Although this page is often accessed as a stand-alone piece, it is part of a work on the history of the British and Irish cement industries, and where statements are made about the historical development of techniques, these usually refer only to developments in Britain.
The raw material preparation stage of cement manufacture results in the production of a rawmix that is in a suitable state for feeding to the kiln in which it is converted by heat into clinker. This is a chemical transformation. The rawmix consists of a mixture of materials that will react together to form the calcium silicates that confer on the clinker its strength-giving properties. The mineral particles in the rawmix usually consist principally of calcium carbonate from the limestone component of the mix, and aluminosilicates from the clay or shale components, together with a certain amount of quartz (silicon dioxide). In order to get these to react together effectively, a number of rules must be followed - rules that have been understood from the earliest times in the development of the industry:
The chemical reactions in the kiln largely take place at the surface of the mineral particles, so to make these reactions take place quickly and at low temperature, the mix must be very finely ground.
The different solid mineral particles must approach very close to one another for reaction to take place, and the composition of the mix must be correct, not only en masse but also on the microscopic scale – in other words the mix must be thoroughly homogenised.
The properties of the clinker are extremely sensitive to the amount of each mineral component in the mix, so to make a consistent product, the composition of the mix must be tightly controlled.
Small mineral particles are needed in order to form the silicates without having to burn the mix at an excessive temperature or to burn for an excessively long time. In this way, the cost of manufacture and wastage of energy are minimised. The different minerals in a rawmix have different grinding requirements.
Calcium carbonate (calcite) decomposes in the kiln, evolving carbon dioxide, and the decomposition "opens up" the crystal structure, making it more reactive, so particles less than 150 μm are satisfactory. Particles larger than this, unless burned very hard, leave particles of "free lime" (unreacted calcium oxide) in the clinker, which result in less alite formation and cause unsoundness (destructive expansion) in the cement.
Clay minerals such as kaolinite (Si2Al2O5(OH)4) also decompose on moderate heating (by de-hydroxylation) and so become more reactive. But unreacted particles have a more serious effect, by leaving an excess of calcium oxide in the rest of the mix. Particles less than 90 μm are considered satisfactory. Larger particles, unless burned very hard, leave masses of small belite crystals, interspersed with melt phases, that are too large for free lime to penetrate from the outside and convert them to alite.
Silicon dioxide (quartz) is unreactive unless very fine, and particles need to be less than 45 μm for easy reaction. Larger particles leave dense masses of belite that are impenetrable to attack by free lime to form alite, and may be unstable enough to invert to inert γ-calcium orthosilicate.
In these particle size requirements, a profound processing problem arises - although quartz needs to be more finely ground than the other minerals, it is by far the hardest mineral present, and this difficulty, recognised from the outset, had a major effect upon the evolution of the industry, by mandating the use of rather rare raw materials in which quartz is fine from the outset.
Effect of Homogeneity
When heated to the peak temperature in the kiln, "clinkering" takes place. This consists of partial melting, the total amount of liquid produced being typically 20-30% of the mass. The liquid acts as a solvent through which ions can be exchanged between the solid particles. It also has the effect of drawing the loosely consolidated solid particles closer together by surface tension. To ensure easy formation of the desired clinker minerals, ions must diffuse a minimum distance through the liquid. These means ensuring that each aluminosilicate or quartz particle must have the required number of calcite particles in close proximity (i.e. within a few tens of micrometres), and requires a high degree of homogeneity.
In general, if the requirements for composition and fineness are met, and the components are ground together, the homogeneity follows naturally. However, there is always the distinct possibility that the well-homogenised components may separate or “settle out” during storage or manipulation of the mix, and dealing with this effect is a major preoccupation, particularly in the “dry process”.
The importance of the first of these rules can be exemplified by looking at a simple binary mixture such as Portland cement manufacturers used from the earliest times. Expressing the chemical analyses in the way usual in the cement industry, typical Thames-side raw materials might be characterised as follows:
NOTE: LoI is "loss on ignition" - the mass lost when the material is heated to 950°C, consisting of water held in clay minerals, carbon dioxide evolved by carbonates, organic substances and so on.
If we combine these two minerals in various proportions and heat the mixture to clinkering temperature, allowing the "losses" to escape, by simple arithmetic we can calculate the analyses of the resulting clinkers, for example:
Despite the apparent similarity of the chemical analyses of these clinkers, their mineralogical compositions vary considerably. The corresponding silicate compositions of these are:
Alite (approximately Ca3SiO5) provides almost all the "early strength" (strength up to seven days of curing) of cement and a moderate amount of late (>7 days) strength. Belite (approximately Ca2SiO4) provides very little "early strength" and a large amount of late strength. So it can be seen that these small compositional changes have a large effect upon the cement's properties. In order to control cement properties, cement manufacturers aim to control alite content within ± 2 %, and it can be seen that, even if there are no sources of variability in the individual components (and this is far from being the case), to achieve this, the percentage of clay in the rawmix must be controlled within ± 0.14 %.
Accurate weighing is not enough!
Bearing in mind that the raw materials are natural products dug out of a hole in the ground, such control can't be achieved merely by weighing the components very accurately. The chalk (a relatively pure raw material), for example, will vary in its content of water and of subsequently-removed flints. Other types of raw materials have much more scope for variation. The problem is only solved by a responsive chemical control system and an elaborate blending process.
The three main topics of rawmix preparation are discussed below:
The early period in the history of Portland cement (and its "artificial cement" precursors) was characterised by a lack of efficient grinding technology. In the modern period, grinding technology becomes progressively more complex as the financial and environmental need to minimise energy use has emerged.
Rendering material into a powder, in the early industrial period, was achieved by technologies that had changed little in two millennia. For coarse grinding, there was the "roller mill" or "edge runner", consisting of one or more heavy stone wheels that were made to roll around a circular stone track, crushing lumps of material placed in their path. For finer grinding, flat stones were used: two circular stones with flat faces are arranged face-to-face: one is stationery and the other rotated in close contact with it. The upper stone has a hole in the middle (the “eye”) through which coarse material is slowly fed. Material passes between the stones and is ground by a combination of crushing and shearing, emerging at the periphery. The traditional motive power for the process was from water or wind. It is interesting to note that when James Parker commenced making his “Roman” Cement, he set up at Northfleet, not for reasons of raw materials or market, but because a water mill was available.
In the cement industry context, this technology was difficult to apply because the grinding stones consisted usually of fine-grained sandstones not much harder – if at all – than the minerals being ground. Quartz in the mix would not grind at all, and would rapidly wear out the wheels. Production of artificial cements became possible when manufacturers side-stepped the problem and adopted the Wet Process of grinding. The main elements of this process were borrowed from the pottery industry that had arisen in Staffordshire in the eighteenth century. The most important apparatus was the Washmill. This corresponded to the “blunger” of the potteries and consisted of a vessel in which the rawmix components were vigorously stirred with a large addition of water. Soft mineral components are broken down by this treatment, to form a slurry, while hard components are scarcely affected, and can be sieved out of the slurry, or are simply allowed to settle out.
A washmill of 1890 vintage.
The adoption of this process placed severe constraints upon the kinds of raw material that could be used. Both limestone and clay had to consist of soft aggregations of very fine-grained material. The quartz content of the clay, in particular, had to be very fine, but also sufficient in quantity to produce the desired silicates. These materials were fortuitously available in Southeast England. Upper Chalk is very soft and fine grained, contaminated only with fist-sized flints that are easily separated out by the washmill. Alluvial clays from estuaries can be selected such that the coarser quartz has already settled out in the up-stream sections of the river valley, and the clay of the Medway estuary became famous for its high content of ultra-fine silica. Easy access to alluvial clay and the already well-developed soft chalk quarries brought William Aspdin from Rotherhithe to Northfleet.
The washmill can be thought of as a giant kitchen liquidiser. Although washmills escalated in size during their history, the basic design became standardised quite early. The mill consisted of a circular (or often octagonal) stone-lined bowl with a rigid centre-post, about which rotates a set (usually four) of radial arms from which harrows are suspended on chains. The harrows each contain 10-20 vertical steel-bar teeth or ”tines”. The rotation of the radial arms drags the harrows around the mill and they agitate the coarse raw materials and water constantly added to the mill. Part of the circumference of the mill is made up of iron grids or screens which allow the liquid slurry, once it is fine enough, to splash out into a sump surrounding the mill.
The action of the mill is to a large degree autogenous – that is, the motion of the larger chunks of raw material crushes the smaller pieces. This is particularly the case where the chalk contains flints: in this case the mill is designed primarily to keep the flints in violent motion, and these in turn grind the softer minerals by a mixture of crushing and attrition. Because chalk commonly contains as much as 5% flint, such washmills rapidly fill up with flint, and they were usually constructed in pairs, so that the flint could be emptied out of one while the other produced slurry.
A washmill of 1950 vintage.
Flat millstone grinding as used in the 1890s. Based on figure in Butler, p 112, modified for wet grinding. The standard design had stone diameter 54", with the "eye" diameter 20". The upper stone rotated at 150 rpm, the clearance being adjusted from the base of the drive shaft. For slurry grinding, output per pair of stones was 2-3 dry t/hr, with a power draw of 20-25 kW.
The typical size of washmills changed with time:
So, for example, in 1890 a block of seven chamber kilns would use around 280 tonnes of dry raw material a week. A single washmill grinding 8 dry tonnes per hour could supply this in 35 hours, or about six hours per day in a six-day working week, allowing adequate time for removing flint and maintenance.
Such washmills were typically fed with raw materials by tipping whole tram wagon-loads over the side, at such a rate that the mill did not overfill. More modern mills were fed by conveyor belt from feeders delivering a constant flow of materials, often with weighers to meter the components. Elaborate arrangements were also added to allow the mill to be easily and quickly "stoned out" by dropping flints through doors in the floor of the mill.
In the earliest process (pre-1870), the slurry was made with 70-80% water content by mass, with a milky consistency. The slurry was run by gravity or pumped into "slurry backs" - large shallow tanks, usually brick lined, but often just holes in the ground. The slurry was left here to settle. Adjustable weirs in the sides of the back were gradually lowered to allow relatively clear surface water to run off, while some water, particularly in the latter stages, percolated through the base and evaporated from the surface. During settlement, the larger particles in the slurry would drop to the bottom. If left long enough, the slurry would dry to firm blocks, containing perhaps 10-15% water. These were removed from the back, taking care to leave behind the layer of coarse material at the base, and transferred to "drying flats", usually consisting of floors of thin masonry or iron, heated from beneath by flues from a coal-fired furnace or waste heat from coke ovens. Preferably, according to Butler, the slurry was removed from the back when at the consistency of "soft butter", because, although wetter, this gave better heat exchange on the drying flat. Settlement of slurry to the "stiff" state could take 4-5 months, requiring huge areas of slurry backs for a modest cement output. Taking the slurry out at the "buttery" stage halved this time, but the process was still expensive in time and area when cement output began to escalate. A further problem with lengthy settlement was the tendency of the mix to separate out mineralogically, usually with excessive chalk in the lower layers and excessive clay in the upper layers.
In 1870, the so-called "Goreham" or "thick slurry" process came into use. This involved washmilling a much thicker slurry with 40-45% water content, which was then led directly to the drying flats. This allowed the process of making feed ready for the kiln in only the few days needed to dry it out, and hence solved the major logistic problem of the slurry backs. This was, of course, achieved at the expense of a considerable increase in energy consumption, on three fronts:
somewhat more washmill power was required
much more drying energy was required
extra grinding was required
Extra grinding was needed because coarse material would not now settle out. Grinding was almost always accomplished by running the slurry through sets of flat stones (right), and this could double the overall power consumption of the process. And despite the extra grinding, raw material fineness never again achieved the high standard achieved in the sedimentation process. Furthermore, there was distinctly less homogenisation effect in the new process - slurry was commonly pumped direct from the mill system to the drying floors without any intermediate storage or blending. The "capacitance" of the system was reduced.
The problem of high drying cost was solved soon afterwards by using kiln exhaust gases in chamber kilns. Chamber kilns had more than adequate waste heat available for drying, so there was no reason to make slurry with water content lower than the typical 40-45%.
Emergence of Modern Grinding Technology
The development of industrial grinding equipment only became possible following development of the steel industry. Bessemer and open-hearth mild steel became readily available in the 1870s and specialised alloy steels were developed in the 1880s. This made available steels for strength, hardness and abrasion resistance, and many different designs of grinding equipment emerged from 1885. The first developments were various forms of roller mill, in which the feed is crushed between a circular steel roller and the steel surface on which it rolls. The new element that became available with these mills was pressure: flat-stone mills had a carefully adjusted clearance between the wheels, and any actual contact would cause them to wear out in minutes. By contrast, mills with hard steel grinding surfaces could be put under pressure, by gravity using heavy rollers, by spring- or hydraulic loading, or by use of centrifugal force. One of the most successful of these was also one of the earliest - the Griffin mill.
The Griffin mill 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.
Roller mills opened up the possibility of grinding suitably dried raw materials in the dry state, and this opportunity was taken up in areas where reasonably dry limestone was available - notably in the Blue Lias districts. The dry meal, slightly wetted, could be made into briquettes in a brick-press - also a new technology of this period - and loaded into a bottle kiln or one of the newly developed continuous shaft kilns. When rotary kilns were developed such mills could be used to prepare dry process feed. The first dry process plant (Norman) had seven Griffin mills for rawmix grinding. As with the other designs, the mills ran "cold", and it was necessary to dry the crushed raw material before feeding to the mills, to prevent damp material from building up on the grinding path.
Roller mills were not suitable for finish grinding of slurry, but were used in the Northeast when preparing the relatively hard local chalk: chalk was crushed then ground in a damp state to millimetre size in a roller mill before washmilling.
Early ball and tube mill set.
Four Newell ball & tube mill sets circa 1912.
Tube mills at Wilmington. Both the ball mills (off-frame, right) and the tube mills were closed-circuited. Rotary screens at the mill outlet caught oversize which was pumped back to the mill inlet.
Unlike roller mills, ball mills represented a new departure from ancient grinding process principles, and developed when sufficiently tough machinery could be produced. The ability to make bearings capable of dealing with high static and dynamic (vibrational) loads was probably the key breakthrough. A ball mill consists of a horizontal cylinder partly filled with pebbles or steel balls that is rotated about its axis, and material is ground by crushing and abrasion in the tumbling action.
Pre-grinding Ball Mills
In early terminology the term "ball mill" had a more restricted meaning: it was a short mill (with length equal to its diameter or less) with stepped liner plates to increase the cascading action of the media. It was fed through a central port at one end: in some mills the fine material left through the periphery through slots between the liner plates: less often, the fines left through grids in the end wall. The mill combined a grinding and sieving action, and the peripheral discharge type was enclosed in a casing to capture the flying dusty material leaving it. These were first developed in the late 1880s. The commonest form of this design in the UK was the F. L. Smidth "Kominor" mill, although all equipment suppliers had their own designs. These were used for dry grinding of dry raw materials, coal and clinker, and because of the absence of sensitive moving parts in its grinding compartment, they could also be used for wet grinding.
Much closer to the modern idea of a ball mill was the tube mill. These were longer with length four to six times the diameter, and rotated on trunnion bearings at either end. The inlet trunnion was hollow and the feed entered through this. The fine product passed through a slotted diaphragm at the outlet and was lifted into the outlet trunnion with scoops Tube mills were designed to "finish" grind material that had already been reduced to about 0.75 mm, and in fact would not work effectively with larger feed. Because feed above the critical feed size passed through unground, a "nib screen" was provided to sieve out the occasional grossly-oversized particle from the product. In the early mills the grinding media consisted of flint pebbles. Cemented-in flint ("silex") blocks were often used as a lining. Because of its length, a progressive size reduction took place, and the meal finally left through a circumferential grating at the far end. Whereas the size of product on the "ball" mill was controlled largely by the size of the outlet sieves, the size of product of the "tube" mill was controlled solely by the feed rate, a reduced feed rate causing the material to remain longer in the mill and receive more grinding action. This meant that, for the first time, there was virtually no limit on the fineness that could be achieved. From the early 1900s, both for cement grinding and for grinding hard raw materials, whether by wet or dry process, the standard arrangement was a ball mill for preliminary grinding (to below 0.75 mm), followed by a tube mill for fine grinding. Typically the ball mill was placed on an upper floor so that material could move by gravity. An example was the original rawmill system at Aberthaw. Four Ernest Newell ball and tube sets were provided to serve two kilns. The ball mill by this stage used forged steel balls as grinding media and had its outlet in the form of grids in the end wall. The slurry from there ran straight into the tube mill inlet. The tube mill now had two chambers: first about one-third of the length, charged with forged steel balls, and the second charged with flint pebbles. The chambers were separated by a slotted steel diaphragm.
The next logical step was to simplify the above system by combining the ball and tube mill functions into a single mill tube. In its earliest form, because it combined two functions, it was known as a combination mill. Today they are termed ball mills or tube mills indiscriminately. This development was facilitated by the escalation in the size of mills. The correspondingly larger mill trunnions were capable of taking a coarsely crushed feed. Multiple chambers, separated by slotted diaphragms, had successively smaller steel grinding media. This became the standard format for wet grinding of hard materials from 1920, and the design changed little thereafter. Pebble media ceased to be used because the power drawn (and therefore the output) is proportional to the mass of charge, and since steel has three times the density of flint, a given sized mill would make three times the output with steel media.
Closed circuit operation of wet ball mills became normal, with the product passing through some sort of size classifier that would return the coarser fractions to the mill inlet for re-grinding. At first various sorts of sieve were used, and later DSM screens, sieve bends and hydrocyclones. Wear of the mill linings was reduced by the use of rubber linings, and media wear was reduced by the use of increasingly sophisticated alloys.
In the closing years of wet process systems a return was made to the concept of the pebble mill, with the FLSwashdrums installed at Northfleet. This was a special case, in which soft chalk, normally washmilled, was to be ground at a rate of 6 million tonnes a year. Since this chalk contained 5% flint, a method was needed to separate out 300,000 tonnes per year of flint with minimal operating cost. The washdrum is an autogenous tube mill divided into two chambers by a dam ring. The chalk feed keeps the first chamber topped up with fresh flint, which grinds the slurry. The fine slurry passes out through a circumferential grid upstream of the dam ring. The flint gradually overflows the dam ring into the second chamber, in which it is washed clean with fresh water. The washed flint finally exits the mill onto a conveyor, and the dirty water is pumped to the first chamber to make the slurry. The slurry produced is coarse, and after separating the fines with sieve bends, the coarse fraction is re-ground in an ordinary tube mill.
Because sand is so much harder than other rawmix components (e.g. on Moh's scale, clay minerals are 1-2, calcite is 3 and quartz is 7), where it is needed to increase the silica content of the mix, it is best to grind it separately, because if mixed with softer components, these simply over-grind, and the resulting fines "cushion" the grinding effect and protect the sand grains from the grinding action. Wet ball-milling remains the best available technology for grinding sand, and results in an easily-handled slurry that avoids the health hazards of finely ground quartz powder. As mentioned above, quartz must be ground below 45 μm, and this is achieved using a mill with a rubber lining and cylindrical zirconia media, 20 mm down to 5 mm. Although expensive and of somewhat lower density than steel, their hardness gives them a much longer life. They also have the advantage that, if the mix is to be used for white clinker manufacture, they avoid the contamination of the slurry with iron and chromium which is otherwise considerable.
The use of roller mills for dry rawmix grinding became extinct as the "old" dry process kilns became extinct around 1930, and when the "new" dry processes arrived in the 1950s, ball mills had become standard. Even dry process raw materials contain a certain amount of moisture, in the range 2-10%, and three strategies were possible for dealing with this:
first dry the material in a separate drying apparatus, then grind in a conventional ball mill
use a mill with an initial drying chamber, swept with hot air from a furnace
use an air-swept mill that dries and grinds simultaneously, swept with kiln exhaust gases.
Separate driers were used on the early dry process plants. For the Lepol process and for long dry kilns, no waste heat is usually available to heat the rawmill, and the rawmill is detached from the kiln system and supplied with a furnace to provide hot drying air. For suspension preheater kilns, air-swept mills were coupled to the mill exhaust.
Separate driers, in the form of cylindrical rotating drums swept with hot air from a furnace or the kilns, were used by the five early dry process plants: Norman, Kirtlington, Premier, Southam and Ellesmere Port. They were also used by the Scottish slag-based plants and by the three Anhydrite Process plants. As mentioned above, Norman used Griffin mills. Kirtlington apparently used a single ball mill, while Premier and Southam used ball-and-tube mill sets, and Ellesmere Port used a mixture of ball-race mills, Griffin mills and a ball-and-tube set. The slag and anhydrite plants all used combination ball mills.
The characteristic feature of dry process rawmills remote from the kiln system is a preliminary drying chamber into which hot air is blown from a furnace that might be fired by coal, oil or gas. This is followed by one or more grinding chambers containing conventional grinding balls. The mill is operated in closed circuit, with an air separator catching all but the finest fractions of the mix and returning it to the inlet.
A design used by FLS consisted of a cylindrical mill with two compartments: the first contained no media but used lifters to cascade the raw material through the air-stream. The second was a conventional ball mill chamber. This design was used at Padeswood.
Among these, the three Blue Circle plants (Cauldon, Weardale and Cookstown) also added Aerofall mills (short, semi-autogenous, air-swept ball mills) ahead of their Double Rotator mills to improve output by feeding them with a fine grit.
The gas-swept ball mill is a mill (typical length 2-3 times its diameter) with large trunnions to allow a large gas-flow in and out. The material is moved through the system entirely by the gas stream. Although few of these were installed, at Plymstock and Hope, they represent the philosophy that ultimately became universal - that of making the rawmill into the last heat exchange stage of the kiln system.
The nearly universal use of roller mills for raw milling today, involving overcoming their considerably greater complexity, has been due to the need for reduction of energy consumption. Ball mills are extremely inefficient, only a few percent of the energy being used for breaking particles, and the rest ending up as heat, much of which may be wasted. A similar degree of grinding can be achieved with a roller mill, using only a third the electrical energy.
A basin 12 ft deep, and with a 10 ft diameter centre post ("dumpling") would when brim-full hold 1136 m3 of slurry (1788 tonnes at 42% moisture, or 1037 dry tonnes, equivalent to 677 tonnes of clinker). A boom, rotating about the centre post, carried stirring and agitating devices.
66 ft "sun and planet" mixer at Wilmington. Four inter-meshed stirrer paddles rotating (1-4 rpm) in the same direction usually carried chains at the lower end to drag the bottom of the basin. The differential moment of the stirrers caused the boom to gradually rotate around the basin in the opposite direction.
Air-agitated 66 ft mixer at Masons quarry. The boom is driven round by a drive wheel running on the rim. On the boom, a compressor supplies air to 10-15 pipes extending to the bottom. Compared with stirrers, air produces effective mixing of the layers of slurry, but can leave dead zones unless the air-flow is carefully maintained and balanced.
In the early days of the industry, when manufacture was guided only by primitive rules-of-thumb, little attempt at mixing the rawmix was attempted, beyond the mixing of the measured amounts of chalk and clay in the washmill. Although the very early industry could do a certain amount of mixing of slurry in filling the large "slurry backs", the situation took a turn for the worse with the introduction of the chamber kiln. With these, the holding of slurry stocks was often reduced to zero, the washmills being only operated when a kiln was being loaded, and the run-of-mill slurry product was pumped direct to the kiln drying chamber. A few larger plants, up to the end of the nineteenth century, had rectangular slurry holding tanks, but only for logistic purposes. With the beginning of applied chemistry in the industry (not in Britain!), clinker with a higher alite content started to be made, and it became clear that, to achieve this without making over-limed clinker, batch-to-batch chemical variations had to be smoothed out. The need for tight control became greater with the arrival of the rotary kiln, where large hour-to-hour chemical variations would render the kiln uncontrollable.
Wet process slurry blending
With the majority of plants using wet process, the start of the 20th century saw the first circular slurry mixer basins. Although larger and smaller sizes are occasionally seen, basins of 66 ft (20.12 m) internal diameter became by far the most common.
The solids in a rawmix slurry usually have a mean density close to 2700 kg/m3, and the density of a slurry with M% water content is given by ρ = 158571 / (M + 58.73) kg/m3.
Such mixers, alone, were very efficient homogenisers if kept in working condition, and if linked in series ("cascade blending") could eliminate variation in the most ill-behaved rawmill product. However, a further blending stage was in many instances added before the storage basins, in the form of cylindrical blending tanks ("doctor tanks"). These were sized in the range 100 - 1000 m3 and were agitated by blowing in high volume compressed air at the base. Tanks were filled then homogenised and tested before being transferred to storage. Often these were filled in pairs, with one slightly over-limed and one slightly under-limed, a calculated quantity from each then being simultaneously run to storage so that target chemistry was obtained.
Probably the most elaborate three-stage blending system, at Shoreham. The washmills delivered slurry with a ±4% calcium carbonate range, filling a "preliminary" silo in about 4 hours. Appropriate quantities of slurry from two "preliminary" silos were pumped to a "final" silo. In the unlikely event of this blend being off-target, two "final" silos could be blended off before running the slurry into one or more kiln feed storage tanks. With such a system, it is easy to control the slurry variability to "undetectable" levels. However, as originally installed, it was necessary to employ one person per shift solely to control the pumps and valves. The whole operation was later (1980s) completely automated. In its heyday, the plant made 59 t/h of clinker, so three full storage tanks gave 32 hours run. The wash plant ran 24 hours a day, 7 days a week.
At the beginning of the twentieth century, the perfection of wet process blending posed major problems for proponents of the dry process, because the blending of dry powders was much more problematic, and could not compete. Throughout the first half of the twentieth century, the lack of reliable dry process blending was put forward as the main reason for sticking with the inefficient wet process. The early (pre-1930) plants that did install dry processes - Norman, Kirtlington, Southam, Premier, Ellesmere Port and Coltness - had, with the exception of Ellesmere Port, "forgiving" rawmixes in which the individual mineral components were not far off the desired chemistry, so poor blending had a comparatively small deleterious effect. In the case of Ellesmere Port, the kilns were soon converted to wet process.
The first installation, at Norman, was described in detail in The Engineer (Nov 13, 1908, p 518). The "raw meal mixer" was designed by William Gilbert and David Butler. It consisted essentially of a silo containing 6 hours' run of raw meal, from which meal was extracted at eight different levels and recirculated by screws and elevator at six times the kilns' usage rate. The philosophy relies upon the idea that the meal flows en masse through the silo, so that the meal extracted from the bottom is much older than that extracted from the top, and eight different "ages" of meal are blended. However, subsequent experience shows that such silos in practice flow by "rat-holing" and the material extracted, whether at the top or the bottom, is essentially that currently leaving the rawmills, while the bulk of the material in the silo remains "dead". There is a further, more subtle defect: even if the mixer functioned as hoped, the "moral hazard" built into the design is that if the extraction feeders block, cut out, or otherwise cease to deliver, then neat run-of-mill rawmix will run directly to the kilns without any interruption, and this will always become the default condition. When flowing, it is likely that a highly chaotic "feast and famine" flow would result, whereas, from the operator's point of view, the equipment would "work best" when completely blocked.
The next four plants made no pretence of blending, putting the emphasis on exact weighing of raw materials at the raw mill. Of course, this, even if it works, makes no allowance for the inherent variability of the raw components. Coltness had a system of four bins around which material could be circulated. This was upgraded in the 1930s to operation with Fuller Kinyon pumps, and a rather fancy (for the time) system of remote controlled valves, allowing mill product and recirculated meal to be directed as required to recirculation to nominated bins, or transfer to the kiln feed bin, operated manually or to an automated program. This system was the subject of a Fuller patent (US 1812604: 1931). In favourable circumstances, this might dilute the mill product variability up to four-fold, but once again, such a system would be prone to disruption by failure to flow, or by a decay of the necessary discipline.
Principle of the air-fluidised quadrant blending silo
A breakthrough in dry blending occurred in the 1950s, when Fuller developed the air-fluidised blending silo (patent US 2844361: 1958). Compressed air is injected into the silo through porous pads in the base, causing the raw meal to become fluidised. In one part of the base - typically a one-quarter sector ("quadrant") or a one-eighth sector ("octant"), a higher-pressure air supply is provided. This causes the column of meal in that sector to become more dilated with injected air than the rest. It thus has a lower density, and the meal rises, while the denser meal in the adjacent sectors flows in at the base. After a few minutes, the high pressure air is moved to the next quadrant, thus applying the convective rotation of material across a different diameter. In behaving as a fluid, the dry material is now capable of the same sort of blending behaviour found in a liquid slurry. The first quadrant blending silos were commissioned at Cauldon in 1957.
Such silos are used in two ways:
batch-wise, in which the silo is filled, agitated, and the meal is then transferred to storage, or
in cascade, in which the silo is kept full and continuously agitated, is fed with rawmill product and continuously overflows to storage.
Batchwise operation achieves the full blending effect, and in favourable circumstances reduces the variability 10-fold or more, but any slight off-target errors in the filled silo analysis cause corresponding step-changes in the kiln feed chemistry. Cascade operation at least avoids these step changes, but the overall blending effect is at best small, and usually non-existent. As originally designed, the systems would be operated in batch mode, with at least two blending silos - one filling, while the other is blended and transferred. The blending silos were typically located on top of the storage silos, so they could be emptied out by gravity.
Rawmix blending and storage system for Padeswood Kiln 3.
Early installations often reduced variations further by maintaining multiple storage silos with large capacity in comparison with the kiln throughput, and recirculating the meal around the silos at a high rate, with the kiln feed stream bled off the recirculation. The power requirement for air-fluidised blending was high because of the large volume of expensive compressed air used, and the screws, airslides and elevators of recirculation systems added considerably to this. Thus, although these systems made possible dry process blending almost as efficient as that of wet process, this was achieved at a considerably greater energy expense. The system for the long dry kiln at Padewood shown had four 500 t bending silos to the right, and two 2800 t storage silos to the left. Meal extracted at 100 t/hr from the base of the storage silos is lifted by elevator to a conveyor which returns some to the silos, while the rest carries on to the kiln feed hopper. A typical kiln feed rate was 66 t/hr, so the total storage capacity was about 115 hours - almost 5 days' kiln run. Subsequent long dry kilns had somewhat smaller relative storage capacity.
The design of dry process systems subsequently standardised on designs such as that at Hope, with aerated blending silos discharging by gravity into sets of storage silos, a compact system being desirable in modern plants in which raw material drying is done by kiln exhaust gas, the preheater, rawmill and blending system being grouped together.
As larger plants were constructed from the 1980s onward, the high energy consumption of dry process blending systems has been addressed. The chemical constancy of rawmill product has been improved by installation of elaborate pre-homogenisation stockpiles and automated on-line analysis systems. At the same time, rawmix storage silos have been radically re-designed so that with comparatively small energy consumption, a single silo can be used as a cascade blending device.
The chemical control of rawmix is required in order to ensure that the rawmix components (at their simplest, limestone and clay) are in the right ratio to make clinker of the desired alite content. To achieve this, a feedback control system is almost always used. This consists of taking a sample of the rawmill product, analysing it, and depending upon the deviation of the result from the "set point" value, to make a change to the feeder(s) up-stream of the mill. The strategy for making changes to the mill feed varies considerably with the system of plant employed, but the objective is always to keep the chemistry close to the target value, and equal to the target value on average, without any excessive "cycling".
The chemistry of the rawmix at the mill outlet varies because of random changes in the composition of the components and in the amount of each fed to the mill. Irrespective of the cause of a variation, the purpose of the system is to make appropriate changes to the mill feed to restore the target composition. The time taken by the process critically affects the success of control. A change in the raw material feed may take some time to appear at the outlet. Ball mills have a residence time of 15-25 minutes, and this is considerably increased if a closed-circuit with high recirculation is employed. Other mills have smaller lags: only a few minutes in the case of a roller mill. Once taken, the analysis sample has to be processed and analysed. Simple methods can do this in 10 minutes or so, and instrumental techniques for full analysis can be turned around in a similar time if automated. Finally, a decision must be made as to the feed-rate changes needed, and the changes implemented. The tendency for the system to produce cyclic variations, or even catastrophic out-of-control situations, increases with the "process lag" - the sum of these delay times.
A further complication arises from lack of precision in the chemical analysis. In addition to the "inherent variability" of the system (due to raw material changes and feeder faults), adjustments made on the basis of incorrect information may result in "induced variability", so that the overall "perceived variability" is the vector sum of the inherent and induced variabilities and the measurement error. Analytical errors arise from three main causes:
random errors in the actual analysis
random variations in sample preparation
random errors in sample gathering
Even in the days of the more primitive test procedures, and certainly today, sampling errors are the most significant, and analysis errors the least. However, sampling is often a "Cinderella" in the allocation of capital spending on analysis systems.
In general, the best way to compensate for the unavoidable measurement errors is always to sample, analyse and adjust the process stream as frequently as is economically feasible. In the automation of the chemical control loop, it is often found that less precise but more frequent data (with correspondingly smaller and more frequent process adjustments) yields better control.
A typical rawmix preparation sampling regime. If the raw materials are relatively consistent, then they may only be infrequently sampled. Some systems may have multiple blending stages. The most important sample, taken hourly at least, is the rawmill outlet sample.
In addition to controlling the rawmill product, a control strategy has to consider the down-stream effects of the blending system. Where there is an elaborate blending system, tight control of rawmill feed is less necessary, and in some systems cycles are deliberately induced, although performance is always best if the amplitude of these is minimised. The performance of the blending system is monitored by sampling and testing at each stage.
Although the amount of feed adjustment might vary, the sense of the adjustment is straightforward enough when a simple pair of raw materials is involved. An early tester, obtaining a "high" carbonate result (i.e. the clay content is low) would increase the clay feed, and continue adjusting it until the carbonate target value was obtained. The situation with more than two feed materials is more complex, and generally involves selecting test results that will isolate the effect of each material. Where more than three components are put into the mix, as was common from the mid-20th century onwards, deducing the adjustments required for each, based on a full analysis, is a mathematical problem that can only be solved satisfactorily by computer. In combining this calculation with the dynamics of the control system, a convergent iterative approach is usually most effective. Linear programming principles are often involved, and a high degree of mathematical proficiency is needed for the design of the control algorithms.
Chemical analysis techniques
An early calcimeter design by Scheibler. Operation: a weighed sample of dry rawmix was put in bottle A. An excess of dilute hydrochloric acid was put in vial S. With the valve on B and valve P open, water was pumped out of reservoir E into the U-tube until it was filled to the zero mark with both limbs level, then both valves were closed. The vial S was emptied onto the rawmix by shaking A. The CO2 produced inflated the bladder in B, dsplacing air into the U-tube limb C. Meanwhile water was let out of U-tube limb D through valve P, and finally adjusted so that limbs D and C were level, ensuring that the air in C was at atmospheric pressure. The reading on C was then taken as the volume of CO2 produced. This was converted to mass of CO2 allowing for temperature and atmospheric pressure.
For the first thirty years of the industry there was no chemical control at all: the components were weighed out (or more usually measured by volume) more or less accurately. By aiming for a composition yielding equal amounts of alite and belite, the chances of making excessive free lime or lower silicates were minimized, but variations in the composition of the individual mix components meant that the mix was still very variable and prone to serious failures. Chemical control began with the use in Germany of the calcimeter in the early 1870s. This apparatus measures volumetrically the amount of carbon dioxide evolved when acid is added to a dried sample of rawmix. This quantity is more or less proportional to the calcium content (although errors occur if the amount of magnesium carbonate or non-carbonate calcium in the mix changes). The method started to be used in Britain in the 1880s – the delay being due to the total absence of competent chemists in the industry before then – and was by 1900 employed in most large plants.
Around 1900, an alternative method arose – that of treating the dried rawmix with a slight excess of standard acid, then back-titrating the excess with standard alkali. This obtained essentially the same result as the calcimeter, and was subject to the same errors, but was simpler and quicker – the calcimeter required careful measurement and compensation for temperature and pressure. It began to displace the calcimeter after WWII, the delay being mainly due to pure inertia. By either method, the calcium carbonate content of the rawmix could be estimated with a precision of 0.25% or better, and, given the facilities for control, allowed control of the final carbonate content of the mix to 0.5% or better. This represented a dramatic improvement over rule-of-thumb non-chemical methods, bringing at least the short-term variability of clinker potential alite content down from around 25 to around 6.5, and thus allowing alite levels to be safely raised.
Longer term variations could still occur due to changes in the composition of the rawmix components – particularly the clay, which is prone to changes in the silica/alumina ratio. Simple “carbonate control” could still work, provided that such changes were slow, and the detailed analysis of the clinker was performed regularly. In typical twentieth century practice up to 1970, clinker would be subject to a “full analysis” (at least SiO2, Al2O3, Fe2O3 and CaO contents) every day, by qualified analysts. The rawmix would be analysed for “carbonate content” by a “tester” at several points in the rawmix preparation process at least hourly. On the basis of the trends in the clinker analysis, the tester’s “carbonate target” would be adjusted as often as necessary to keep the clinker alite content more or less constant. The situation was more complex where more than two components were being used in the rawmix. This was initially rare but became more common during the twentieth century. The tester’s calcimeter reading could only be used to control the amount of high-carbonate component in the mix. If this was being combined – say – with a mixture of shale and sand, then these two had be kept in a constant ratio until a “full analysis” became available. Rarely, a more desperate option was to employ qualified chemists for rawmix control around the clock, using a "high-speed" (i.e. corner-cutting) classical technique. Rawmix would be clinkered, and analysed gravimetrically for silica (once per shift), and titrimetrically for iron and calcium (four times per shift), allowing - in theory - control of a four-component mix. More frequent testing, although needed, was not possible by this method.
The earlier classical analysis techniques were slow (if done properly) and might not be available at weekends, so this process had inconveniently long time lags. The use of instrumental techniques was intended to speed the process up. Colourimetric methods for measurement of SiO2, Al2O3 and Fe2O3 were first on the scene in around 1960, but atomic absorption and XRF techniques followed shortly afterwards. The colourimetric and atomic absorption methods were still fairly lengthy “white-coat” laboratory techniques, but XRF revolutionised the situation, in that a “full analysis” could be completed in 20 minutes or less. Furthermore (although this involved a long period of cultural change that is still not complete) the XRF spectrometer could be operated by non-chemists and be placed outside the formal laboratory setting, in the midst of the process plant. Despite its high cost, the very high productivity of an XRF (doing anything up to 800 analyses per week) means that it is now standard equipment on every cement plant.
A rawmix sample prepared as a pellet pressed in a steel ring. The ring is held down on a tungsten carbide flat "anvil" surface, the powder (around 10 g) is added, and is pressed against the flat surface with a hydraulically powered tungsten carbide die.
A rawmix sample prepared as a glass bead. The bead is made from 2 g (loss-free) of rawmix and 8 g lithium tetraborate, fused at 1100°C and cast in a 38/40 mm mould made of 95%Pt + 5%Au. The flat base of the bead is presented to the x-rays.
XRF gradually evolved after WWII. The technique involves irradiating a flat specimen of the material to be analysed with “white” (i.e. broad, continuous spectrum) x-rays from an x-ray tube operating usually at an EMF of 20-70 kV. The incident x-rays cause the specimen to fluoresce, emitting x-rays of lower energy that have specific sharply-defined wavelengths characteristic of the elements present. Detectors arranged around the specimen sort and measure the intensities of the emitted x-rays, and a computer performs the extremely complex calculations that convert these measurements into analytical data. Typically seven or eight elements are measured, but should it be necessary, more can be included, and XRFs measuring up to sixteen elements routinely are not unknown.
In addition to the high cost of the spectrometer, costly sample preparation equipment is also necessary, typically doubling the cost of the installation overall. This is because sophisticated techniques are required to ensure that the material exposed at the flat face of the specimen in completely homogenous and representative of the sample. There are two broad techniques of sample presentation, both of which are typically used. The “cheap and cheerful” method is to grind the material to be analyzed to below about 5 μm, and press it in a die at 100-400 MPa pressure, to form a flat-faced “pellet”.
Measurements on this can be highly precise, but can have poor accuracy unless complex ad hoc continuous calibration processes are used. The more accurate (but slower and more expensive) technique is to dissolve the sample in a molten lithium borate glass at around 1100°C., and cast the resulting liquid into a flat-surfaced “bead” in a platinum/gold mould.
By producing a homogenous, non-crystalline glass, this technique renders the sample into a form that is independent of the sample’s mineralogical history, and because it can be calibrated with pure chemicals, yields an “absolute” analysis. Typically, both techniques are used, with “pressed pellets” producing fast, high-productivity results, and “glass beads” supervising the accuracy of the pressed pellet calibration.
Take-up of XRF as a technique was slow because the cost of the equipment was daunting, but mainly because the poor standard of calibration and operation in the early days made the technique seem second-rate. Successful production of high-accuracy data requires a high degree of specialized knowledge in the supervision of the instrument. In an early, notorious case, an expensive instrument was used un-calibrated with the operators controlling rawmix to an “x-ray intensity target”. While dressed up as a no-nonsense, practical approach, this was merely an expression of the low competence of the plant personnel. Another problem was that the early instruments were not provided with enough computer power to handle the complex matrix correction calculations necessary for accuracy. By 1970, there were 7 instruments in place on large plants. Smaller plants could not justify the expenditure, but these gradually disappeared, and by the end of century, all plants had at least one. Modern software allows the straightforward application of "fundamental parameter" corrections, providing accuracy, for many elements, better than any other technique.
Practically ever since XRF became available in the 1960s, attempts were made to use it “on-line”, provided with samples from automatic samplers, and feeding back results into a control system without human intervention. The early attempts were failures, but sufficiently reliable systems began to be installed on new cement plants from the late 1970s onward, the key factor being the development of systems by the kiln equipment suppliers – notably Polysius and FLS, who recognised that in the future such systems would be part of the standard equipment of a cement plant. There are two competing approaches: a small "ruggedised" instrument directly attached to the process stream, or an "auto-laboratory" in which automated sampler(s), automated sample preparation and automated XRF instrument are run by a robotised laboratory. The first suffers (usually) from compromises in precision, and the latter from slower sample turn-around leading to longer process lags.
Systems using automatic samplers, conveying materials to a central robotised sample preparation and analysis module, still stand or fall by how representative the sample is of the process stream. This presents major problems, particularly where analysis is attempted on materials that have not yet been ground. This has led – from around 1995 onwards – to the use of an alternative technique. Prompt gamma neutron activation analysis (PGNAA) produces a similar element-specific spectrum – this time of gamma rays, and instead of stimulating these with x-rays (which can only penetrate a few micrometres into a sample), neutrons are used. Most light elements are nearly transparent to neutrons, so a neutron beam can penetrate a mass of material a metre or more in thickness. The PGNAA analyzer can be wrapped around a vertical chute or a horizontal conveyor belt on which the entire process stream is being carried. This avoids the sampling problem, and allows, for example, the entire feed to a rawmill (consisting of crushed rock, say 50 mm down in size) to be analysed. A typical PGNAA analyser produces analyses every minute, and these data can be fed back to the mill feeders to control the ratios of rawmix components on the run. The fast reaction time of the system means that less money need be spent on complicated arrangements for blending out the variations in the rawmix after the rawmill.