Cement Kilns

Size Reduction and Grinding

All sections of this page are still in development.

This article is divided into the following topics:

Size of materials

Plant Layout

In the cement industry, natural raw materials are processed to produce an end product which is a fine powder. As shown in the flow diagram, several size reduction processes are needed. The major raw materials are usually quarried, so that they begin the process as geological strata. The process of quarrying extracts the material in sizes ranging from 1 cm to 1 m. These have to be combined and reduced to a powder or slurry of maximum size about 100 μm before they can be fed to the kiln system. It is not usually possible to achieve this in a single stage. Rawmill systems typically require a feed of maximum size 5-10 cm, so a crushing stage is first required.

In the kiln, the finely-divided rawmix is sintered, resulting in production of hard clinker of typical size 1-50 mm. In order to produce cement, this has to be ground, again to a powder of maximum size about 100 μm.

In addition, other process streams are involved. When the kiln is fired with solid fuel, this will typically be received in a size range 10-50 mm, and must be reduced to below 200 μm. Gypsum and other materials interground with the cement are also received in lump form and may need crushing before use.

Size distribution and measurement

The way in which the size of a particle is defined is always somewhat arbitrary. Most of the mineral particles encountered are irregular in shape - some more irregular than others. Materials such as limestone, clay, gypsum and clinker are quite isotropic in their breakage characteristics, and form ground particles that approximate spheres at least to the extent that, when placed on a three-dimensional grid, their maximum dimensions in the x-, y-, and z-directions are roughly the same. A few other materials, such as micas, produce plate-like particles, and coal has a slight tendency to be plate-like too.

The intuitive way of summarizing the size of roughly spherical particles is to test their ability to pass through, or be retained on, a real or imaginary sieve. The idealized sieve has identical flat square apertures, and the length of the side of the square is specified. If the particles tested on the sieve are perfect spheres, then the diameter of a sphere that just passes the sieve is assumed to be equal to the size of the sieve opening. All these assumptions are questionable, but the sieve concept has the advantage of simplicity and is easily reproducible.

When materials are subjected to a size reduction process, even if the unground particles are completely uniform in size, the final product will always have a range of sizes, because of the random nature of the breakage process. The range of sizes can be expressed graphically as a size distribution curve, in which the mass proportion of each size is plotted against the size.

Cement PSD

In the cement industry the shape of such particle size distributions is commonly modelled using the Rosin Rammler Distribution function:

where Rx is the % retained at size x, xc is a characteristic size, and S is the "slope".

The Rosin Rammler Distribution (RRD) is applicable to materials that have been ground by a process that breaks uniform brittle particles in a random manner, without any agglomeration. Real distributions depart from the RRD, for two main reasons:

  • Grinding any material containing two distinct minerals, such as clinker and gypsum, or limestone and shale, necessarily results in a bimodal distribution.
  • Almost all grinding processes involve some form of size classification, which truncates the otherwise asymptotic distribution.

Despite these reservations, the RRD is an effective method of characterising the grinding processes used in cement manufacture. The RRD summarises a set of particle size data in terms of just two parameters:

  • the "characteristic size" xc, which is the size exceeded by 36.8% (i.e. 100/e) of the mass, and is an expression of the overall fineness
  • the "slope" S, which is an expression of the "narrowness" or "tightness" of the distribution: a low value of S indicates a broad range of particle sizes.

Sieve Analysis

Sieves were used to separate coarse and fine particles in milled materials such as flour from earliest times, so it was natural that the measurement of particle size began with the use of sieves with standardised mesh sizes. Sieves, however, have a number of disadvantages:

  • The manufacture of sieves with precisely uniform mesh size is difficult and expensive, and in use, the uniformity and size of the mesh can change.
  • Undersize material does not simply fall through a sieve: a certain amount of energy has to be supplied, and the material on the sieve has to be agitated. These processes can have a grinding action on the material, and can distort the mesh.
  • Sieves below about 30 μm are extremely fragile, and yet require more energy to separate materials, so sieves below this size are impractical.

However, at least for early cements, the powders were sufficiently coarse that sieve analysis was considered adequate to characterise their fineness, and some statement of the proportion retained on some standard sieve size was always considered necessary when certifying the quality of early cements. But a given value was no guarantee of quality, because all early cements were sieved after grinding, and the quoted fineness value was more an indication of the thoroughness of sieving than of the thoroughness of grinding.

For those familiar with modern cements, it is hard to conceive just how coarse the early cements were. Using data on sieve analysis from the nineteenth century writers, it is possible to describe the evolution in fineness of British cements.

Fineness History

As can be seen, the 1890s saw a critical transition. New technology was taken on to bring this about. In Britain's case, this was only to a limited extent due to the availability of new technology: the change to improved technology that already existed was driven by user pressure. Continental cement, generally much finer and more effectively quality controlled, began to be imported, and had already largely robbed Britain of its export markets.

The first British Standard for Portland cement, published in 1904, specified fineness in terms of maximum amounts retained on 76 mesh (223 μm) and 180 mesh (96 μm) sieves. The dangers of manufacturing cement merely to meet such a specification were emphasised by Butler (1899, pp 121-122), referring to stone grinding in 1898:

The use of sieves in the manufacture of cement is found to be very economical, as they help the stones considerably, but care must be taken that the proper amount of flouring is done by the stones. It is possible, by excessive sieving, to have a cement which may all pass through a certain sieve, and yet contain little or no flour, it being simply cracked till it is just fine enough to pass that sieve. A practical illustration of this fact—familiar no doubt to many who have tried to make cement in a small experimental way—may be obtained by pounding some clinker in a pestle and mortar with frequent sifting, and comparing the powder thus produced with that of the same clinker emanating from millstones. It will be found that the mill-stone-ground material contains a much larger proportion of flour than that ground in the mortar, and gives infinitely better results when tested for strength in the ordinary way. There is no doubt that this is just where the many kinds of edge-runner mills now largely used for grinding cement, fail to come up to the old-fashioned millstones, viz. in the flouring of the cement. They depend too much on their sieves: the clinker is passed under the runners and all cracked slightly, and then passed up to the sieves; any that is cracked fine enough to pass the sieves is conveyed to the warehouse, the remainder being returned to the mills to be cracked again; so the process continues; the material often passing under the rollers ten to fifteen times altogether before the reduction is completed. So long as the cement will pass the sieves, the promoters of these mills fondly imagine they have done their duty, and lose sight of the fact that it is the impalpable powder or flour which is the essential part of the cement.

Suffice it to say that sieve data were insufficient to describe the fineness of cement as it relates to cement strength. In the case of rawmix or fuel grinding, the production of excessively fine material is detrimental, and sieve analyses are adequate for control of their fineness (Note 1). The concept of "flour" in cement (i.e. material below about 20 μm: see Note 2) understood by Butler was quantified after 1900 by elutriators, which separated out the fines by flushing a sample with an upward current of air or a liquid, the size range extracted being related to the sweeping velocity. However, elutriators were only ever research instruments and unsuitable for routine use, and no universally-recognised standard was ever developed.

The specification of maximum levels of sieve oversize in Standards occurred during the period when grinding technology was developing rapidly, and the levels specified reflected more the current technology than any desirable fineness. Successive British standards included the following maxima:

Size19041907191519251931*1958
96 μm 22.5181410
223 μm3311
90 μm10--
212 μm1--
*The standard sieves were redefined in 1931. From 1958, due to use of specific surface, sieve data were no longer relevant.

Cement specifications clearly needed a way of prescribing the fineness of cement, and the sieve method was written into standards in the absence of any more useful measure, despite the fact that it provided little guidance as to the likely quality of the cement, particularly as the number of different grinding technologies multiplied. The need for a fineness measure relatable to cement properties such as strength and setting time became ever more urgent until finally a breakthrough occurred in 1939, with the development of the measurement of specific surface by air permeability.

Specific Surface

Specific surface is a measure of the total surface area of the particles in a given mass of cement: SI unit m2/kg. Assuming spherical particles of cement, the mass of a particle of diameter D μm is M = 3150 × πD3/6 × 10-18 kg (for particle density 3150 kg/m3), and the surface area is A = πD2 × 10-12 m2. Thus the specific surface of the particle is A/M, or 1905/D, and the smaller the particle, the higher the specific surface. The specific surface of a cement sample is therefore largely influenced by the mass of small particles present, in contrast to sieve analysis which focuses only on the largest particles. The specific surface concept also has the advantage that, because reaction with water occurs at the particle surfaces, it accurately correlates with the rate of reaction of the cement.

The air permeability method was first published by Lea and Nurse in 1939. The method consists of measuring the equilibrium rate of flow of air through a bed of cement compacted to a standard porosity, when a measured pressure differential is maintained across it. The mathematical treatment was based on the Carman Equation (published 1938), from which it followed that, for a constant particle density, air viscosity and bed porosity, the specific surface is proportional to √P/V where P is the pressure differential and V is the air flow-rate. There are no empirical constants in the equation, so no "standard sample" is required for calibration.

During the following ten years, a number of variants of the Lea and Nurse apparatus were developed, with a view to simplifying operation for rapid use in hour-to-hour quality control, the original Lea and Nurse method being somewhat laborious. These included the Rigden and Blaine methods, and rather than employing equilibrium flow conditions, they compared a constantly-changing flowrate with a constantly-changing applied pressure. These methods were fairly rapid for the coarser cements of fifty years ago, but had the distinct disadvantage that a primary standard cement was needed for calibration, with an assigned value traceable to the Lea and Nurse method. An automated Lea and Nurse apparatus remains the best, providing virtually instantaneous results.

Adoption of specific surface for control of cement grinding began during WWII, and a specific surface minimum of 225 m2/kg was introduced into the British Standard in 1947, and remained unchanged until 1989, when it was raised to 275 m2/kg.

Crushers

Crushers are used to reduce material in lump form to a size sufficiently small to be fed to a mill for fine grinding. The cement industry processes requiring fine grinding are rawmix preparation, fuel preparation and grinding of clinker to make cement. All three may require a preliminary crushing stage.

In the case of raw materials, attempts have always been made to minimise the amount of crushing needed by extracting the rock in the finest possible state. In the chalk areas, chalk was until well into the twentieth century quarried by hand, and by a process termed "milling", chalk was scratched from the quarry face producing material that was mostly 30 mm or smaller. With the use of explosives in hard rock quarries, the amount and disposition of the charges is designed to maximise the amount of small material produced. However, crushing is invariably employed to further reduce the rock. Crushers are usually more energy-efficient than the finer-grinding equipment.

Coal, as used by the cement industry, rarely needs crushing. Clinker, as produced by early static kilns, emerged as large lumps, and crushing was always required before grinding could take place. Rotary kilns by contrast produce clinker that is typically below 50 mm and does not need crushing. However, every rotary kiln occasionally produces larger clinker, or may contain large slabs of kiln coating material, so some facility for occasional crushing is required.

There are a number of different crusher types in use in the industry:

Manual Crushing

Roll Crushers

Jaw Crushers

Gyratory Crushers

Impact crushers

Manual Crushing

Cement PSD Women breaking limestone at a lime works. This pen and wash drawing is often stated to show Joseph Aspdin's town centre cement works at Kirkgate, Wakefield. It clearly does not.

During the first half of the nineteenth century, mechanical crushing was not cheaply available. Stone had been crushed for applications such as Macadam road metal exclusively by hand, using the cheapest available labour, such as convicts and workhouse inmates. On smaller cement plants, clinker was crushed by hand throughout the century. In chalk areas, chalk was quarried by pick and bar on a piece-rate basis, pay being obtained only for undersize material. Even the ultra-modern West Thurrock plant began operations in 1912 with hand-quarried chalk.

Roll Crushers

Roll Crusher

Roll crushers are the oldest of the types listed here, having been introduced for ore crushing in the Cornish mining industry as early as 1804. Smooth-faced rolls were driven by steam engine - a technology not generally available elsewhere. Later in the century, rolls with interlocking teeth were used for light crushing of chalk and marl, and for coal. Towards the end of the twentieth century, the process underwent a revival with the invention (Schonert, Germany, 1982) of the flood-fed high pressure roll. This, instead of generating grit, took clinker-sized material and compressed it into a highly-fractured compact, which, when fed to a conventional mill, dramatically reduced the remaining grinding to be done, and thus reduced overall energy consumption.

Newell Roll Crusher

Newell roll crusher installed at Aberthaw in 1914 for coal crushing. The design had toothed rolls of 24 inch diameter and 16 inch length. As with all crushers, the drive includes a flywheel so that there is sufficient angular momentum to crack occasional large hard pieces. There is also a substantial spring-loaded frame applying horizontal load to the roll bearings, so that the rolls can jump apart if an ungrindable (e.g. iron) object is encountered. In the absence of this, tramp metal would smash up the hard alloy roll surfaces.

Jaw Crushers

Jaw Crusher

Jaw crushers have their origin with the Blake Crusher (Eli Witney Blake, USA, 1858) which became the most common crusher type in the cement industry, and many others. A number of variants with slightly different jaw motion were introduced in competition with the Blake design, but the latter remains in common use today, with a design little changed from the original.

Newell Jaw Crusher

Newell jaw crusher installed at Aberthaw in 1914: the design is the standard Blake double toggle machine. It had inlet dimensions 36 inch by 22 inch and was used for crushing Blue Lias limestone. It could thus accommodate stone of about 400 mm down. Again, there are substantial flywheels to provide momentum. Early Blake's crushers could have cast iron jaws, but with the arrival of Bessemer steels, manganese steel castings were used for the jaws, providing a compromise between strength and hardness. The reciprocating movement of the jaw is typically only a few millimetres. Material is cracked on the "in" stroke and the product falls further into the jaw on the "out" stroke.

Gyratory Crushers

Gyratory Crusher

Gyratory crushers were a logical development of jaw crushers that emerged after an initial patent (Philetus Gates, USA, 1881), but were first used in the British cement industry in the first decade of the 20th century. The design allowed a larger inlet for very large stone, less liable to block. Unlike other crushers, the gyratory crusher could in theory be flood-fed - i.e. filled to the brim and allowed to empty out gradually. This made it useful for stone being brought by rail, a whole truckload being tipped at a time without need for a feeder.

Hadfield Gyratory Crusher

Hadfield gyratory crusher similar to that installed at Southam in 1908. The Southam machine had a diameter of 5 ft, crushing 50 t/hr, but much larger crushers were already available, and by 1913, Wilmington had installed a 10 ft machine crushing 300 t/hr. The central "dolly" does not rotate, but its axis is moved at the base in a circular path, so that the crushing zone rotates, with an small amplitude analogous to that in the jaw crusher. The design resists blockage because material can move sideways as well as downwards.

Impact Crushers

Impact Crusher

Impact crushers differ from the other crushers in that they do not work by squeezing a captured lump - instead a relatively high-speed hammer strikes the lump in mid-air and bounces it against a fixed anvil surface. Breakage occurs both with the hammer blow and with the bounce. The chamber is typically designed so that the hammers hit lumps several times against different anvil surfaces. The effect can be extended by adding a second hammer wheel.

Impact Crusher

Mills

Mills are used to reduce materials of maximum size around 30 mm to fine powder. Different types of mill used in the industry include:

Flat Stones

Roller Mills

Tumbling Mills

Separators

In the early history of fine grinding, most of the technology was developed for milling cereals, which are soft. Powered machines for flour milling developed directly from the techniques for hand grinding and so the flat quern gave rise to the flat stone mill, while roller mills developed from the hand roller and the pestle and mortar. The relatively minor processes for grinding harder materials, such as the production of paints and inks, tended to use the same sort of equipment, often using wet grinding.

The limitations of what was still essentially a Stone Age process for fine grinding of hard materials led to the development of the continuous tumbling mill, and the cement industry led the way with its development, although today the method is used in many other heavy processing industries. As with the development of kilns, the main thrust of development 1875-1975 consisted in find ways of applying ever-increasing amounts of cheap energy to the process, and tumbling mills, although extremely inefficient in terms of energy conversion, allowed unlimited scale-up. As increasingly fine cement grinding was required, it became more difficult to dissipate the enormous amount of waste heat generated during grinding in tumbling mills.

The somewhat earlier development of new crushing and grinding techniques in Germany and the USA consisted of moving from the use of "natural" abrasives such as millstones to custom designed steel alloys, and the development of these alloys - particularly those with manganese and chromium - only began with the introduction of the Bessemer process in the 1870s, producing a cheap pure mild steel base that could be modified at will.

The ancient observation that hard materials were easier to grind in a liquid medium led to the use of "wet process" grinding of raw materials in the cement industry, and grinding by this technique was always much more energy-efficient than dry grinding, although the energy required to remove the water from rawmix was always vastly greater than that saved during grinding. Flat stones were easily modified for wet grinding, but roller mills were exclusively used for dry grinding. Tumbling mills, when developed, were also easily adapted for wet grinding.

From 1975, concern to improve energy efficiency led to a reversion to the use of roller mills, which are considerably more efficient. Roller mills are now almost exclusively used for dry process raw milling on new installations, and the latest finish mills also employ roller mills, in combination with high efficiency separators.

Evolution of types of mill used for finish milling
Finish mill types

Flat Stones

Flat Stones

Flat stones were used for both raw and finish grinding in the Portland cement industry from the outset. In raw grinding they were used to produce both wet and dry rawmixes. The industry inherited them as a completely mature technology and continued using "standard" equipment, operated by craftsman millers recruited from the flour industry. This accounts for the almost complete lack of any innovation during the half-century in which the industry used them. The centuries-old technique amassed a massive glossary of arcane terminology which the old text-books usually say are "too familiar to require explanation".

stone mill

The standard mill had a diameter of 54 inches, and the hundreds of mills installed rarely varied from these dimensions. Mills would usually be arranged in sets of four or five, driven from a common shaft. The mill shown (from Butler, p 112) is for finish grinding, and has a rotating upper ("runner") stone 20" thick, with a 20" diameter feed hole ("eye"). The runner rotated at 100-150 rpm. The large eye and 3-4 inch "swallow" (compared with flour milling) were required in order to allow material of up to 2" to be fed and drawn between the stones. Runners for raw milling were 12" thick because less pressure was needed for the softer material. For finish grinding, extra crushing force was sometimes obtained by cementing a heavy iron plate to the top of the runner. The lower stone ("bedstone") was 12" thick. The stones were made from a hard rock, and white "French burr" was almost variably used, because its low wear-rate more than compensated for the high cost of the stone. French burr is an Oligocene chert mainly found in Seine-et-Marne, and is obtained in 10-30 cm pieces: these have to be shaped and assembled into a wheel by cementing with Keane's or Portland cement. The pieces were arranged with their bedding planes perpendicular to the grinding face, so that the surface remained uniform as it wore down.

stone dressing

The figure shows some of the stone dressing terminology. The stones were both "dressed" by pecking at the surface with picks and chisels, firstly to obtain exact planarity of the "skirt" part, and then to cut the furrows in a characteristic profile, with a steep "back" on the leading edge of the furrow and a gentle "feather" in the trailing edge. Furrows were usually cut to about ½" depth, getting shallower towards the outside, and 1½-2" wide. The furrows pulled the feed into the mill, and had a scissoring action, but the main grinding effect was due to crushing and shearing between the feather edges of the furrows.

The dress of the stones gradually wore away in use, and had to be re-cut. In rawmilling of chalk, re-dressing might be needed every 300 operating hours, but much more often when grinding clinker. As late as 1900, commentators such as Butler and Spackman still maintained that millstones produced the best cement, but admitted that the maintenance cost was very high. Referring to cement grinding, Butler (p 115) says:

The dressing of the stones is a very expensive item, both as regards labour, and the consequent wear and tear of the stones, to say nothing of the enforced idleness and unproductiveness of a pair of stones when up for dressing. As a rule, a pair of stones will not run more than thirty-six hours, this period of course depending upon the nature of the clinker under treatment, and the fineness to which it has to be reduced. This means that, with a mill containing four pairs of stones, one pair is always up for dressing. The proper dressing of a pair of stones, and looking after the feed and lubrication of those that are running, constitutes a fair day's work for the miller and his mate.

An excellent video showing the tedious stone dressing process can be found on YouTube. This puts the productivity of flat stone milling of cement into context: two men, one of whom is a relatively high-paid craftsman, supervise a set of four mills, three of which are running, for a twelve hour shift. Each running mill would make 1.3-1.6 t/h, so they might typically make 52 tonnes in a shift - a labour usage rate for finish milling of 0.46 man-hours per tonne. A large plant with 35 kilns, making 1000 t of clinker per week, would need 13 pairs for daywork operation. Bevans, making 1700 t/week in 1897, used flat stones exclusively, and had five six-pair sets, allowing six day single-shift operation even at busy times of the year. It can be seen why the Thames-side manufacturers, who used flat stones almost exclusively until the turn of the century, regarded calls for finer cement with trepidation. The power drawn by each pair was around 28 kW (so consuming about 19 kWh/t to make a cement with a specific surface of around 230 m2/kg). This was sufficient power to obtain a product at around 70°C, and it was clear that the process was inefficient. This led manufacturers to experiment with alternative methods of grinding, even if quality was compromised. It is interesting to speculate that the drawn-out attachment to the use of flat stones in the UK was associated with the high belite content of the nineteenth century clinkers. Belite crystals have a tendency to plastic deformation, rendering them tough to grind by impact processes, whereas the shearing action of flat stones probably tore them apart relatively effectively.

Waldringfield mill

The picture shows the original four-pair finish mill installation at the Waldringfield plant, supplied by the Pulsometer Engineering Company, as shown in The Engineer (19/10/1883, 56, p 303). It was rated at 50 tons per day, so the standard day was one 12-hour day shift.

When used for cement milling, a certain amount of coarse material ("nibs") always escaped grinding, especially when the dress of the stones was worn. Because of this, cement was always put through some sort of separator. The traditional method was a rotary sieve of about 100 μm opening, with brushes to encourage the material through, much as used in flour mills. Maintenance of these sieves was always a high-cost item, and they were of no use for finer grades of cement. Towards the end of the century, air-separators began to be used, and with the oversize returned to the mill inlet, flat stones were used in closed-circuit. By far the most common separator was the Askham Air Separator.

The emerging need for fine grinding led to either closed-circuit operation, or the addition of a tube mill. The latter allowed the flat stones to be operated at a coarse setting, allowing high throughput and a much reduced frequency of re-dressing. Towards the end of the 19th century, a few variants of the flat-stone mill were developed, using improved abrasive materials and higher speeds. One form that developed was the disc mill, with the axis of rotation horizontal. These had hard surfaces spring-loaded into close contact. The most-used development was the Cormorant mill - a horizontal mill with 24" conical grinding elements made of cement-bonded rock emery (corundum), and driven at high speed. These, although expensive, avoided the dressing costs of conventional mills, and were used at Swanscombe and other JBW plants 1895-1901.

Flat stones remained in use both for raw and finish grinding on many of the larger plants until around 1910, and on small plants until 1930. The use of flat stones as the final, fine-grinding stage for both applications was displaced by tube mills, particularly after steel fine-grinding media became available.

Roller Mills

Roller Mill

Roller mills are as old a technology as flat stones. Early forms were often called edge-runners and consisted of stone wheels running on edge on a circular table. It is clear that the action of such mills was quite different, lacking the shearing effect of flat stones, and relying for their grinding effect almost entirely on the crushing effect of the weight of the wheels. They had the advantage that, whereas the size of flat stones was limited by the problems associated with suspending the runner, edge-runners could be scaled up more or less indefinitely.

In the cement industry, early use was mainly as a pre-crushing stage prior to use of flat stones. However, a number of technical improvements made them more attractive as a stand-alone grinding process. Iron grinding surfaces allowed higher pressures, and steam power allowed higher speed machines in which the weight of the rollers could be supplemented by centrifugal force.

The elementary roller mill has a number of defects in comparison with flat stones: the material does not naturally progress from inlet to outlet, and flowable materials tend to move out of the path of the approaching roller in a bow-wave movement. The bed of material in the grinding path has to be prepared so that it can be captured in the nip of the roller. In practice, scrapers and brushes of various designs were added to the rotating mechanism to place the feed in the correct position and to flush out the fines. The problem of bed preparation explains the fact that 20th century developments began with the grinding of damp materials - first coal, then dry process raw materials - which form a cohesive crushable mass on the grinding path.

Two distinct forms of roller mill emerged:

  • the ancient form, in which the common axle of the rollers is rotated, causing the rollers to rotate around a static grinding path
  • the form used exclusively in modern mills, in which the rollers are on individual fixed axles projecting in from outside the mill, and the grinding path takes the form of a rotating table.
Dutrulle & SolomonDutrulle & Solomon Edge Runner 1886

Dutrulle and Solomon were a London firm. Their mill appeared in 1886, and was used, with varying degrees of success, for cement fine grinding. The design produces a 3-stage grinding effect by successive concentric raceways, the first with a 4 foot roller weighing 3.8 t, the second with a 4'6" roller weighing 4.8 t, and the third with two 5 foot rollers, each weighing 5.9 tonnes. The rollers were rotated by a common shaft, which also actuated brushes and scrapers to move the material. The product was discharged into an elevator which took it to a rotary sieve. Allegedly, the recirculating load could be 2400%. The rollers had replaceable hard alloy steel tyres. The rollers, compared with the 1.7 tonne weight of a flat-stone runner, were cumbersome to remove for maintenance. A distinct disadvantage of this design was that, because of the different sized rollers, the drive shaft was subject to out-of-balance forces.

Neates GrinderNeate's Dynamic Grinder: picture from Engineering, 31/1/1891, 51, p 131

The Neate's Dynamic Grinder appeared in 1891, first installed as the sole cement grinding plant at Percy J. Neate's Borstal Manor plant. The mill had four rollers, each 5 ft in diameter and weighing 3 tonnes. The rollers were set at 30° to the vertical and had a 1 ft wide replaceable steel tyre. The inclination of the rollers produced a slight extra centrifugal pressure, but was mainly in order to improve stability at full speed. Scrapers detached material from the steel grinding path and tipped into elevator pits on either side of the mill. These lifted it to a pair of rotary screens, from which the rejects dropped back into the mill. Mills were said to make around 4 t/h at a fineness typical of 1890, and drew around 80 kW, so the energy consumption of 20 kWh/t was about two-thirds of typical flat-stone cement. However, critics suggested that the cement, ground closed-circuit to a coarse residue specification, contained correspondingly less "flour" than flat-stone cement, and so gave poor strength performance.

Neate went on to produce a more compact, higher speed mill that was small enough to fit in the mounting of a standard flat-stone mill. These mills became quite widespread, although at most locations they were used only as pre-grinders for flat stones.

Various other edge-runner designs were produced, but none achieved more than experimental use at single locations, until a number of innovations arrived from the USA from the 1890s onwards.

Griffin MillGriffin Mill

Norman Finish MillsGriffin mills in use for cement grinding at Norman in 1908

By far the most important US import was the Griffin Mill. This was a pendulum mill, developed in 1886. The vertical drive shaft enters at the top of the mill and is connected through a universal joint to the pendulum shaft, the "bob" of which is the cylindrical roller. The static grinding path is a vertical cylindrical surface surrounding the mill chamber. On turning the drive shaft, the roller swings outwards and makes contact with the grinding path, whereupon, by friction, it begins to rotate around the mill in the opposite direction to that of the drive. This counter-rotation produces a significant amount of shear to material in the grinding path, in addition to the crushing action. In common with other centrifugal mills, increasing the speed also increases the centrifugal effect and therefore increases the grinding pressure.

The mill had a built-in sieve separator: fan blades on the rotating shaft above the grinding bob lifted the fines and blew them through sieve plates placed above the grinding path.

The mill came into use in the 1890s, initially for re-grinding rejects, both in dry rawmix and cement grinding. Later, they came to be used for the entire grinding process. Martin Earles first installed them for autonomous cement grinding in 1899, and in 1904, they were used exclusively at Norman for rawmilling, coal and cement grinding. As the mills "ran cold", for raw milling, the material had to be completely pre-dried. A slightly more equivocal commitment to these mills occurred at Ellesmere Port in 1912. The twenty years or so during which they were used were a period in which the market was "upping the ante" on cement fineness, and although in 1900, Griffin mills could easily meet industry fineness requirements, by 1920 they had been left far behind by ball-and-tube mills. An interesting case is the Saxon and Norman plants: the 1901 Saxon plant was installed with Schneider Kilns and ball-and-tube mills, while the "more modern" Norman (1904) had rotary kilns and Griffin mills. However, the Griffin mills were abandoned from 1912 and the Norman clinker was taken to Saxon for grinding. With the fading of dry process, the only remaining duty for Griffin Mills was in coal grinding, for which their effect was perfectly adequate, although their open discharge must have made for a rather explosive atmosphere.

Raymond MillRaymond Mill

Raymond Bowl Mill. This was a variant of the pendulum mill that lasted much later in the century. Multiple pendulum rollers were suspended from a central rotating spider, and bore on the sides of an air-swept bowl. The mill was often used for drying/grinding coal, and was occasionally encountered in Britain.

Centrifugal Ball-Race Mills

Tiger MillAskham's Tiger Mill

Fuller Lehigh MillFuller Lehigh Mill

Although the grinding elements are steel balls, these are essentially roller mills, grinding by pressure. They are similar to Griffin mills in that grinding pressure can be increased by increasing the speed.

The Askham Tiger Mill dates from the early 1890s, and demonstrates the fact that, if operated sufficiently fast, centrifugal mills need not be arranged on a vertical axis. The Tiger mill consisted of a horizontal cylinder with a central shaft, fed at one end and discharging at the other. "Pusher" paddles on the shaft rotate four steel balls around the circumference. The mill was commonly supplied in combination with an Askham Separator. They were regarded as suitable for dry raw materials and soft clinker. At Saxon, they was used for pre-grinding rawmix prior to finishing in Griffin mills.

The Fuller-Lehigh Mill arrived around 1905, and was a much more substantial vertical mill, in direct competition with Griffin Mills. Its use for cement grinding came and went in much the same way, but it formed the basis of a new generation of coal mills that was much longer lasting. Here again, "pushers" on the central shaft propel four steel balls around the circumferential grinding path. The largest mills, with a 57 inch diameter grinding ring, had 17 inch balls weighing 330 kg each. A key feature is that the mill for the first time has a chamber above the grinding zone that functions as an air separator with a separately controlled fan. Variations on this theme remained in use for coal milling and later raw milling ever since.

Vertical Pressure Ball-Race Mills

E-MillEssential design of ball-race mills with vertical pressure and a spinning table

During the 1920s, Fuller-Lehigh mills became restricted to coal milling, although this became a major market, because firing with pulverised coal - a technology invented by the cement industry - was taken on by the electric power stations that were proliferating at that time. Desirable characteristics for a coal mill were:

  • increased output
  • a sealed device that would not leak dust
  • elimination of oversize particles
  • the ability to simultaneously dry and grind the coal

Claudius Peters obtained rights to manufacture the Fuller-Lehigh mill, and subsequently modified it incorporating a number of features that became characteristic of all roller mills:

  • they kept and developed the air separator above the grinding zone
  • they abandoned the centrifugal principle, and instead pressed the balls down on to a horizontal race, using a spring-loaded upper race, which could be further loaded by external spring devices
  • they built the bottom race on to a spinning table
  • material passing under the rollers were centrifugally ejected from the edge of the table into an air-stream blown from below, and lifted by this into the separator
  • the latter air-stream could be hot, so producing a drying effect, and hot rejects from the separator fell back to the centre of the table mixing with the fresh coal feed, to provide a relatively dry layer under the grinding balls.

This was the E-Mill, and mills on the same principle continue to be built. Babcock & Wilcox, which had an interest in Claudius Peters, and were mainly engaged in power generators, produced their own version of the E-Mill, and many of these were used for coal milling in the cement industry.

Modern Roller Mills

Modern Roller MillEssential design of modern roller mills

The successful key design features of the Claudius Peters E-mill were also applied to roller mills almost contemporaneously by Loesche (1925). Many other suppliers and designs followed, all with the common features:

  • the rollers are on fixed axes and bear down upon a rotating table, with the pressure applied by external adjustable tensioners
  • the table is surrounded by a ring of ports which supply air to lift the ground material to the top of the mill: the air may be hot, if required
  • in the top of the mill a separator sorts the fine and coarse particles: the fines leave the mill suspended in air, and the coarse rejects fall back into the mill

As with the ball-race mills, the roller mills developed initially for coal milling, because coal is relatively soft, fineness requirements are not difficult, and the low density of coal means that it is relatively easy to entrain in a slow air-stream.

Different makes and applications of mills were reflected in some minor variations:

  • Rollers might be flat-faced on a flat table, or round-faced on an indented table
  • Rollers might have horizontal axes, or be tilted
  • Separators might be static, or 1st, 2nd or 3rd-generation dynamic designs.

The application of roller mills to raw milling and cement milling took longer. In these applications, roller mills had to compete with ball mills, which were able to meet fineness and output requirements with ease. However, ball mills always had specific energy consumptions nearly double those of roller mills. As long as energy remained cheap, there was little incentive to improve roller technology to deal with the more difficult materials. So serious developments occurred in Britain and the USA after the 1973 energy crisis, but occurred in relatively energy-poor countries such as Germany and Japan much earlier. A roller mill was grinding rawmix in Germany as early as 1937, but the first in Britain were the two Berz mills installed at Cauldon in 1957. These were removed in 1963 and the experiment was not repeated for a long time, subsequent dry process rawmills all being ball mills until in 1977 roller mills were installed on Ketton Kiln 7 and Platin Kiln 2. From here on, all dry process rawmills were roller mills, with the exception of Derrylin in 1989.

The modern roller rawmill is heated by the exhaust gas from the kiln's preheater, and the mill product and kiln exhaust dust are captured in a common filter system. The rawmill is increasingly to be considered as an integral heat-exchange element of the kiln system. Roller rawmills can be supplied to suit any foreseeable kiln size, by increasing the size of the table. A large number of modern mill specifications shows on average table diameter (m) D = 178 P0.45 where P is the table power in kW, with a 7% error margin.

Roller Mill PowerPower requirement for cement grinding by roller mills and ball mills

The remaining milestone for roller mills was their application to cement grinding. The technical problems were the difficulty of obtaining a stable table bed that would present itself to the rollers, high separator bypass causing excessive recycle, and insufficient production of ultra-fines - the latter being the fatal flaw of the roller mills used a century before, as discussed above. However, these technical problems have been solved, and all roller mill suppliers now offer versions designed for grinding clinker and slag. These allow uniformly lower energy consumptions as compared with ball mills - typically 50% lower on the main drive for standard cement grades. The power requirement for the ancillary equipment (separator, pumps, conveyors, fans) is greater than that for a ball mill, mainly because of the fan power needed to recirculate the dense clinker particles, but the overall power of the system is still typically 30% lower than that for a ball mill system. As with rawmilling, very large outputs can be achieved with mills with increasing table diameters. A large number of modern mill specifications shows on average table diameter (m) D = 187 P0.41 where P is the table power in kW, with a 14% error margin. The power consumption depends on product fineness, and is shown here in comparison with ball mills. The first plant in which all grinding was by roller mills was installed in Vietnam in 1999. However, it was not until 2009 that a roller mill for cement was installed in Britain and Ireland - in fact at Platin. The absence of any other projects would appear to indicate that electricity prices still remain acceptably low in Britain.

Tumbling Mills

Tumbling Mill

A tumbling mill consists of a rotating chamber - usually cylindrical - containing grinding media which may be balls, rods, or the material itself. The rotating action causes the media to cascade, producing a grinding effect by both percussion and attrition.

Tumbling mills have dominated grinding in the British cement industry during the period of study (1895 to date) of this website. They also typify the guiding principle of the industry's history as discussed regarding kilns - that, provided that energy is cheap and emissions are of low priority, there has been little incentive to progress beyond a state of complacent reliance on low technology.

When first introduced in the industry in the 1890s, they easily solved an increasingly worrying problem - how to greatly increase cement fineness while also increasing output. Tumbling mills afforded a technology which changed little in the ensuing century, and which could be scaled up almost indefinitely.

Tumbling Mill Theory

eccentric load

In a tumbling mill, the fraction of the internal volume occupied by the media (in bulk) is referred to as the volume load, usually expressed as a percentage (%VL). In the mill at rest (Fig. 1), containing a typical 30% volume load, the centre of gravity of the mass of media is below the central axis of the mill. On turning the mill, the media are rotated without movement until the surface of the media is at its angle of repose (Fig. 2). Because this raises the centre of gravity of the charge, energy must be supplied. On continuing to turn the mill, the media on the sloping surface begin to slide (Fig. 3). As the mill continues to turn, media are lifted around the periphery to the top of the charge, replacing the media sliding to the bottom. Material fed into the mill, and occupying the voids between the media, are subjected to attrition between the sliding layers of charge. If the speed of the mill is increased, then the media at the top edge of the charge acquire sufficient velocity to become airborn (Fig. 4), and then describe a ballistic trajectory across the mill until they land at the bottom ("toe") of the charge. Material in this zone is subjected to percussion, the crushing effect being the result of the absorption of kinetic energy.

The rotational speed of a mill is selected to achieve the desired blend of attrition and percussion. As speed increases, the airborn media are lifted higher, until a speed is reached where the centrifugal effect exactly balances the effect of gravity upon the contents of the mill, and the mill then "flywheels" with the media distributed around the circumference. This is the "critical speed" and is given by NC (rpm) = 42.29/√D where D is the diameter (m) inside the lining (Note 3). Grinding action ceases altogether at this speed so mills are usually driven more slowly - typically 60-85% of the critical speed. The precise media trajectory achieved at a given percentage of critical speed depends upon the form of the mill's lining. Mills often contain "lifter" linings that lift the media higher, while a smooth lining may transfer less kinetic energy due to slippage of the charge.

The power required to rotate the mill is a complex function of its dimensions, the nature of the lining, and the volume load, density and nature of the media. However, for the comparatively narrow range of operating conditions used in cement plant mills, a simplified equation is often used:

P = 0.2846 D A W N (Note 4)

where P (kW) is the power drawn (excluding drivetrain and motor losses); D is the diameter inside lining in metres; A = 1.073 - α (the fractional volume loading); W = mass of charge in tonnes; N = rotation speed in rpm. Power can be calculated for individual chambers or for the entire mill. The actual energy consumption of the mill includes also frictional losses associated with mill bearings and drive gearbox, and electrical losses associated with non-optimal operation of the motor. However, with good practice, these can constitute less than 5% of the overall energy.

All grinding processes are very inefficient, because the actual energy difference between ground and unground materials at constant temperature is very small - only a few J/kg - whereas the energy expended in grinding operations is considerable:

materialmill typekWh/t
cementball33
cementroller20
rawmix (dry)ball16
rawmix (dry)roller9
rawmix (wet)washmill4

1 kWh/t = 3.6 kJ/kg. Virtually all the energy put into these processes ends up as heat. In the case of cement made by ball mill, the dissipation of all this heat becomes a major technical problem in itself, the energy being sufficient to raise the product temperature by 150°C.

In grinding cement, tumbling mills are most efficient in the early stages of grinding, and the energy requirement rapidly increases with the fineness of the product.

grindability
The "standard" curve for well-tuned ball mills was derived for cement typical of the mid-20th century: modern clinkers tend to be harder, and the energy cost for modern cements is generally higher than indicated here.

The form of the curve shows demonstrates the utility of air separators for making finer cements. The mill is operated to deliver SSA around 300 m2/kg, in the more efficient zone of the curve: the separator extracts the finer fractions and returns the rest to the mill.

Early Tumbling Mills

As with other modern crushing and grinding processes, tumbling mills for use in the cement industry could not develop until a wide variety of engineering-grade alloy steels became available. However, tumbling mills of sorts date from perhaps a century earlier. The first were batch mills, and had their earliest use in the ceramics and pigment industries. They consisted of iron drums one-third filled with - usually - flint beach pebbles, and lined with ceramic or squared flint ("silex") blocks. Flint pebbles have the advantage that they are naturally smooth and rounded, but have the hardness of quartz, and so can be used for grinding most minerals. The material to be ground was added in quantity sufficient to fill the voids between the pebbles, often with added water. The drum was usually rotated on rollers. After rotation, often for many hours or days, the closure of the mill was removed, the whole contents were emptied out, and the pebbles were separated out with a sieve. Some designs instead used a sieve grating across the door, and were emptied by rotating with the door off. Such mills are still used for small-scale grinding processes.

For industries such as cement, there was a need for continuous grinding processes, such as were already available from flat stones, etc. So it was necessary to be able to put unground materials into the mill continuously, and to withdraw ground product at the same time. Because the mill was to run non-stop, the structure and grinding elements had to be strong and wear-resistant.

Ball Mills

Discussion of early tumbling mills is complicated by the fact that the terminology has changed. A Ball Mill was originally a short (compared with its diameter) mill with media consisting of steel balls, used for coarse grinding. A Tube Mill was originally a relatively long mill with media consisting of small pebbles, used for fine grinding. Later, mills that combined these two functions were called Combination Mills. However, since 1930, all mills have been combination mills, and are now called ball mills or tube mills, the terms being more-or-less interchangeable.

Krupp Grusonwerk ball millEarly Krupp Grusonwerk ball mill

The first continuous tumbling mill to find application in the cement industry was patented in 1876 (Gebr. Sachsenberg, Rosslau, Sachsen-Anhalt) and after redesign by Jenisch and Lohnert, was manufactured by Grusonwerk (Dessau) in 1885. These were certainly in use in Britain by 1890. The mill was built around a drive shaft carried on bearings at either end. Between two end-plates, the drum was built up from steel castings in a slatted arrangement. The length of the drum was around 50-55% of the internal diameter. Steel balls at about 25% volume loading were the grinding media. The mill was run at around 50% of critical speed, since cascading of the charge was not desirable because of the central shaft, and to avoid breaking the peripheral castings. Feed entered the mill through slots around the shaft on the inlet end-plate, and fine material ran out through perforations in the slatted castings. Surrounding this was a network of sieves that allowed the fines through, while coarser particles were carried up and dropped back into the mill. The whole mill was surrounded with a dust enclosure to catch the fines, which in practice were thrown out in all directions. The mill as described would draw power P = 2.3 D3.5 kW where D is the diameter inside the lining in metres, so a 2 m mill would draw 26 kW at the shaft and might grind 2-3 t/hr on coarse grinding duty. After Grusonwerk was acquired by Krupp in 1893, it was known as the Krupp Grusonwerk Ball Mill. The working principle was used by other manufacturers and mills of this type were still being produced up to WWI. Later designs attached half-shafts to the two mill end plates, avoiding having the shaft pass right through the mill.

As has been discussed, mills that rely upon sieves for their fineness control may be suitable for rawmix or coal grinding, but for cement, particles leave the mill as soon as they are just fine enough to pass the sieve, and the ultra-fines needed for cement strength are not produced in quantity. For this reason, such ball mills were only used as a pre-grind stage, followed by - for example - flat stones. In 1904, Krupp produced a self-contained cement mill using a ball mill and their version of the Askham separator in closed circuit, but I think this was never used in Britain. However, from the mid 1890s, ball mills were used in combination with tube mills (see below) and this setup became the normal cement milling installation until 1920.

Early Kominor millearly kominor

Although many others made peripheral-discharge ball mills, the main competitor in Britain for the Krupp Grusonwerk mill was the FLS Kominor. This was produced specifically as a pre-grinder for the Davidsen tube mill, but was often used on its own for such duties as coal milling. The design abandoned the barrel assembled from slats, and introduced the familiar modern format, with a solid cylindrical mild steel shell, lined with with stepped cast steel plates. Discharge was through peripheral slots adjacent to the end casting. The perforations in the Grusonwerk mill plates were problematic because the percussion of the charge would cause the holes to close up and choke the mill, and the Kominor end-discharge, although having a smaller area, avoided this defect. Scoops gathered up the material retained on the sieves and returned it to the mill inlet. The original mills were 1 m long by 1.5 m diameter, with a 1.15 t charge, and probably drew about 7 kW.

later kominorLater Kominor design with cylindrical sieve mountings ("Fastax") for extra area and accessability

The later mills increased in size up to 1.5 m long by 2.4 m diameter, with a 5 t charge, and could draw up to 50 kW. The "Fastax" sieve modification was mainly designed to allow fast and independent replacement of both the sieves and the liner plates, emphasising the fact that on all ball mills, these items both wore very quickly. These mills are described in detail in Redgrave's 1922 edition. The Kominor was provided by FLS as part of their package along with kilns, etc., and appeared in many large installations before WWI.

Krupp 1907 ball millKrupp ball mill with trunnion discharge (1907) as installed at Southam

Ideally, a mill should both be fed at one end, and discharge at the other end, along the centre line, so that grinding takes place progressively along the length of the mill. However, lifting the fines back to the mill centre line was a difficult problem, and the peripheral discharge design evaded the problem. The obvious disadvantage is that much of the feed could pass out of the mill almost immediately. A transitional design appeared in 1907, when Krupp designed a pregrind mill with steel ball charge, from which the product was lifted and discharged through the fine-end trunnion. The enlarged outlet trunnion had a tyre running on rollers, thus minimising the size of the plain bearings, and easing the problem of dealing with thermal expansion. Until the mid 1920s, many mills were designed along these lines, with the inlet end supported by a fixed plain bearing on a small trunnion, and the outlet end carried on rollers, either on the outlet trunnion or on the mill body.

Tube Mills

Davidsen tube mill Davidsen Tube Mill

The Tube Mill, used for finish grinding of cement, but also for rawmix and coal, was first patented in 1891 (Meyer Joseph Davidsen, a Dane living in Paris) and subsequently manufactured in 1893 by FLS. The mill had many features that have continued into modern designs. The tube was of welded mild steel plate, of length 3-5 times the diameter. The ends were heavy steel castings, each with a central hollow trunnion. The mill was supported by lubricated white metal bearings on the trunnions. The feed passed into the mill through the inlet-end trunnion. Because the early trunnions were of small diameter (150 mm or less externally), the feed had to be screwed in, and this also effectively limited the feed to material that was already less than 1 mm. As with the ball mill, the original designs had the product leave the mill peripherally, but this time through small ports close to the end plate, into a dust hood. By 1906, lifters and larger trunnions allowed the product to leave through the trunnion. The mill cylinder had a smooth lining around 40-50 mm thick, which could be made of steel, but in the early versions was usually silex (shaped flint blocks, usually sourced in France or Belgium) and the grinding media were flint beach pebbles of 25-50 mm diameter and 30-40% volume load. The mill was rotated with a spur wheel attached to one of the end plates, and was run at 75-85% of critical speed, which resulted in not much cascading, because of slippage of the charge on the smooth lining. The combination of flint media and silex lining was particularly advantageous for wet grinding of rawmix, since it resisted corrosion, and a market for tube mills for finish grinding slurry established itself quickly.

In cement grinding, the tube mill established itself as a complement to the ball mill, and they were installed in pairs, with the ball mill discharging directly into the tube mill inlet. They often worked on a common drive. Tube mills were also used in combination with flat stones, the latter being run "light" so that stone wear was minimised, and greatly increasing their output. During the 1890s, in response to the need for finer cement demanded by the export market, many tube mills were added to the flat-stone mills that were universal in the Thames-side plants.

Krupp Tube Mill Krupp tube mill with trunnion discharge (1907) as installed at Southam

An obvious disadvantage of the pebble media was its low density - a third that of steel - so that the power draw was proportionately low, and mills could be upgraded (provided that the bearings could take the weight) by use of steel media. But steel balls of the size required - say 40 mm down to 15 mm - were not cheaply available until WWI, and the early mills were not built to cope with the greater static and dynamic loads associated with denser media, so once again, technical progress had to wait until more robust designs were available. The reliance on pebble media during this period helped preserve the distinction between "ball mills" and "tube mills". A further, somewhat questionable justification for installing a two-stage grinding process was the ability to inject water in the second stage, a process that was thought by some to be necessary with rotary kiln clinker. A mill of the early design would draw power P = 2.9 L D2.5 kW, so an early small mill 4.8 m long and 1.15 m inside the lining drew 19 kW and might make around 1 t/h. In 1907, Krupp produced a mill with trunnion discharge, designed to operate with its trunnion discharge ball mill. As with the ball mill, the outlet end was supported on rollers on the trunnion. The mill could be used with steel or flint media.

Wilmington Rawmills Tube mills for wet rawmilling at Wilmington

Aberthaw Finish Aberthaw finish mills

Similar large FLS installations took place in 1912-1914 at West Thurrock and Wilmington. The raw milling differed, because West Thurrock used soft chalk, whereas the Wilmington chalk was hard. The soft chalk was ground with added clay slurry by washmills. The hard Hessle chalk was ground with added slurry in ball mills. In both cases, the coarse slurry was then passed through Trix separators, the rejects being returned to the first-stage mill. The fines were reground in open circuit tube mills - loaded with flint media - with peripheral discharge at the end, as shown. The West Thurrock installation required only one such mill. Apart from this, the grinding facilities were the same: the coal was ground in a single Kominor mill followed by a flint-loaded tube mill, and finish grinding was by two sets of Kominor and tube mill.

By 1914, ball-and-tube mill sets had developed to the point where the ball mill was considerably lengthened, and both mills ran on two trunnion bearings of increased diameter, although in the Aberthaw installation shown still used flint media in the tube mill.

Combination Mills

Solomill The Polysius Solomill

Newells Chamber Newell's chamber grinding mill

By 1910, most new finish mill installations consisted of a ball mill and tube mill, with the ball mill product going directly to the tube mill inlet. In order for the material to flow between mills by gravity, it was necessary to mount the ball mill on a higher level, necessitating high, heavy-duty piers. It was obvious that the installation could be made simpler and cheaper if the two mills could be combined into a single unit. This gave rise to the combination mill just before WWI. The first attempt was probably the Polysius Solomill, and Newells subsequently produced a mill along the same lines. The design reverted to the idea of the peripheral discharge ball mill, and the first, coarse-grinding chamber discharged through sieves into a surrounding collection space. The grit was then lifted from here into the central feed port of the fine grinding chamber. In the Newells mill at least, the fine chamber used flint media. The Newells mill ran on a small inlet trunnion bearing and rollers on the shell towards the outlet end. The Solomill had shell-mounted tyres and rollers at both ends.

Modern ball mills

mill schematic 1930s four-chamber ball mill for cement

From 1920, the transitional "combination" mill morphed into the simple multi-chamber design that was standard for the ensuing eighty years. A mill now consisted of a single cylindrical steel shell, closed at both ends by cast steel end-walls, from each of which projected a concentric trunnion. The mill was supported on bearings on the trunnions. The size of the trunnions had increased to the extent that normal-sized crushed stone or clinker could be fed to the mill. Large particles need large media, with high momentum, to crush them. On the other hand, to grind small particles, the large surface area provided by small media was needed to provide the necessary attrition. For grinding large particles down to finished product in a single mill, it is necessary to use successively smaller media, which must be prevented from mixing with each other. The simplest solution was to divide the cylinder into a number of chambers with perforated diaphragms that allowed the ground material through, while keeping the media sizes separate.

The basic technology of these mills was well established by the mid 1930s, and remains more-or-less unchanged today. During 80 years, there has been a steady increase - ten-fold overall - in capacity, and this has been accompanied by improvements in the alloys used in the wearing parts and in the details of the bearings and drive.

Typical mills in the 1930s had drive power in the range 300-1000 kW, while typical modern mills have ten times these ratings, and there is no obvious upper limit to future developments, although the poor energy efficiency of the design means that the cement industry will not see larger mills.

The development of all-steel media allowed maximum grinding effect in these mills, but the rate of wear of these media was very high, often around 1 kg per tonne of material ground. Although hard alloys were available, they tended to be brittle, and hard media would break into irregular pieces, causing a rapid reduction in grinding efficiency. It was preferable to have media that wore fast but maintained their shape, and chilled white iron was most commonly used for media and liner plates. Periodic sorting of the media with screens allowed the more worn media to be used in the next chamber of the mill, with most of the "top-up" media being added to the first chamber. The development of alloys that were both hard and tough - mainly chromium steels - began in the 1950s, but these alloys were expensive, and it was not until the early 1970s that their price was reduced sufficiently to eliminate the use of softer steel. On modern, large mills, the use of chromium steel alloys for media and liner plates is indispensible.

The original mills were turned by means of a spur ring meshing with a pinion. This was connected to an electric motor, usually through a gearbox. Because conditions were often very dusty in mill houses (or wet in wet-mills) the drive shaft was usually run through a wall, with the gearbox and motor in a separate, "clean" room. From the 1950s, it became common to dispense with the spur-ring, and drive the mill directly through a concentric shaft attached to the outlet end trunnion. This became less viable with larger modern mills, but an alternative drive philosophy for these that is occasionally encountered is the "wrap-around" induction motor delivering torque directly to the mill shell, so that drive bearings and gearboxes with their frictional losses are no longer required.

The classic bearings for these mills consisted of cylindrical "low-friction metal" surfaces, separated by a film of oil. Oil is continually pumped into the gap, and control systems automatically stop the mill if the oil pressure drops or the temperature of bearing elements rises due to friction. As the size of mills increases, the pressure needed to lift the mill on its trunnions increases beyond the viscous capacities of lubricants, and in order to provide more bearing area, modern large mills dispense with trunnions, and run instead on pads bearing on a slip-ring on the mill shell itself.

A significant development in mill design occurred with the introduction of the classifying liner plate. These plates provide a conical surface sloping towards the inlet, which bounces cascading balls towards the inlet. This effect is greatest for the largest media, and as a result, the larger media congregate towards the inlet, while the smaller madia are displaced towards the outlet. This potentially allows the removal of diaphragms. Classifying liners are usually thicker than conventional plates, so that the mill diameter is reduced, with a potential reduction in mill capacity only slightly offset by the increased effective length of the mill provided by removing the diaphragms. However, the mill is simplified, and draught through the mill is freed up, so modern mills generally have only two chambers: the first with a lifter lining, and a much longer second with a classifying lining. In a few instances, single chamber mills have been designed, with classifying lining throughout.

Ball mills for raw milling

For raw milling and coal milling, it is not necessary to produce ultra-fines, so two-chamber mills were most common. The development of the multi-chamber mill corresponded with the abandonment of dry process in Britain, so raw mills were used to grind slurry. In hard-rock areas making slurry with ball-mills, the raw- and finish-mills were very similar, and were often installed side-by-side.

Although rawmix components are generally softer than clinker minerals, any quartz content makes the mix very abrasive, and wear is exacerbated in wet milling conditions. In the case of linings for wet mills, the wear problem was solved by use of rubber liner plates, which are not corroded and resist abrasion. Grinding continues to occur primarily by abrasion between the media.

In the case of dry process rawmilling, early plants preceded the raw mill with separate driers. With the arrival of the new dry process in the 1950s, it was desirable to combine the drying and grinding processes. It was also necessary run the mill in closed circuit, using a separator. Where ball mills were used in Britain, four different process designs were tried:

  • double rotator mills
  • semi-autogenous mills
  • drier + single chamber mills
  • air-swept mills (including coal mills)

On modern installations, ball mills have been entirely supplanted by more efficient roller mills.

Double rotator mills

Double RotatorDouble rotator grinding circuit

The Double Rotator® was the design supplied by Polysius, and used at Cauldon, Weardale, Aberthaw, South Ferriby, Cookstown and Rugby. Although in theory hot kiln exhaust gas could be used to heat the mill, only the Aberthaw mill did this. The Rugby mill uses cooler exhaust gas, and all the others used separate furnaces to generate hot drying air. The design has the advantage that both first- and second-chamber product pass through the separator, so that softer components that grind finely in the first chamber are removed and avoid over-grinding, while the second chamber can be graded to deal with the harder components only. At Cauldon, Weardale and Cookstown, the mill was preceded by an Aerofall mill and performed only regrind duty.

Semi-autogenous mills

Aerofall7 m Aerofall mill being installed at Weardale

Semi-autogenous mills are used for grinding coarsely-crushed damp ores, and the grinding action is obtained in part by the rock itself ("autogenous grinding"). The few installations in Britain (Cauldon, Dunbar, Weardale, Cookstown) have all been Aerofall Mills. These were all the same size: 7 m diameter by 2.2 m long, with a 1350 kW motor, grinding 150-200 t/h of raw material. They ran on wide inlet and outlet trunnions, allowing a 1.3 m diameter feed space, and were fed with material 250 mm down. About 15% volume load of 130-150 mm media were used, and the mills maintained about 30% full. The mills were air-swept, requiring about 70 m3/s to carry out the product. The mills were followed by three-stage separators, which separated 30% of the product as of rawmix fineness, the remainder going to regrind mills.

Mills with drying and grinding chamber

The FLS dry rawmill design had a drying chamber, followed by (usually) a single grinding chamber with a classifying lining, operated in closed circuit. Only two of these were installed in Britain and Ireland, at Padeswood and Platin. Although it was feasible to heat these with kiln exhaust gas, in practice hot air was supplied from auxilliary furnaces.

Air-swept mills

The first raw ball-mills to be designed specifically to be heated by kiln exhaust gas were those designed in the early 1950s by KHD as part of their suspension preheater kiln process, and were often known as Humboldt mills. Various other suppliers subsequently produced similar designs. In Britain and Ireland, the only installations were at Plymstock, Hope, Limerick and Derrylin. The mill is swept with hot gases from the kiln exhaust, suitably diluted with ambient air to avoid overheating the bearings. The large gas mass-flow is capable of lifting all the product in suspension. The rawmix-laden gas is then led into the kiln's gas filtration system: the solid product is passed through a separator, and any oversize is returned to the mill inlet. The system has the advantage that the sweeping gas need not be dust-free: the dust that leaves the preheater - typically 10% on clinker - is drawn through the mill, and is cooled and wetted by the grinding process, making subsequent filtration much more efficient. It also ensures that the dust is well homogenised with the rawmix before return to the kiln. However, rawmills are always designed with excess capacity to allow for maintenance time and for periods of reduced output, so for times when the rawmill is stopped, a parallel gas-cooling system must be provided.

When fuel was ground by ball-mill single-chamber air-swept mills were used from the 1920s, operated in indirect or semi-indirect mode. For 40 years, most FLS installations had "Tirax" ball mills and indirect firing. Edgar Allen had a similar arrangement. These mills usually removed the requirement for a separate coal drier: with hot sweeping air, the fuel could be dried and ground simultaneously. The hot air was usually extracted from the kiln hood, but was sometimes provided by a separate furnace. The hot air supply temperature was controlled by bleeding in ambient air so that the coal/air mixture leaving the mill was below the 80-100°C upper limit to avoid explosive conditions. All such installations were further protected with explosion doors.

Ball mills for cement milling

The early multi-chamber mills had typically 3 or 4 chambers. A typical four-chamber mill might have:

chamberlength, mball size, mm% VLmass, tDAWN kW
13.8990, 80, 70, 6029.223.12253
22.2950, 40, 3034.115.75161
33.8925, 22, 1927.022.35259
42.841626.716.00186
total12.9077.22859

As discussed above, it became common to install classifying linings to the outlet chambers from the 1960s, and new mills commonly had only two chambers.

Closed circuit mills (with separators) were rarely installed, from the 1930s, but became common from the late 1960s, and recently-installed mills are exclusively closed-circuit.

Evolution of types of tumbling mill used for finish milling
ball mill types

During the period of development from 1930 to date, cement mills have increased in size and drive power, in response to an increase in fineness of cement, an increase in the hardness of clinker, and a desire to operate a smaller number of grinding units.

Mean installed power of tumbling mills used for finish milling
ball mill power

Separators

Closed CircuitClosed-circuit grinding

Most grinding processes work best if the fine material is removed as soon as it is formed, since the fines have a "cushioning" effect, preventing the grinding surfaces from making good contact with the coarser particles. In processes such as cement milling, where the particle size has to be reduced about a thousand-fold, the product-sized particles produced in the early part of the process impede the grinding of the rest, and will tend to end up finer than they need to be. With "cushioned" grinding, the amount of extra surface area produced per unit of grinding energy is reduced, although the product continues to be heated up by the energy expended.

In the case of rawmix grinding, there is no need to produce ultra-fine particles at all, and over-grinding can be deleterious. Once again, it is best to remove particles from the milling system as soon as they are below the "critical size" - 125 μm for calcite and 45 μm for quartz.

Closed-circuit grinding is used to remove fines from the milling system, so that the mill can function with a feed that is optimally coarse for its grinding action.

In early grinding, both for cement and rawmix, sieves were originally used as the "separator". In the case of rawmix, if coarse quartz is avoided in the raw material, the critical cut-size is around 100 μm, and sieves have remained fairly effective. In the case of cement, an extremely coarse product was initially thought acceptable, but successively finer sieves came to be used. As discussed above, sieves are progressively more expensive and less effective at smaller particle sizes. The use of grinding equipment other than flat-stones also necessitated more sieving. The approaching problem is described by Butler (1899, p 129), referring to a closed-circuit edge-runner mill:

. . . for every 2 tons of material that went forward to the warehouse no less than 50 tons passed over the elevators, indicating that the material passed under the edge-runner as many as twenty-five times altogether; . . . the renewals of the gauze of the sieves under these conditions cost no less than 1½d (2015 £0.70) per ton of cement ground, which is not surprising under the circumstances.

An alternative method was needed. Air separators met this need, and remain essential equipment today.

Separators for Wet Rawmilling

Stag ScreenerAllen Slurry Separator from The Engineer, 145, 27/1/1928, p 94

Thick slurry, of the consistency of paint, will pass a fine sieve only slowly, or not at all, under gravity. Various designs were developed to increase the pressure on the sieve. In the example shown, installed at Dunstable in 1926, paddles ("beaters") of 56" (1.42 m) diameter rotate at 140 rpm, throwing slurry with a centrifugal action of 15.6 g against sieve plates with 500 μm circular holes. Sieves as fine as 150 μm were sometimes tried, but their fragility in this punishing environment is obvious. Rawmixes such as that at Dunstable had the advantage that a marl close to kiln feed chemistry was being used. Typical performance was actually something like:

>μm%
6000.8
3002.5
1507.5
9015.0
7518.5

Such a psd would not work well with Thames raw materials, where the large particles would be mainly flint fragments. The grit from this screener was mainly fossiline pieces of marl, still quite grindable, and so the grit was returned to the washmill. At locations with harder grit such as flint, e.g. British Standard, the grit was passed through a regrind tube mill before returning to the washmill.

FLS Trix Separator

A particularly common variant of the centrifugal sieve was the FLS Trix separator, used on most FLS wet process plants. Shown here is a format used in the 1920s. First used in combination with Kominor wet ball mills in the 1890s, these remained in use as late as 2005 on the wet process line at Padeswood.

DSM screen120° DSM screen

An improvement on the sieve principle was the DSM (Dutch State Mines) screen or sieve bend. The sieve has tapered slots across a curved surface, and the slurry is applied tangentially under moderate pressure. The tangential flow of the slurry means that the effective screen size is lower (by the sine of the glancing angle) than the actual slot size. Thus the screen can be robust and less liable to blockage. Screens with gaps down to 150 μm are available and these reject most particles over 75 μm. The DSM screen continued in use until the wet processes ceased operation. The remaining Kensworth chalk plant uses vibrating screens.

HydrocyclonesHydrocyclones

Separators for Dry Grinding

Askham SeparatorAskham Air Separator, based on the patent drawing

For dry grinding, impractical sieving was replaced by air separators. The Askham Air Separator was patented (Mumford and Moodie, UK) in 1885. The principle is the ancient process of winnowing out fine particles by use of a draught of air. To obtain good results, the material must be well dispersed in the air stream before separation, and the fines must be efficiently recovered from the air stream. The Askham Separator achieved this using an arrangement that remained in use by all air separators for the next 90 years.

The manufacturers were Askham Brothers & Wilson, Ltd, of Sheffield - a steel producer that also specialised in grinding equipment for a wide variety of industries. The separator was originally designed to work with flat stones, and subsequently with roller mills. It has a rotor comprising, on a common shaft, an air circulating fan and a spinning material distribution plate. The fan induces a circulation of air between the inner and outer annuli of the machine. The coarsely-ground mixture is dropped on to the distribution plate, the rotation of which throws it out in a thin layer into the rising air stream. Particles that are too large to be entrained drop into the inner cone and are recycled to the mill. The entrained fines pass with the air stream into the outer annulus, where most of the solids drop out, partly due to reduced velocity, and partly because of the centrifugal effect as the air-stream re-enters the inner annulus. Fineness is adjusted by varying the rotor speed or by means of a sliding damper.

The separator began to be used at many larger plants during the 1890s as the need for finer cements developed. However, the introduction of tube mills during the same period provided an alternative means of achieving high fineness, and air separators disappeared once again for half a century.

The Askham company was taken over by Edgar Allen in 1903, and Askham brands such as "Stag" and "Tiger" were subsequently used by Allen. Allen continued to produce air separators, but few were sold in Britain.

Newells 1914 design (based on that of Pfeiffer), with longer recirculating air path.

The Askham design principle was adopted by other equipment suppliers, mainly in the USA and Germany. Gebrüder Pfeiffer obtained the German rights for the Mumford & Moodie patent in 1887, and produced separators from 1888. Before 1900 a modified Pfeiffer separator was being used in combination with a Krupp ball mill to produce finished cement. In the USA, the Sturtevant version of the design became dominant. All the developments varied the original design by considerably lengthening the airflow path in the outer annulus and by more sophisticated dampers, both designed to improve the removal of the fines from the recirculating air. All these first-generation air separators had the common problem that, although oversized material was efficiently removed from the product, the tailings contained a substantial proportion of completely ground material. The result of this high "bypass" rate was that closed circuit grinding systems had high recirculating loads, and had to be designed to cope with these high mass-flows. Design limitations on the recirculating load meant that there remained limitations on the fineness that could be achieved, both in terms of increased specific surface, and in reduction of coarse residues. Subsequent changes in design aimed to rectify this.

Tromp curve

Development of air separators

Separator performance is usually described in terms of the Tromp function, which is the probability that particles of a given size will end up in the rejects. An ideal separator puts all particles above a critical size into the rejects, and all those under that size into the product. Real world separators deviate from the ideal in that the separation is not sharp, and because a proportion of the fines always end up in the rejects - defined as "bypass". The main reason for bypass is that the fines remain attached to the coarse particles as agglomerates. Ultrafines tend to adhere electrostatically to larger particles. Early air separators could have bypass as much as 50%. Developments of design were therefore designed more completely disperse the feed to reduce bypass, to reduce the cut size, and to increase the sharpness of cut. The Tromp curves show the changes achieved in first, second and third generation separators.

2nd generation separator principle: cyclone separator

3rd generation separator principle: spinning cage separator

The great advantage of the first-generation separators was their simple, "one-pot" design, but inevitably the separation of the fines from the recirculating air was inefficient, largely relying on gravity. The second-generation design was introduced in 1960 and involved entraining the material in air in a similar manner, then increasing the separation g-forces by ducting the air through exterior cyclones.

Third generation designs began in 1979 with the Onoda O'Sepa, and other designs followed. The disagglomeration and dispersion of the feed was greatly improved by drawing it under fan power through a spinning cage which provides opposing centrifugal and impact forces. The fines are then retrieved in external cyclones.

The improved designs used progressively more power - up to 10% of the mill system total - but this could be more than offset by increased mill throughput. Further advantages were more possible adjustments to fine-tune the product size distribution, and improved cooling of the mill system.

Although commonly used elsewhere, and although in Britain their benefits were well understood, separators ceased to be used for cement milling for a long time mainly because of their capital cost: although the separator itself was not expensive, a closed-circuit mill building was much more elaborate than that for open circuit milling - much higher, and incorporating large elevators and conveyors rated for moving the high recirculating loads. The gradual reintroduction only took place from the late 1960s onwards, as fineness requirements increased, and larger mills came to be used, which are susceptible to over-heating when used in open circuit.

Cement fineness increased through the 1970s and 1980s, leading to a 30% increase in grinding energy, and this prompted conversion, in some cases to more efficient systems with later-generation separators.

Meanwhile, for raw milling, the return of the dry process in the 1950s and 1960s necessitated various separator types. In the case of roller mills, separators were built into the mill casing, with rejects returned direct to the grinding path. Combinations of third generation separators with roller mills finally made these a viable proposition for cement milling around 2000. This provides the possibility of reducing power usage in finish milling by 30% or more, but by 2015, only one such mill had been installed, at Platin.

Notes

Note 1. As discussed on the rawmix preparation page, large particles are deleterious, rendering the mix difficult to burn. Erling Fundal of FLS (1979) showed that the critical sizes were 45 μm for quartz and 125 μm for calcite. These parameters can be measured by sieve. There is no good reason to grind any finer than this, and rawmix with one component (usually calcite) ground much finer than the others will tend to segregate in unpredictable ways in a suspension preheater. In the case of solid fuel, the critical size is around 150 μm.

Note 2. Cement particles react with water at their surface. The outer layers of a cement particle react, and are replaced with a somewhat thicker layer of hydrate. This is exemplified by the progress of reaction of a spherical particle of 14 μm diameter:

hydration reaction

The cement reacts with about 25% of its mass (or 79% of its volume) of water. The degree of completion of hydration is found to correlate closely with strength development. Many years of modern experience show that the rate of hydration (and corresponding increase in strength) diminishes with time. The first 1 μm layer is reacted in about 40 hours, the second in about 80 hours, and so on. This is understandable because, in order to reach unreacted material, water has to penetrate increasingly thick and impermeable layers of hydrate. This observation leads to the familiar "7-micron" rule: after the outer 7 μm layer of the particle has hydrated, the rate of reaction becomes negligible. This stage is reached in about 2 months, and in most concrete, wet curing has already ceased by then. This means that particles with a diameter over 14 μm always leave an unreacted core. In a simple spherical particle model, the mass proportion unreacted can be calculated as (1 - 14/D)3, as shown here:

proportion unreacted

Using this function in combination with a particle size analysis, it is possible to calculate the overall amount of un-reactable material present. Early British cements, up to 1885, left 50-60% of their mass unreacted. Good modern cements, made with high-efficiency separators, leave less than 5% unreacted - the objective is, as far as possible, to make full use of the clinker, given its high energy and environmental costs.

Note 3. A particle of negligible size and mass m just remains in contact with the mill lining at the "12 o'clock" position if the centrifugal effect (mv2/r) is equal to the gravitational force (mg), where v is the peripheral velocity in m/s and r is the radius inside the lining in metres. The circumference of the mill is 2πr metres so v = 2πrN/60 where N is the speed in rpm. Substituting v in the first equation we have 4π2r2N2/3600r = g. This rearranges to N = 30/π√(g/r) or 30/π√(2g/D). Using standard gravity of 9.806 m/s2, the constant reduces to 42.29.

Note 4. The multiplier 0.2846 is an empirical value. It is a conversion of an original 0.1 D W N with units of horsepower, feet and Imperial tons, originating in the 1920s. The more modern formula gives the same result at an odd volume load of 22.7%, although the new multiplier may simply be the result of miscalculation. The number of significant figures, here as elsewhere, should not be construed as an indication of precision.

Note 5.

Note 6.

Note 7.

Note 8.

All sections of this page are still in development.

© Dylan Moore 2015: commenced 01/09/2015: last edit 23/06/2017.