amount of extremely fine material (Figure 7.19). This selective grinding gives a product of relatively narrow size range, with little oversize or slimes. Rod mills are therefore suitable for preparation of feed to gravity concentrators, certain flotation processes with slime problems, magnetic cobbing, and ball mills. They are nearly always run in open circuit because of this controlled size reduction.

Ball Mills

The final stages of primary comminution are performed in tumbling mills using steel balls as the grinding medium, and so are designated “ball mills.” A typical ball mill setup is shown in Figure 7.20.

Since balls have a greater surface area per unit weight than rods, they are better suited for fine grinding. The term ball mill is restricted to those having a length to diameter ratio of 2 to 1 and less. Ball mills in which the length to diameter ratio is between 3 and 5 are designated tube mills. The latter are sometimes divided into several longitudinal compartments, each having a different charge composition; the charges can be steel balls or rods, or pebbles, and they are often used dry to grind cement clinker, gypsum, and phosphate. Tube mills having only one compartment and a charge of hard, screened ore particles as the grinding medium are known as pebble mills. They have the advantage over ball mills when iron contamination needs to be avoided. Since the weight of pebbles per unit volume is about 35–55% of that of steel balls, and as the power input is directly proportional to charge weight, the power draw and capacity of pebble mills are correspondingly lower (see Section 7.2.2). Thus in a given grinding circuit, for a certain feed rate, a pebble mill would be much larger than a ball mill, with correspondingly higher capital cost. The increment in capital cost might be justified by reduction in operating cost attributed to the lower cost of the grinding medium, provided this is not offset by higher energy cost per ton of finished product (Lewis et al., 1976).

Ball mills are also classified by the nature of the discharge. They may be simple trunnion overflow mills, operated in open or closed circuit, or grate discharge (low-level discharge) mills. The latter type is fitted with discharge grates between the cylindrical mill body and the discharge trunnion. The pulp can flow freely through the openings in the grate and is then lifted up to the level of the discharge trunnion (Figure 7.21). These mills have a lower pulp level than overflow mills, thus reducing the residence time of particles in the mill. Consequently, little overgrinding takes place and the product contains a large fraction of coarse material, which is returned to the mill by some form of classifying device. Closed-circuit grinding, with high circulating loads, produces a closely sized end product and a high output per unit volume compared with open circuit grinding. Grate discharge mills usually take a coarser feed than overflow mills and are not required to grind so finely, the main reason being that with many small balls forming the charge, the grate open area plugs very quickly. The trunnion overflow mill is the simplest to operate and is used for most ball mill applications, especially for fine grinding and regrinding. Energy consumption is said to be about 15% less than that of a grate discharge mill of the same size, although the grinding efficiencies of the two mills are similar (Lewis et al., 1976).

The trend in recent years has been to use fewer comminution machines per grinding line with the result that units have increased considerably in size and thus capacity. For example, in the 1980s, the largest operating ball mill was 5.5 m in diameter by 7.3 m in length driven by a 4 MW motor. Today the largest ball mills in operation are over 8 m in diameter utilizing gearless (wrap-around) motors with power output of more than 16 MW.

Grinding in a ball mill is effected by point contact of balls and ore particles and, given time, any degree of fineness can be achieved. The process is completely random—the probability of a fine particle being struck by a ball is the same as that of a coarse particle. The product from an open-circuit ball mill therefore exhibits a wide range of particle size, and overgrinding of at least some of the ore can be a problem. Closed-circuit grinding in mills providing low residence time for the particles is almost always used in the last stages to overcome this.

Several factors influence the efficiency of ball mill grinding. The pulp density of the feed should be as high as possible, consistent with ease of flow through the mill. It is essential that the balls are coated with a layer of ore; too dilute a pulp increases metal-to-metal contact, giving increased steel consumption and reduced efficiency. Ball mills should operate between 65% and 80% solids by weight, depending on the ore. The viscosity of the pulp increases with the fineness of the particles, therefore fine-grinding circuits may need lower pulp densities. The major factors affecting the pulp rheology and its effects on grinding circuits have been discussed by a number of researchers (Klimpel, 1984; Kawatra and Eisele, 1988; Moys, 1989; Shi and Napier-Munn, 2002; He et al., 2004; Pax, 2012). It was found that not only the viscosity of the pulp but also the rheological type, Newtonian or non-Newtonian, would affect ball milling performance. An extreme case of high viscosity slurry is grinding fibrous ore particles where mills have to operate at low %solids.

The efficiency of grinding depends on the surface area of the grinding medium. Thus, balls should be as small as possible and the charge should be graded such that the largest balls are just heavy enough to grind the largest and hardest particles in the feed. A seasoned charge will consist of a wide range of ball sizes and new balls added to the mill are usually of the largest size required. Undersize balls leave the mill with the ore product and can be removed by passing the discharge through trommel screens or by magnets mounted near the discharge trunnion. Various formulae have been proposed for the required ratio of ball size to ore size, none of which is entirely satisfactory. The correct sizes are often determined by trial and error, primary grinding usually requiring a graded charge of 10-5 cm diameter balls, and secondary grinding requiring 5-2 cm. Concha et al. (1988) have developed a method to calculate ball mill charge by using a grinding circuit simulator with a model of ball wear in a tumbling mill.

Segregation of the ball charge within the mill is achieved in the Hardinge mill (Figure 7.22). The conventional drum shape is modified by fitting a conical section, the angle of the cone being about 30°. Due to the centrifugal force generated, the balls are segregated so that the largest are at the feed end of the cone, that is, the largest diameter and greatest centrifugal force, and the smallest are at the discharge. By this means, a regular gradation of ball size and of size reduction is produced.

There are also non-spherical shapes of grinding media such as Doering Cylpebs. The Cylpebs are slightly tapered cylindrical grinding media with length equaling diameter, and all the edges being rounded (Figure 7.23). They are available in sizes from 85 mm×85 mm to 8 mm×8 mm. Because of their geometry, these grinding media have greater surface area than balls of the same mass. For instance, Cylpebs with a diameter to height ratio of unity have 14.5% more surface area than balls. It was thus supposed that these grinding media would produce more fines in a grinding mill than balls. However, in laboratory tests it was found that the comparison depended on what was held constant between the two media types: mass, size distribution, or surface area (Shi, 2004).

Grinding balls are usually made of forged or rolled high-carbon or alloy steel, cast alloy steel and some of white iron (Rajagopal and Iwasaki, 1992). Consumption varies between 0.1 and as much as 1 kg t⁻¹ of ore depending on hardness of ore, fineness of grind, and medium quality. Medium consumption can be a high proportion, sometimes as much as 40% of the total milling cost, so is an area that often warrants special attention. Good quality grinding media may be more expensive, but may be economic due to lower wear rates. Finer grinding may lead to improved metallurgical efficiency, but at the expense of higher grinding energy and media consumption. Therefore, particularly with ore of low value, where milling costs are crucial, the economic limit of grinding has to be carefully assessed (Chapter 1).

As the medium consumption contributes significantly to the total milling cost, great effort has been expended in the study of medium wear. Three wear mechanisms are generally recognized: abrasion, corrosion, and impact (Rajagopal and Iwasaki, 1992). Abrasion refers to the direct removal of metal from the grinding media surface. Corrosion refers to electrochemical reactions with oxygen, sometimes enhanced by galvanic effects with sulfide minerals (see Chapter 12); continual removal of oxidation products by abrasion means the fresh surface is always being exposed to such reactions. Impact wear refers to spalling, breaking, or flaking (Misra and Finnie, 1980; Gangopadhyay and Moore, 1987). Operational data show that abrasion is the major cause of metal loss in coarse (primary) grinding, with more corrosion occurring in finer regrinding (Dunn, 1982; Dodd et al., 1985). Corrosion products (iron oxy-hydroxides) can be detrimental to flotation and has led to the use of “inert” media, including high chrome balls, but also ceramic and other non-iron media (see Chapter 12).

Attempts have been extended to predict media wear by developing a model incorporating the abrasive, corrosive, and impact wear mechanisms (Radziszewski, 2002). An example is the following empirical ball wear model for forged carbon steel based on fitting data from 46 Peruvian mills to ore hardness, pulp acidity and particle top size given by Rabanal and Guzmán (2013):

${\mathrm{\Omega }}_E{d}^R=K_d^E=1.36\left[{\left(\frac{A_i-0.05}{0.20}\right)}^{0.166}{\left(\frac{F_{80}}{5000}\right)}^{0.069}{\left(\frac{pH}{10}\right)}^{-0.243}\right]$ (7.10)

(7.10)

where Ω_E is the energy-corrected ball wear rate, g per kWh, d^R is the diameter of the largest balls in the mill (the recharge size), mm, $K_d^E$ is the linear wear rate of balls, µm per kWh t⁻¹, Ai is the Bond abrasion index determined in a laboratory test (unitless), F₈₀ is the feed 80% passing size of the ore, µm, and pH is the water acid/base measurement of the mill pulp.

The charge filling in a ball mill is about 30–45% of the internal volume of the mill, about 40% of this being void space. As discussed in Section 7.2.2, the reason is the power draw passes through a maximum at about a filling of 45%. Given the cost of media it may be economic to operate at the lower end of the filling range, even at the expense of a loss in energy efficiency, the same point noted for rod mills.

Ball mills are often operated at higher speeds than rod mills, so that the larger balls cataract and impact on the ore particles. The work input to a mill increases in proportion to the speed, and ball mills are run at as high a speed as is possible without centrifuging. Normally this is 70–80% of the critical speed.

Autogenous/Semi-autogenous Mills

The highest throughput grinding circuits in the mining industry use autogenous grinding (AG) or semi-autogenous grinding (SAG) mills. An AG mill is a tumbling mill that uses the ore itself as grinding media. The ore must contain sufficient competent pieces to act as grinding media and preferably be high specific gravity (s.g.) which, for example, favors AG milling of iron ores (s.g. 4 vs. 2.7 for high silicate ores). A SAG mill is an autogenous mill that uses steel balls in addition to the natural grinding media. A typical setup is shown in Figure 7.24. The ball charges in SAG mills have generally been most effective in the range of 4–15% of the mill volume, including voids. As noted in Section 7.2.2 there is a case to increase ball charge to increase power draw. In South African practice SAG mills can have a ball charge as high as 35% of mill volume (Powell et al., 2001).

The first paper describing ore as grinding media was delivered to the American Institute of Mining and Metallurgical Engineers in 1908. In the early 1930s, Alvah Hadsel installed the Hadsel crushing and pulverizing machine, which was later improved by the Hardinge Company and was called the Hardinge Hadsel Mill. Early deployment of AG mills can be attributed primarily to a need in the iron ore industry to economically process large quantities of ore in the late 1950s. The high density of iron ores compared to many others probably favored AG milling. Non-ferrous operations (mostly copper and gold) utilized AG milling to a lesser extent before recognizing that SAG milling was better able to handle a variety of ore types. By December 2000, at least 1075 commercial AG/SAG mills had been sold worldwide with a total installed power exceeding 2.7 GW. SAG mills installed in Chile now regularly exceed 100,000 tons per day treated in a single mill.

AG/SAG mills usually replace the final two stages of crushing (secondary and tertiary) and rod milling of the traditional circuit (Chapter 6). This gives some advantages over the traditional circuit: lower capital cost, ability to treat a wide range of ore type including sticky and clayey feeds, relatively simple flowsheets, large size of available equipment, lower manpower requirements, and reduced steel consumption (noting that these mills replace some crushing/rod milling stages). The use of AG/SAG milling has grown to the point where many existing plants are retrofitting them, whilst new plants rarely consider a design that does not include them. This may not continue in the future, as new technologies such as fine crushing, high pressure grinding rolls, and ultra fine grinding (stirred milling) offer alternative flowsheet options (Erikson, 2014).

A schematic of various sections of the autogenous grinding mill is shown in Figure 7.25. AG or SAG mills are defined by the aspect ratio of the mill shell design and the product discharge mechanism. The aspect ratio is defined as the ratio of diameter to length. Aspect ratios generally fall into three groups: high aspect ratio mills where the diameter is 1.5–3 times of the length, “square mills” where the diameter is approximately equal to the length, and low aspect ratio mills where the length is 1.5–3 times that of the diameter. Although Scandinavian and South African practice favors low aspect ratio AG/SAG mills, in North America and Australia the mills are distinguished by high aspect ratios. The largest SAG mills are up to 12 m in diameter by 6.1 m length (belly inside liners), driven by a motor power of more than 20 MW. Many high aspect ratio mills, particularly the larger diameter units, have conical rather than flat ends. Attention should therefore be given to the mill length definition.

Inside the milling chamber the mill is lined with wearing plates held by lifter bars bolted to the shell. Lifter bars are essential to reduce slippage of the mill load, which causes rapid wear of the liners and also impairs the grinding action. The shape and geometry of the lifter bars, particularly the height and the face angle, have a significant influence on milling performance and wear rates.

The grate, as shown in Figure 7.25, is used to hold back the grinding media and large rock pieces and allow fine particles and slurry to flow through. Shapes of the grate apertures can be square, round, or slotted, with size varying from 10 to 40 mm. In some installations there are large holes varying from 40 to 100 mm in the grate, designated as pebble ports, which allow pebbles to be extracted and crushed (to ca. 12–20 mm) prior to recirculation to the mill. The total open area of the grate is ca. 2–12% of the mill cross-sectional area.

The slurry of particles smaller than the grate apertures discharges into the pulp lifter chambers. The pulp lifters, which are radially arranged, as shown in Figure 7.25, rotate with the mill and lift the slurry into the discharge trunnion and out of the mill. Each pulp lifter chamber is emptied before its next cycle to create a gradient across the grate for slurry transportation from the milling chamber into the pulp lifter chambers.

There are two types of pulp lifter design: radial and curved (also known as spiral, refer to Figure 7.26). The radial (or straight type) is more common in the mineral processing industry.

While the major portion of the slurry passing through the grate is discharged from the mill via the discharge trunnion, a proportion of the slurry in the pulp lifter chamber flows back into the mill as the mill rotates. This flow-back process often leads to higher slurry holdup inside the mill, and may sometimes contribute to the occurrence of “slurry pooling”, which has adverse effects on the grinding performance (Morrell and Kojovic, 1996). To improve the slurry transportation efficiency of the pulp lifters, particularly where slurry pooling limits capacity, a new concept of a Twin Chamber Pulp Lifter was developed (Latchireddi and Morrell, 2003, 2006) and first tested in a bauxite plant (Alcoa’s 7.7 m diameter mill) in Western Australia (Morrell and Latchireddi, 2000). In this design, the slurry first enters the section exposed to the grate, the transition chamber, and then flows into the lower section, the collection chamber, which is not exposed to the grate. This mechanism prevents the pulp from flowing backward into the mill, which can significantly increase the capacity of the mill. The Turbo Pulp Lifter from Outotec is a similar concept. DEM analysis has helped refine lifter design to minimize pulp flow back. With curved pulp lifters the rotation of the mill cannot be reversed (used to extend liner life); in consequence asymmetric liner profiles are employed designed to resist wear on the leading side.

AG/SAG milling may be wet or dry. Dry AG mills were once common in iron ore grinding. They have more workplace environmental problems (dust), do not handle materials containing clay well, and are more difficult to control than wet mills. In certain talc and mica grinding applications dry SAG mills are operated.

AG/SAG mills can handle feed ore as large as 200 mm, normally the product of the primary crusher or the run-of-mine ore, and achieve a product of 0.1 mm in one stage. The particle size distribution of the product depends on the characteristics and structure of the ore being ground. The main mechanism of comminution in AG/SAG mills is considered to be abrasion and impact. Due to the relatively gentle comminution action, fractures in rock composed of strong equidimensional mineral grains in a weaker matrix are principally at the grain or crystal boundaries. Thus the product sizing is predominantly around the region of grain or crystal size. This is generally desirable for subsequent mineral separation as the wanted minerals are liberated with minimal overgrinding, and the grains keep their original prismatic shape more intact. A pilot plant study was made of the liberation characteristics of a nickel sulfide ore with full autogenous and semi-autogenous milling conditions. The mill products were sized and assayed and then analyzed by QEMSEM. Evidence indicated that selective breakage was occurring in both cases, leading to preferential liberation of sulfides (Wright et al., 1991).

Investigations have shown that ores ground autogenously may float faster than if ground with steel media (Forssberg et al., 1993). Grinding with steel media can suppress the floatability of minerals, due to release of iron oxidation products into the slurry (Chapter 12).

In comparison with high aspect ratio mills, SAG mills in the gold mines of South Africa (locally known as “run-of-mine” mills, or ROM mills) are low aspect ratio (largest mills are 4.88 m diameter and up to 12.2 m in length), and are normally operated at high ball charge (up to 35%), high total filling (up to 45%), and high speeds (up to 90% critical). They are often operated in a single stage of grinding to produce final product of 75–80% passing 75 µm from a feed top size around 200 mm, but the mill throughput is relatively low. The initial cost of low aspect ratio AG/SAG mills is less than high aspect ratio mills, but they consume more power per ton of product. The development of the different operating practice in South Africa occurred for historical rather than operational reasons. The ROM mills evolved from pebble tube mills used for fine secondary grinding that were converted to primary mills by directing the full run-of-mine feed to them.

The influence of feed size and hardness on AG or SAG mill operation is more significant than that on rod mill or ball mill operation. In rod or ball mills, the mass of the media accounts for approximately 80% of the total mass of the charge and dominates both the power draw and the grinding performance of the mills (see Section 7.2.2). In SAG mills a significant proportion (or all of it, in AG mills) of the grinding media derives from the feed ore. Any change in the feed size distribution and hardness will therefore result in a change of the breakage characteristics, and the mill charge level will be changed, which affects the mill power draw. As a result, the measured AG/SAG mill power draw often varies widely with time. This is one of the significant differences in operation between the AG/SAG mill and the rod/ball mill, power draw of the latter being relatively stable. In response to the variation in feed size and hardness, the mill feed rate has to be changed. At BHP-Billiton OK Tedi Mine (Papua New Guinea), the ore hardness varied between 5 and 16 kWh t⁻¹, and the throughput to the 9.8 m by 4.3 m (7.5 MW) SAG mill could vary between 700 and 3,000 t h⁻¹ (Sloan et al., 2001). The effects of feed size and hardness on AG/SAG mill operation have been discussed (Bouajila et al., 2001; Hart et al., 2001; Morrell and Valery, 2001).

The AG and SAG mills respond in a different manner to feed size changes. In an AG mill sufficient numbers of large rocks need to be provided to maintain a high enough frequency of breakage collisions. In general, AG mill performance is better with coarser feeds, up to 200 mm. In SAG mills, however, ball charge tends to dominate rock breakage, and the contribution of rock grinding media will decrease (see Section 7.2.2). The coarser feed rocks provide less of a grinding media role and will instead provide a rock burden that requires to be ground. By reducing the feed size in these circumstances, the grinding burden will be reduced (Napier-Munn et al., 1996).

Since the feed size and hardness exert an important influence on AG/SAG mill performance, there are incentives to take account of the mine to mill operation through controlling blasting practices, mining methods, run-of-mine stockpiling, partial or fully secondary crushing, and selective pre-screening of AG/SAG mill feed. The mine-to-mill exercise has resulted in significant benefits to mining companies in terms of improving AG/SAG mill throughput, energy consumption, and grinding product size distribution (Scott et al., 2002).

The effect of feed characteristics on AG/SAG mills also differs from that on high pressure grinding rolls (Chapter 6). HPGRs are close to constant tonnage units: changes in feed characteristics will affect the power draw of the HPGR but tonnage remains fairly consistent compared to AG/SAG mills. This provides for greater circuit stability, and is a contributing reason to the growing interest in using HPGR technology (Lichter, 2014).

Gearless

The largest mills, any requiring over 18 MW of power output, are almost always driven by gearless drives where the rotor of the motor is mounted to an external circumferential flange of the mill and the stationary stator of the motor is mounted securely to a foundation wrapped around the mill flange. Because of the large stator structure, these drives are sometimes called wrap-around or ring motors. Gearless drives operate at the same speed that the mill is turning and require a sophisticated drive control system that includes an inherent variable speed capability.

Synchronous

Widely used in North and South America, the synchronous motor is mounted beside one end of a mill and is connected to the mill via gears. The motor drives a pinion, which is a small geared shaft that connects to the bull gear that wraps around the mill circumference on a flange at one end of the mill. Synchronous motors derive their name from the way that the motor speed synchronizes with the waveform of the electric alternating current (AC) phases. This means the motor operates at a fixed speed in its simplest form. A maximum of two motors can be connected to the mill’s bull gear via separate pinions, called a twin-pinion arrangement, developed for large mills (>5.5 m).

Wound-rotor Induction

Common in Australia and Africa, the wound-rotor motor also uses a gear system to transmit energy to the mill. It operates at a higher motor speed than the synchronous motor and needs an additional mechanical gear-box to reduce the motor shaft speed to the pinion speed needed to drive the mill. A slip energy recovery system is commonly used to permit modest speed adjustments when the motor is operating near its design speed. Wound-rotor induction motors can also be in a twin-pinion arrangement.

It is often desirable to adjust the speed at which a tumbling (primarily AG/SAG) mill operates; in such a circumstance, the mill drive system design must include a mechanism to achieve a variable speed capability (Barratt et al., 1999). Both the synchronous and wound-rotor induction motors can be made fully variable speed if the electric AC waveforms delivered to them by the drive electrical system are modified. A variety of technologies can be used to adapt the AC waveform. Two common variable speed drives for synchronous motors are the cycloconverter and the load commutating inverter (Grandy et al., 2002; von Ow, 2009).

The choice of motor and drive type is dictated first by the motor size (von Ow, 2009):

The electrical efficiency of fixed speed synchronous motors is greater than fixed speed wound-rotor induction motors, but the initial cost is higher. The gearless motor has a higher efficiency than a variable speed synchronous drive, which has a higher efficiency than a fully variable speed wound-rotor induction drive. The initial capital cost, however, and installation time favor the wound-rotor induction motor, followed by the synchronous motor, with the gearless being the most expensive motor with the longest installation time. A financial and plant power grid evaluation is often used to choose between motor options.

Rod and Ball Mills

The specific energy needed for a certain required capacity may be estimated by the use of Bond’s equation, written here as:

$W=10Wi\left(\frac{1}{\sqrt{P_{80}}}-\frac{1}{\sqrt{F_{80}}}\right)E{F}_x$ (7.11)

(7.11)

where W is the specific energy consumption of the mill, kWh t⁻¹; Wi is the work index measured in a laboratory mill, kWh t⁻¹; F₈₀ and P₈₀ are, respectively, the circuit feed and product 80% passing sizes, µm; and EF_x is the product of the Rowland efficiency factors applicable to a grinding stage. The Rowland efficiency factors dependent on the size of mill, size and type of media, type of grinding circuit, etc. (Rowland and Kjos, 1980; Rowland, 1982; King, 2012). The power required (in kW) is then given by W×T, where T is the throughput tonnage (t h⁻¹) (see Example 7.2).

Example 7.2

Using the Bond equation, select the size of ball mill in closed circuit with a cyclone for the following conditions: F₈₀=600 µm, P₈₀=110 µm, Wi=10.5 kW t⁻¹, T=150 t h⁻¹.

Solution

Note that the P₈₀ is the circuit product (i.e., cyclone overflow) and circulating load is assumed to be 250% (Rowlands, 1982). Solving for W, and ignoring efficiency factors (for simplicity) then:

$W=10\times 10.5\times \left(\frac{1}{\sqrt{110}}-\frac{1}{\sqrt{600}}\right)=5.72(\mathrm{kWh}{\mathrm{t}}^{-1})$

Therefore the power draw P is;

$P=5.72\times 150=859(\text{kW})$

The size of mill is then selected from either a calibrated power model (Section 7.2.2) or, more generally, from manufacturer’s data, such as in the table below (Table ex 7.2).

Table ex 7.2

Ball Mill Specifications (overflow discharge, 40% charge filling: see Rowlands, 1982)

Diameter (m)	Length (m)	Ball Size (mm)	Mill Speed (% critical)	Charge Weight (t)	Mill Power (kW)
3.96	3.96	50	71.7	82.8	843
4.12	3.96	64	71.7	89.4	945
4.27	4.27	64	70.7	104	1093

The mill (4.12 m×3.96 m) meets the target power.

AG/SAG Mills

Extensive testing is being conducted for sizing autogenous or semi-autogenous mills because of the lack of equivalent methodology as represented in the Bond approach for rod and ball mills. Pilot scale testing of ore samples, more specifically for an autogenous mill where the grinding medium is also the material to be ground and consequently a variable itself, has become a necessity in assessing the feasibility of AG/SAG milling, predicting the energy requirement, flowsheet, and product size (Rowland, 1987; Mular and Agar, 1989; Mosher and Bigg, 2001).

The use of energy-based or mathematical modeling and simulation can help reduce the cost of the pilot tests by narrowing the choice of processing route at the pre-feasibility stage. This approach includes collection of a large quantity of representative ore samples that are typically drill cores or samples taken throughout the ore body. The rock breakage characterization data are obtained through standard laboratory tests such as the drop-weight impact and tumbling tests (Napier-Munn et al., 1996), SAGDesign tests (Starkey et al., 2006), the MacPherson autogenous mill work index test (MacPherson, 1989; Mosher and Bigg, 2001) and more recently, the Rotary Breakage Tester (Shi et al., 2009). The breakage characteristics of the ore are compared with extensive databases of ore types and plant performance, which provides an indication of the ore’s relative strength and amenability to processing in various circuit configurations. Following pilot scale tests, computer simulation software such as JKSimMet (Napier-Munn et al., 1996) is used to scale-up to the full-size plant. There are other computer software available for plant simulation, such as MODSIM developed by King and co-workers (Ford and King, 1984), USIM-PAC (Evans et al., 1979), and CEET (Comminution Economic Evaluation Tool) (Dobby et al., 2001; Starkey et al., 2001). Even in the absence of pilot data, energy-based models or simulation can still provide reasonably accurate prediction of full-scale plant performance.

The cost of a thorough pilot testing or standard testwork program can be prohibitive, particularly for deposits with highly variable ore types. Instead of collecting large representative samples, a large number of small quantity drill core samples, representatives of the different geological domains of the deposit, are collected throughout the ore body to be characterized by simpler tests such as the SPI test (SAG Power Index) or abbreviated versions of the standard laboratory tests, such the SMC (Morrell, 2004), the RBT Lite (JKTech Brochure) or the SVT (Brissette et al., 2014). These tests are recognized to be geometallurgical tests (Chapter 17). Geometallurgy methodology is increasingly used as a standard in both design and production forecasting of AG/SAG circuits worldwide, giving a much better understanding of the ore response to milling for any deposit.

Power Draw

From a review of existing power models for stirred mills, Radziszewski and Allen (2014) proposed that if all grinding is accomplished via shear then grinding power can be represented by the following:

$P_{\tau }=\mu {\omega }^2{V}_{\tau }$ (7.12)

(7.12)

where P_τ is the shear power (W), µ the viscosity of the mill contents (N.s m⁻²) (for which a relationship was given), ω the angular velocity (rad s⁻¹), and V_τ the “shear volume” (referring to all the shear surface pairs between an impeller and the mill chamber). With appropriate calibration the predicted power matched the measured power for the data of Gao et al. (1996) and Jankovic (1998).

Stress Intensity

This is a measure of the pressure acting on the particles and is fixed during the comminution process as the conditions—grinding media size and density, stirrer speed and slurry density—are normally constant during operation. In the fluidized slurry mills the stress derives mainly from the action of the media, while for lower stirrer speed mills gravity (the weight of the charge) also contributes stress. Expressions for stress intensity have been developed by Kwade et al. (1996) and Jankovic (2001). There is a relationship between stress intensity and product size and an optimum exists that gives maximum size reduction (lowest product P₈₀) and thus is the most energy efficient condition (Jankovic, 2001). As with tumbling mills, the ratio of media size to particle size is of particular significance in optimizing energy consumption (Jankovic, 2003).

Levin Test

This test requires “baseline” data from an operating mill (i.e., feed size, product size, mill tonnage, and total grinding energy consumption). A sample of the baseline reference material is used to calibrate the lab scale mill. The test sample is ground to the same product size as the baseline reference and the number of revolutions required to achieve the target size is recorded. Once calibrated, the energy delivered per revolution is assumed to be the same. Any future material’s specific energy requirements can be estimated by recording the number of revolutions required to achieve a particular target size. The challenge of the method is obtaining high quality baseline data.

Metso Jar Ball Mill Test

This method can be carried out wet or dry in a tumbling grinding mill (8˝ D×10˝ L). Grinding is carried out over several grinding periods (e.g., 10 min and 20 min). Both product size and energy consumption are recorded for each subsequent grinding period. The measured energy is multiplied by 0.65 (VTM factor) and 1.10 (SMD factor) to predict the energy required by a Vertimill® and Stirred Media Detritor, respectively.

Signature Plot Technique

The signature plot takes a 20 kg sample at 50 wt% solids and runs the sample through a test mill. At fixed periods, with a known energy expenditure, a sample is taken to determine the particle size distribution. (See also Chapter 5.) Typically, the method takes either 12 data points or as many until a reduction in size is no longer taking place. This technique has been used for sizing both the Xstrata IsaMill® and the FLSmidth VXP® mill.

Pilot testing of a stirred mill can normally be done with a scale model of the milling technology. Fluidized mills (such as the IsaMill®) have been found to scale up directly to industrial units as long at the test units are 1 L in volume or greater (Karbstein et al., 1996). This is attributed to the fact that the stress intensity responsible for breakage is due to the high speed stirring. Technologies that employ both lower stirring speeds and gravitational stress intensity to induce grinding (such as the Vertimill®) need larger pilot units in order to ensure that the significant gravity stress component is being reproduced.

TowerMill®/Vertimill®

The predecessor of today’s Vertimill® is the TowerMill, which was introduced in 1953 by the Nichitsu Mining Industry Co., Ltd. By 1954 a new company, Nippon Funsaiki, was founded to exclusively produce the Abrasive Toushiki (steeple-like) Crusher. Hereafter it was named “Tower Mill” and in 1965 the Japan Tower Mill Co., Ltd. was established. In 1983, Kubota Ironworks Co. purchased the Japan Tower Mill Co. and supplied the technology as Kubota Tower mills. The latest “owner” of the tower mill technology is Eirich. The Koppers Company, Inc. adopted the new fine grinding technology in the early 1980s, after which the Tower Mill was manufactured by MPSI under a license agreement. In 1991, the license expired and Svedala Industries, Inc. (now Metso Minerals Ltd.) obtained all rights to the technology, except the name, which was changed to Vertimill®.

A schematic of a tower mill is shown in Figure 7.29. Steel balls or pebbles are placed in a vertical grinding chamber in which an internal screw flight provides medium agitation. The feed enters at the top, with mill water, and is reduced in size by attrition and abrasion as it falls, the finely ground particles being carried upward by pumped liquid and overflowing to a classifier. Oversize particles are returned to the bottom of the chamber. An alternative configuration is to feed the mill by the bottom.

Stirred Media Detritor (SMD®)

English China Clays developed an attrition stirred “sand mill” during the 1960s and in 1969 the first production units were installed in a kaolin plant. Currently, ECC, now Imerys, operate more than 200 attrition sand mills in their kaolin and calcium carbonate plants around the world. In 1996, Svedala and ECC signed a license agreement enabling attrition sand mills to be supplied for the Century Zinc Project in Australia. In the following year, this license was expanded, enabling Svedala (now Metso Minerals Ltd) to manufacture and supply the Stirred Media Detritor (SMD®) globally for all applications other than the white pigment industry.

Figure 7.30 shows an SMD®. Normally, natural silica or ceramic is used as the grinding media. Grinding media are added through a pneumatic feed port or the manual feed chute located on top of the mill. Feed slurry enters through a port in the top of the unit.

IsaMill®

Developed by Mount Isa Mines Ltd. (now Glencore Technology) and Netszch-Feinmahltechnik GmbH in the 1990s, the IsaMill® is a large version of the Netszch horizontal stirred mill that was being used for ultrafine grinding applications in various chemical industries. To make this horizontal mill suitable for the mining industry, the engineering challenge was to expand the volume by a factor of 6.

Inside the horizontal shell is a series of rotating discs mounted on a shaft that is coupled to a motor and gearbox (Figure 7.31). The high-speed discs fluidize the media and the slurry that is continuously fed into the feed port. A patented product separator keeps the media inside the mill, allowing only the product to exit. The grinding media can include granulated slag, river sand, ceramic beads or a sized portion of the ore itself.

VXPMill®

Originally developed in the mid-1990s for the fine pigment industry by Deswik Ltd., in 2010 Deswik and Knelson signed a partnership and the name was changed to the Knelson-Deswik mill. In 2012, Knelson became a part of the FLSmidth group and the mill received its present name: the FLSmidth VXPmill®.

HIGMill®

In 2012, Outotec launched a new fine grinding technology for the mineral processing industry. The technology has been utilized for more than 30 years in the calcium carbonate industry, but until recently, was not available for mineral processing.

Media

Proper media selection is key for a successful stirred milling application. A balance needs to be achieved between milling efficiency (kWh per ton fresh feed) versus operating costs (cost of media replacement per ton of fresh feed). Milling efficiencies vary with media size, density, and shape.

Media size selection is dependent normally on the feed top size. The design tries to ensure that the top feed size to the mill is controlled and that the medium is a size no larger than needed to handle this feed. In applications where large variations in the feed top size are expected, energy efficiency will need to be sacrificed in order to select a larger media size.

Media density and shape have varying degrees of effect on grinding (Nesset et al., 2006). Typically, spherical media are considered to be energy efficient, although there is evidence that rod shaped media can be more efficient in ultrafine grinding applications (Tamblyn, 2009). For vertical mills, low media density is preferred as dense media sink and results in grinding inefficiency. In horizontal mills, high density media were found to be more efficient (Xu and Mao, 2011).

Media Loading

The energy drawn by a mill is directly related to the media loading. It is recommended to keep a mill fully loaded during operation. Many mills have automated the addition of media so that when mill power draw drops below a set value, a charge of media is added. It is generally easier to control a mill if media loading is constant leaving control dependent on percent solids and slurry flow rate.

Slurry Percent Solids

Optimization of slurry density can result in gains in energy efficiency. If there is too high a water content, the number of contact events between particles and media drops, resulting in less grinding; if the solids content becomes too great, then fluidization of the mill is lost and the only media movement is that directly caused by the impeller.

Vibratory Mills

These are available for continuous or batch grinding materials to a very fine end product. Two vibrating cylinders charged with ball or rod media are placed either directly above the other or inclined at 30° to perpendicular. Located between the two cylinders is an eccentrically supported weight driven to produce a small circular oscillation of a few millimeters, which fluidizes the mill contents and provides impulses to the material that is ground by attrition. The material, wet or dry, is fed and discharged through flexible bellow-type hoses. Vibratory mills have a small size and low specific energy consumption compared to competing mill types. Three-tube vibratory mill configurations are also available.

Centrifugal Mills

The concept of centrifugal grinding is an old one, and although an 1896 patent describes the process, it has so far not gained full-scale industrial application. The Chamber of Mines of South Africa studied centrifugal milling (Kitschen et al., 1982; Lloyd et al., 1982). Operation of a prototype 1 m diameter, 1 m long, 1 MW mill over an extended period at Western Deep Levels gold mine proved it to be the equivalent of a conventional 4 m×6 m ball mill.

Roller Mills

These mills are often used for the dry grinding of medium soft materials of up to 4–5 mohs hardness. Above this hardness, excessive wear offsets the advantage of lower energy consumption compared with conventional mills.

Table and Roller Mills

These units are used for dry grinding of medium-hard materials such as coal, limestone, phosphate rock, and gypsum. Two or three rollers, operating against coiled springs, grind material which is fed onto the center of a rotating grinding table (Figure 7.32). Ground material spilling over the edge of the table is air-swept into a classifier mounted on the mill casing, coarse particles being returned for further grinding.

Pendulum Roller Mills

Used for dry fine grinding non-metallic minerals such as barytes and limestone, material is reduced by the centrifugal action of suspended rollers running against a stationary grinding ring (Figure 7.33). The rollers are pivoted on a spider support fitted to a gear-driven shaft. Feed material falls onto the mill floor, to be scooped up by ploughs into the “angle of nip” between the rolls and the grinding ring. Ground material is air-swept from the mill into a classifier, oversize material being returned.

Instrumentation

Typical instrumentation is illustrated in Figure 7.40 for a primary (AG/SAG or rod) mill-secondary ball mill/cyclone circuit. Control of the feed rate is essential, and devices such as variable speed apron feeders or feeder belts, in conjunction with weightometers, are often used. Slurry density is measured mainly by nuclear density gauges, and flow rates by magnetic flowmeters or, increasingly, CiDRA’s SONARtrac® technology (Chapter 3). Sump level is commonly indicated by bubble tube, capacitance-type detectors, or other electronic devices.

Newer mills are often equipped with variable speed drives, allowing mill rotational speed to be adjusted as an additional (and important) degree of freedom. Control of the grinding media charge can be made by continuous monitoring of the mill power consumption. A fall in power consumption to a certain level necessitates charging of fresh grinding media. Continuous power monitoring finds its greatest use in autogenous grinding, the power drawn being used as a measure of mill load and a constraint on mill feed rate. AG/SAG mills with crushers in pebble recirculation loops can utilize the crusher operating time or closed side setting (Chapter 6) to control the size of crushed pebbles returning to the mill, which in turn affects mill load and circulating load.

Arguably, the most important parameter required to provide good control of mill performance (especially AG/SAG mills) is the load (or amount of charge) inside the mill. Volumetric load (the percentage of mill internal volume taken up by ore and grinding media) is the key variable as it affects how the charge behaves inside the mill (e.g., shoulder and toe positions, Figure 7.3), but cannot currently be measured online, so it must be inferred from other, measurable parameters. Two categories of instruments are available: “off-shell” stationary sensors and “on-shell” sensors attached to the shell (Tano, 2005). On-shell load monitoring is an advancing field and includes a variety of measurement techniques. The strain experienced by the mill shell or lifter bars/bolts can be measured to determine when the lifter is within the charge volume and thus the position of the toe and shoulder of the charge (Herbst and Gabardi, 1988; Kolacz, 1997; Tano, 2005). Conductivity inside the mill, as measured by two probes mounted on a lifter, can provide a similar measurement of the shape and position of the charge (Moys, 1988; Vermeulen and Schakowski, 1988). The mechanical vibrations caused by grinding can be detected by accelerometers attached to the mill shell or bearings and analyzed to determine charge motion and detect various operational conditions (Campbell et al., 2001, 2006).

Acoustic systems use a single sensor or an array of microphones to detect the sound vibrations produced by charge movement inside the mill and can be mounted close to the mill or on the shell itself. These have been increasingly included in mill control systems in recent years. They can have considerable advantages over load cells (Pax, 2001) due to their ease of maintenance and the additional process information they provide. One system is to use two off-shell microphones, one placed above, and the other placed below the normal level of impact of the charge on the mill shell. As the mill load increases, the point of impact moves toward the upper microphone, and conversely down toward the lower microphone as it decreases. By comparing the output from the two microphones, it becomes possible to determine whether the charge level is rising or falling. This information is correlated with the power draw of the mill and used to calculate the rate of addition of the new feed. It was found that the sound level changed ahead of detectable changes in the mill bearing pressure, power, torque, motor current, and mill weight. Therefore the plant took the sound reading as an indicator of changes in SAG load for milling circuit control. Other systems adjust mill speed in response to acoustic information to control the mill load and ensure maximum grinding efficiency by moving the point of impact to the toe of the charge (Pax et al., 2014). Several commercial systems are available to analyze the sound produced by the mill and detect factors such as ball-on-liner impacts, charge geometry (toe and shoulder position), water holdup or specific information on the behavior of balls and ore, and is used to infer the mill volumetric loading (La Rosa et al., 2008; Pax, 2011, 2012; Pax et al., 2014; Shuen et al., 2014). Sound-based systems tend to be more successful in the long term as a tool to detect steel-on-liner collisions and therefore indicate when the mill should be slowed to protect the liners rather than a real time indicator of load. This is primarily due to calibration drift and the influence of unmeasured acoustic disturbances in the concentrator.

Measurement of mill weight has mainly been by load cells, which directly measure the mass of material in the mill (including ore, grinding media, water, shell, and liners), from which the volumetric load can be inferred. These instruments, however, rely on accurate calibration, do not take into account changes in rod/ball charge or liner wear, require downtime for maintenance and are prone to drift, so alternative measurements are often used. The most common approach is inferring mill weight by monitoring the pressure of lubricating oil flowing through the mill bearings. However, this measure can be susceptible to signal noise and is impacted by the physical properties of the oil (temperature, viscosity) and system hydraulics. Bearing pressure is nonetheless utilized in many AG/SAG mills as an indicator of the load in the mill and is incorporated into control systems. Other systems use the power draw (more specifically the amplitude and frequency of power oscillations) to calculate the position of the charge and determine mill load (Koivistoinen and Miettunen, 1989).

Particle size is obviously a desirable measurement in a grinding circuit. The particle size of the circuit product (cyclone overflow) can be inferred by the use of mathematical models (Chapter 9) but it is preferable to measure directly by the use of on-line monitors, which recent developments are making increasingly robust (Chapter 4). As noted by Flintoff et al. (2014), on-line measurement of liberation would be a notable breakthrough for grinding circuit control.

Mill feed particle size is significant for AG/SAG load management. Monitoring of the feed to the mill by image analysis, for example, provides a measure of the particle size distribution (Chapter 4). Felsher et al. (2015) describe stabilization of SAG power draw by adjusting apron feeder ratios to target higher fines fraction in the SAG feed.

Upset conditions in cyclone operations, such as roping and plugging, have a significant impact on the plant stability and performance, but they often go undetected for extended periods of time due to lack of instrumentation. There are technologies and sensing methods today for monitoring cyclone performance (Chapter 4). When integrated with the plant control system, “offending” cyclones can be automatically switched off so that corrective action is taken within seconds rather than many minutes.

Pumpbox control plays a significant role in stability of the circuit. Many mineral processing plants run their cyclone feed pumpboxes using traditional level control, where the pump speed adjusts to maintain a constant level in the pumpbox. The issue with this philosophy is that a continuously varying flowrate to the cyclopac leads to variability in cyclone pressure and cut-point, which negatively impacts recovery in downstream operations. Pumpbox level management strategy, focused around stabilizing the flowrate to the cyclones, is designed to allow the level in the cyclone feed pumpbox to vary while maintaining the pump speed (Westendorf et al., 2015). The result is more consistent grind size to downstream operations as well as stabilized circulating load within the grinding circuit. Reductions of variability in excess of 50% have been reported.

Control Objectives/strategies

In implementing instrumentation and process control for grinding circuits, the control objective must first be defined, which may be to:

Aspects of grinding circuit control can be introduced using Figure 7.40. Of the variables shown, only the ore feed and the water addition rates can be varied independently, as the other variables depend upon and respond to changes in these items. It is therefore ore feed rate and water addition to the mill and classifier that are the major manipulated variables used to control grinding circuits. There is an important difference in the dynamic response of the circuit to changes in ore feed rate and to changes in water addition rate. Changes in ore feed rate initiate a slow progressive change to the final steady state condition, while changes in classifier water addition initiate an immediate response. Increase in water addition rate also results in a simultaneous increase in circulating load and sump level, confirming the necessity of using a large capacity sump and variable speed pump to maintain effective control.

If the strategy is constant product size at constant feed rate (objective 1), then the only manipulated variable is the classifier water, and the circuit must therefore tolerate variations in the cyclone overflow density and volumetric flow rate when ore characteristics change. Such variations have to be considered with respect to downstream operations (objective 3). Variations in circulating load will also occur. Changes in mill rotational speed provide a fast response in the circuit because the mill discharge rate will be rapidly affected; however, this can lead to instability in the circulating load, hence speed control loops are generally tuned to be slow-acting to ensure stability. Additionally, mill speed has a direct relationship to power draw, meaning that in an overload or high-power situation, increasing the speed of the mill will cause a further increase in power before the mill will begin to unload.

In many applications, the control objective is maximum throughput at constant particle size (objective 2), which allows manipulation of both ore feed rate and classifier water addition rate. Allowing for the fact that the circuit has a limited capacity, this objective can be stated as a fixed product particle size set-point at a circulating load set-point corresponding to a value just below the maximum tonnage constraint. The circulating load is either calculated from measured values, measured directly, or inferred.

Since both ore feed rate and classifier water addition rate can be varied independently, two control strategies are available for objective 2. In the first, product size is controlled by ore feed rate and the circulating load by classifier water addition rate. In the second, product size is controlled by the classifier water addition rate and circulating load by the ore feed rate. The choice of control strategy depends on which of the control loops, the particle size loop or the circulating load loop, is required to respond faster. This depends on many factors, such as the ability of the grinding and concentration circuits to handle variations in flow, the sensitivity of the concentration process to deviations from the optimum particle size, time lags within the grinding circuit, and the number of grinding stages. If the particle size response must be fast, then this loop is controlled by the classifier water, whereas if a fast mill throughput response is more important, then the product size is controlled by the ore feed rate.

It is important to consider the impact of unmeasured disturbances within the grinding circuit, such as ore hardness, liner, lifter, and grate condition, media charge and holdup in the mill. The importance of these parameters makes traditional control difficult, resulting in the need for more advanced techniques.

A common problem in the operation of a milling circuit, particularly when controlling particle size, is that the control actions are often constrained by physical or operational limitations. This problem is particularly relevant in the case of multivariable control, in which control of the process depends critically on the correct combination of several control actions. The saturation of one of these control actions may result in partial loss of control, or even instability of the process. In order to deal with this problem, a method has been developed, in which the multivariable control scheme includes a robust “limit algorithm”, thus retaining effective stabilizing control of the milling process (Hulbert et al., 1990; Craig et al., 1992; Metzner, 1993; Galán et al., 2002). Alternatively, an expert system is incorporated to “manage” the controllers and modify the action based on the situation in the plant.

Expert systems are an accepted form of control for complex grinding circuits. Expert systems are computer programs that compile the process knowledge and rules employed by the control room operators into a heuristic model (Chapter 3). They allow continuous monitoring of the grinding circuit, applying operational best practice consistently. They require little additional investment and infrastructure and can operate from a desktop computer. Several commercial systems are in use.

Model-predictive control has its origins in work done in the 1970s (Lynch, 1977) and gained ground with the advent of more reliable models and more powerful computing platforms (Herbst et al., 1993; Evans, 2001; Galán et al., 2002; Lestage et al., 2002; Muller et al., 2003; Schroder et al., 2003; Yutronic and Toro, 2010). This approach involves development of mathematical models of a unit or a circuit based on existing knowledge, operating mechanisms, and experimental data. Such models often incorporate a number of model parameters that need to be calibrated by experimental data from a specific unit or circuit. Once the model parameters are determined and the ore breakage characteristics are known, the models can predict the performance of the unit or the circuit in response to changes in the feed condition or the operational conditions, either steady-state performance or the dynamic response. The model can also be used in conjunction with an expert system (Felsher et al., 2015) whereby the model predictive controller provides a fundamental foundation for stabilization and the expert system uses heuristics for optimization, providing a framework for the controllers to work within.