Crushing and pre-blending
A shed full of layered rock does more for kiln stability than any control system, costs almost nothing to run well, and is the most commonly neglected asset in a cement plant — because when it is run badly, nothing goes wrong here.
On this page · 16 sections
- 01The one idea
- 02The parts — four machines, only one of which blends
- 03First principles — blending is a square root
- 04What must be true
- 05The quantitative anatomy — the variance cascade
- 06What goes wrong here
- 07Scenarios from the field
- 08The numbers that matter
- 09Build the model — the variance chain, and what each stage is worth
- 10Upgrade paths
- 11The frontier, and who is running it
- 12What’s changing now
- 13Check yourself
- 14Mini case — the silo that was not needed
- 15Go deeper
- 16Carry forward
§ 01 The one idea
A pre-blending stockpile is a low-pass filter made of rock. It takes a chemically noisy stream from the quarry and returns a smoother one, and it does so by a mechanism with an exact mathematical form — which means the improvement available is calculable rather than a matter of judgement.
The problem is diagnostic, not technical. A badly run stockpile has no symptom of its own. The pile looks like a pile. The reclaimer runs. Tonnes go out. The symptom appears four machines downstream as an unstable kiln, wandering free lime and a fuel rate set defensively high — and it gets diagnosed there, where it is expensive to fix, by people who have never been asked to look at the stacker.
F6 §03 established that a kiln’s fuel rate is set by the worst hour, not the average one, so every reduction in variability converts directly into heat rate. This lesson is where most of that variability is actually removed — and where it is removed most cheaply, because the mechanism is geometry rather than machinery.
§ 02 The parts — four machines, only one of which blends
| Stage | What it does | Blending effect |
|---|---|---|
| Primary crusher impact, hammer, jaw or gyratory | Takes run-of-mine rock at up to a metre down to 25–75 mm. Reduction ratio around 20:1 in one stage; the whole quarry-to-raw-meal chain is a reduction of 5,000–10,000:1. | None. Crushing changes size, not composition. A crusher cannot fix chemistry and is not asked to. |
| Stacker | Lays the crushed stream down in thin horizontal layers along the pile — chevron (roof-shaped), windrow (side by side) or a circular equivalent. | Creates the structure. The blending happens later, but it is designed here. |
| The pile | Holds typically several days of kiln feed. Its length sets how many layers can be laid at a given stacking rate. | Stores the structure. |
| Bridge or side reclaimer | Cuts a vertical face across the pile, taking a slice of every layer at once. | This is where the blending happens. If the reclaimer does not cut the full cross-section, the layers do not mix and the pile is a storage heap. |
§ 03 First principles — blending is a square root
Reclaiming across the face takes a small quantity from each of many layers and combines them. Each layer was laid down at a different moment, from a different part of the quarry, so their compositions are roughly independent. Averaging independent samples is the oldest result in statistics: the standard deviation of the mean of n independent values falls as 1⁄√n.
So the blending ratio a stockpile achieves is, to a first approximation, the square root of the number of layers cut in cross-section. This is not an analogy. It is the published relationship, and the numbers confirm it: simple storage systems reclaim only 4 to 25 layers at once and deliver blending ratios of 1:2 to 1:5 — which is exactly √4 to √25.
End cones. A longitudinal pile has a cone at each end where the layer structure breaks down. Everything reclaimed from a cone is poorly blended, and the shorter the pile, the larger the cones are as a fraction of it.
Layers thinner than the rock. This is the non-obvious one. If you keep raising the layer count, layer thickness eventually falls below the top particle size — and past that point more layers make blending worse, because coarse particles roll down the pile face and segregate at the ends. There is a genuine optimum, not a monotonic improvement.
Layers are not independent. The √n law assumes independence. Consecutive layers come from the same quarry face minutes apart and are correlated, which flattens the curve well before geometry does.
Partial reclaim. A reclaimer working the upper part of the face cuts fewer layers than the pile contains. This is the failure that costs the most and shows the least — the design ratio is on the drawing, the achieved ratio is not measured anywhere.
§ 04 What must be true
Pre-blending has no symptom of its own, so it needs conditions stated explicitly — otherwise there is nothing to check. These five are geometry and statistics. They have not changed since the first blending bed and they will not.
| What must be true | Why | Target | How you verify it |
|---|---|---|---|
| 1. The reclaimer cuts the full cross-section | Blending is the square root of the number of layers actually cut. A machine working the upper face is averaging a fraction of the layers the pile contains. | full face | Watch it for ten minutes, then check the harrow condition and ask whether any standing instruction limits the reclaim depth. This is the single most common failure and it costs nothing to fix. |
| 2. Enough layers, but not too many | Ratio rises as √n, so the first hundred layers are worth far more than the next three hundred. But once layer thickness falls below the top particle size, coarse rock rolls down the face and end-cone segregation gets worse. | ~100 layers thickness > top size |
Pile length × stacker passes, computed, not taken from a drawing. Then layer thickness against the crusher product top size. There is a genuine optimum, not a monotonic improvement. |
| 3. The pile is long enough that the end cones are a small fraction of it | Layer structure breaks down in the cones at each end, and everything reclaimed from a cone is poorly blended. | — | Cone volume against total pile volume. A short pile at a high stacking rate is paying twice: fewer layers and proportionally more cone. |
| 4. Stacking and reclaiming do not happen on the same pile | Reclaiming a pile still being built means cutting an incomplete cross-section — the layers above the current face do not exist yet. | two piles, alternating |
Ask how many beds there are and how they are used. A plant with two beds running them as one large one has quietly halved its ratio. |
| 5. The achieved ratio is measured, not assumed | The design ratio is on a commissioning drawing. Nothing since has checked it, and nothing in the plant will tell you when it falls. | in vs out, a fortnight |
Two sample sets, in and out, hourly, for two weeks. Until this exists every argument about this section is an opinion — including the argument for spending money on it. |
The whole section is about 2% of plant power, so none of this is an energy argument. It is a variance argument, and F6 §03 priced variance: the kiln’s fuel rate is set by its worst hour, so every point of deviation removed here is heat rate that nobody has to buy equipment for. Rows 1, 4 and 5 are free. Row 2 costs a stacker speed setting. That is the best ratio of value to capital anywhere in the plant, and it is the reason this lesson sits where it does.
§ 05 The quantitative anatomy — the variance cascade
Four stages, each with a blending ratio, and the ratios multiply. That single fact governs where it is worth spending money.
| Stage | Blending ratio | Source and basis |
|---|---|---|
| Simple storage, no designed layering | 2–5× | 4 to 25 layers reclaimed simultaneously. This is a heap, not a blending bed. |
| Designed prehomogenisation pile | up to 10× | The practical ceiling for chevron and windrow systems, against a theoretical √n that would allow far more. |
| Continuous blending silo, one silo | 5–7× | The modern standard. Air-fluidised, continuously drawn. |
| Continuous blending, two silos | 7–12× | Better, and roughly twice the capital. |
| Gravity blending silo, well designed | up to 15× | The high end of what silo technology delivers. |
And the power, which is not the point but gets asked about
| Item | kWh/t | Note |
|---|---|---|
| Primary crushing, general estimate | 0.5–1.0 | Of material processed. |
| Primary jaw crusher, range by size | 0.21–1.38 | Large crushers at the low end, small ones at the high end — a 6.6× spread driven by machine size, not by operating skill. |
| Best-in-class crusher, Indian benchmark | 0.57 | Per tonne of limestone. |
| Whole crushing and stacking section | 1–2 | Per tonne of raw meal, including conveying and the stacker-reclaimer. |
Hold that against the plant total of 70–80 kWh/t of cement. The entire crushing and pre-blending section is about 2% of plant power. If you are here to save electricity, you are in the wrong building. You are here because this section decides how hard the kiln has to burn, and that is worth an order of magnitude more than its own power bill. A consultant who argues this section on kWh/t will lose to a more visible project, and will deserve to.
§ 06 What goes wrong here
| The failure | The tell, and the usual misdiagnosis |
|---|---|
| The reclaimer is not cutting the full face. Worn harrow, conservative operating practice, or a machine set up to maximise throughput rather than cross-section. | Tell: compare the LSF standard deviation entering the pile against the deviation leaving it. If the ratio is under 4, the pile is a storage heap. Usual misdiagnosis: kiln instability, treated at the kiln. |
| The pile is too short for the stacking rate. Layer count falls, end cones become a large fraction of the pile. | Tell: ask for the layer count — pile length × stacker passes, not a design figure. Most plants cannot produce it. Usual misdiagnosis: “the deposit has become more variable”, which may also be true and is not the whole story. |
| Only one pile, so it is stacked and reclaimed at the same time. | Reclaiming a pile still being built means cutting an incomplete cross-section. Two piles alternating is the design intent, and a plant running one pile continuously has quietly halved its blending ratio. |
| Nobody measures the ratio. The design blending ratio is on a drawing from commissioning. The achieved ratio is measured nowhere. | This is the root failure and it is free to fix: two sets of samples, in and out, over a fortnight. Until it is measured, every argument about this section is an opinion. |
| Chasing layer count past the optimum. | If layer thickness falls below the top particle size, end-cone segregation increases. More is not monotonically better, and a vendor proposing a faster stacker should be asked what layer thickness results. |
| Proposing a new blending silo to fix an upstream problem. | The ratios multiply, so a silo bought to compensate for a bad pile has to be extraordinary. It is also tens of crores against a stacker-speed change. Proposing it is a reliable way to be told that the real problem was the reclaimer. |
§ 07 Scenarios from the field
Reddipalayam needed a secondary shredder. The quoted price was ₹42.84 lakh. The plant built its own for ₹11.3 lakh — ₹31.5 lakh avoided on one machine.
Two things to take from this. First, it is not an energy project at all, and it appears in the plant’s CII energy award submission because that is where the improvement register lives — which is exactly why F6 §07 tells you to ask for the register rather than the improvement plan. Second, the in-house fabrication capability that produced it is a real and unpriced asset. A plant with a working fabrication shop has a different opportunity set from one without, and it is worth establishing which you are dealing with on day one.
The upgrade case here is almost entirely operating discipline, crusher hammers, screens and reclaimer maintenance inside the annual shutdown window. The capital involved is small — which sounds like an advantage and is not.
A small-capital project has to be justified on downstream benefit, and downstream benefit is contested. The kiln team will not credit the stockpile for a heat-rate improvement, because they cannot see the mechanism from where they stand. So the business case gets made on crusher kWh/t — where the whole section is 2% of plant power — and loses to something visible.
The fix is measurement, not argument. Sample in and out for a fortnight, establish the achieved blending ratio, and show what the kiln-feed deviation would be at a ratio of 8 instead of 3. That converts an opinion into a number that F6 §03’s arithmetic can price. Nothing else works.
§ 08 The numbers that matter
| Metric | Reference | Basis | Your plant |
|---|---|---|---|
| Blending ratio, designed pile | 5–10× | Practical range. Theoretical is √(layers reclaimed). | — |
| Blending ratio, storage heap | 2–5× | 4–25 layers reclaimed. If your pile measures here, it is not a blending bed. | — |
| Blending silo, one / two | 5–7 / 7–12× | Continuous air-fluidised. Gravity systems up to 15×. | — |
| Crusher product size | 25–75 mm | From up to 1 m run-of-mine. Reduction ratio ~20:1. | — |
| Whole-chain reduction ratio | 5,000–10,000:1 | Quarry rock to raw meal fineness. | — |
| Primary crushing power | 0.5–1.0 kWh/t | Jaw crushers 0.21 (large) to 1.38 (small). Indian best-in-class 0.57 per tonne of limestone. | — |
| Crushing + stacking section | 1–2 kWh/t | Per tonne of raw meal. About 2% of plant power. | — |
| Kiln feed LSF standard deviation | <1 | The target the whole chain exists to hit (A5). | — |
What is not published. There is no benchmark for achieved blending ratio in Indian cement, because almost nobody measures it. The √n relationship and the 2–5 / up-to-10 ranges are from the standard process literature and are sound; the quarry-face standard deviation of 12 used in the diagram above is illustrative, not a benchmark — it is entirely deposit-specific and your client’s number may be a third of it or double it. Get it from their lab before using any of this quantitatively.
§ 09 Build the model — the variance chain, and what each stage is worth
How to use it in a room. Set the quarry deviation and the target from the client’s own lab data, then ask them for the pile ratio. They will not have it. That is the finding — and the two weeks of sampling needed to get it is the cheapest recommendation in this lesson.
§ 10 Upgrade paths
| Tier | Intervention | What it preserves, and the honest assessment |
|---|---|---|
| 0 | Measure the achieved blending ratio | Two sample sets, in and out, over a fortnight. Costs lab time. Every other item on this list is unarguable without it and unnecessary if it comes back at 8. |
| 0 | Reclaim the full cross-section | An operating-practice change. Free, immediate, reversible, and frequently the entire gap. Check the harrow condition and whether the operator has been instructed to prioritise throughput. |
| 0 | Run two piles in alternation | Stack one, reclaim the other. If the plant has two beds and uses them as one big one, this is free and material. |
| 1 | Raise the layer count — to a point | A stacker travel-speed change. Watch the layer thickness against top particle size: past that point it makes segregation worse. Diminishing returns are steep (§03). |
| 1 | Crusher hammers, screens, reclaimer harrow | Normal shutdown work. Restores design performance rather than adding any. The right framing is deferred maintenance, not improvement — and it belongs on the deferred list F6 §10 asks for. |
| 2 | Cross-belt analyser on the crusher discharge | Properly an A3 item, because its value is in raw-mix control rather than pre-blending. But it is also the instrument that lets you see quarry variability in real time and stack against it, which turns the pile from a passive filter into a controlled one. |
| 3 | Extend the pile / add a bed | Civil work and a shutdown. Justified only when the layer count is genuinely constrained by pile length and Tier 0 has been exhausted. Verify Tier 0 first — the answer is usually there. |
| 4 | New blending silo | Tens of crores to buy a ratio of 5–7 when the pile can be moved from 3 to 8 for the price of a shift instruction. Almost never the right answer, and proposing it before Tier 0 is how you get told the real problem was the reclaimer. |
§ 11 The frontier, and who is running it
This is the thinnest frontier section in the course, and that is the finding. No genuinely new stacker-reclaimer or pre-blending technology could be identified anywhere in the world since 2020. The area is mature. What has moved is the instrument in front of it — using an analyser to control stacking rather than merely to record it.
| Status | What | Plant, company, country | The number |
|---|---|---|---|
| OPERATING | Analyser controlling the pile | Al-Rashadiya — Jordan Cement, Jordan. Two cross-belt analysers on the limestone and marl feed to the pre-blend stockpiles, updating once a minute. | Kiln Feed Uniformity Index brought inside the group standard with visibly lower variation. Raw material cost down 0.08 JD/t; cement-to-clinker ratio 1.14 → 1.16. Peer-reviewed, 2011. The benefit landed on the clinker factor — two machines downstream of where the money was spent. |
| OPERATING | Analyser on the quarry crusher | Suwannee American Cement (Votorantim), Florida, USA. Sodern CNA, raw mill feed 2003, quarry crusher discharge 2007. | About 98% availability sustained for nearly twenty years. Lower LSF variability and faster turnaround than lab sampling. |
| OPERATING | The earliest case of controlling, not measuring | Saint-Pierre-la-Cour — Lafarge, now Holcim, France. | On-line neutronic analyser used specifically for control of the pre-homogenisation piles — the first named example of an analyser steering the stacker rather than reporting on it. The same plant is now Holcim’s calcined clay flagship. |
| CONVENTIONAL | Large circular store | Jizzak — Huaxin Cement, Uzbekistan. Bedeschi circular store, 80 m diameter. | 300 tph stacking, 150 tph lateral scraper reclaim, on a 5,000 tpd line. Included to make the point that it is conventional — the novelty is the application, not the machine. Ordered 2023. |
A mature area with no technology frontier means the entire available improvement is operating discipline and instrumentation, which is exactly what §10 says. That is good news for a consultant and bad news for a vendor, and it is worth saying out loud in a meeting where someone is proposing capital here.
It also means something sharper. If this section of a plant is underperforming, there is no technology to buy that fixes it. The Al-Rashadiya and Suwannee cases both bought an instrument, not a machine, and in both cases the measured benefit appeared downstream — on clinker factor and on raw meal variability. Write the case that way or it will not survive the capital committee.
Two gaps worth knowing: no benchmark exists for achieved blending ratio at a named plant, and no named cement plant has published a crusher specific-energy figure. Section-level crushing kWh/t appears only in aggregated benchmark publications.
§ 12 What’s changing now
Current as of August 2026 · refresh every six monthsThere is no structural technology change in crushing and pre-blending. What is changing is the input: as newly auctioned blocks come on stream (A1 §12) with more variable and generally lower grade, pre-blending capacity sized for the original deposit becomes the binding constraint on chemistry.
What it means: a pile that was adequate for twenty years can become inadequate without anything about it changing. Ask when the pile was sized and against which deposit — and whether the quarry’s incoming deviation has been measured since the new block came on.Online elemental analysis on the crusher discharge is now standard specification on new Indian lines. CII puts the direct energy effect at −3 to −5 kcal/kg — but that understates it, because the instrument’s real value is as the enabling measurement for tighter control (A3 §10).
What it means: a plant with an analyser can stack against quarry variability instead of merely averaging it. That converts pre-blending from a passive filter into a control loop, and it changes what the pile is capable of. Write the business case on control, not on kcal/kg.§ 13 Check yourself
§ 14 Mini case — the silo that was not needed
A 2.2 Mt/yr plant with chronic kiln instability. Free lime deviation high, fuel rate defensive, refractory life short. The plant had proposed a second blending silo at ₹28 crore, arguing that its existing single silo could not hold kiln feed deviation below 1.0.
The number nobody had. The silo was rated at 6:1 and there was no reason to doubt it. But when asked what deviation was entering the silo, the plant gave the raw-mill outlet figure — measured, as F6 warns, on the wrong side of the process. Nobody had sampled the pre-blend pile in and out.
Two weeks of sampling. LSF deviation entering the pile: 11.4. Leaving it: 3.7. An achieved blending ratio of 3.1, against a design figure of 8. The reclaimer harrow was worn and the operator had for two years been instructed to reclaim from the upper face to protect throughput during a period of high kiln demand. The instruction had outlived the demand.
The arithmetic. At a pile ratio of 3.1, with the mill at 1.5 and the silo at 6, kiln feed deviation is 11.4 ÷ 27.9 = 0.41. So the silo was not the constraint — on paper the chain already met target. What the plant was actually experiencing was the pile’s poor blending arriving as long-period drift the silo could not filter, which shows up as excursions rather than as a raised average deviation. Standard deviation alone had hidden the problem, because the silo smooths short-period noise and does nothing about a slow wander.
What was done. The harrow was replaced in the next shutdown and the reclaim instruction rescinded. Pile ratio went to 7.2. Cost: the harrow, and a conversation. The ₹28 crore silo was withdrawn.
A composite scenario. The blending ratios, the √n relationship and the silo ratings are from the process literature; the plant is not a real one.
The transferable move. When someone proposes capital to fix variability, ask for the measured ratio at every stage upstream of it first. The ratios multiply, so the constraint is almost never where the proposal is aimed — and a stage nobody has measured is where you should look before a stage everybody has.
§ 15 Go deeper
§ 16 Carry forward
- “Blending is a square root. Your pile’s ratio is roughly the square root of the number of layers the reclaimer actually cuts — so the question isn’t what the machine is rated at, it’s what face it’s working.”
- “The ratios multiply. Pile times mill times silo. Which means a silo bought to fix a bad pile has to be extraordinary, and they aren’t.”
- “This whole section is about two percent of your power. It’s not an electricity conversation — it decides how hard the kiln has to burn, and that’s worth ten times more.”
- “What’s your achieved blending ratio, in against out? If nobody’s measured it, that’s two weeks of lab time and it’s the first thing I’d do.”