The raw mill
One machine grinds, dries and classifies at once, using waste heat from a kiln two hundred metres away. Understanding where its power actually goes is what lets you tell a mill problem from a quarry problem from a kiln problem — three things that present identically.
On this page · 16 sections
- 01The one idea
- 02The parts — and where the power goes
- 03First principles — the drying duty nobody costs
- 04What must be true
- 05The quantitative anatomy
- 06What goes wrong here
- 07Scenarios from the field — two plants, the same problem, two methods
- 08The numbers that matter
- 09Build the model — the raw-mill power and heat bridge
- 10Upgrade paths
- 11The frontier, and who is running it
- 12What’s changing now
- 13Check yourself
- 14Mini case — the boiler that was sized on the dry season
- 15Go deeper
- 16Carry forward
§ 01 The one idea
A vertical roller mill is usually described as a grinding machine. It is at least as much a dryer, and the drying is the part that couples it to the rest of the plant.
Grinding is a local problem: rollers, table, grinding pressure, fineness. Drying is not. The heat comes from preheater exhaust gas, which is the same gas a waste heat recovery system wants, and the same gas whose temperature tells you how the preheater is running. That single connection is why a raw-mill question so often turns out to be a kiln question, a quarry question or a power-project question wearing a mill’s clothes.
A mill that will not make throughput can be a mill problem (worn rollers, wrong grinding pressure, a separator past its design), a quarry problem (harder rock, more quartz, wetter feed) or a kiln problem (exhaust gas too cool or too little of it to dry the feed). All three present as “the mill is short of capacity”, and the fix is completely different in each case. The power bridge in §09 separates them.
§ 02 The parts — and where the power goes
| Component | kWh/t raw meal | What moves it |
|---|---|---|
| Mill main motor | 10–14 | Grindability of the feed, fineness target, grinding pressure, roller and table wear, bed depth and stability. |
| Mill fan | 5–7 | System resistance and gas volume. Rises with false air, with duct restrictions, and with the extra gas needed to dry wetter feed. |
| Separator | ~2 | Fineness target and separator generation. A third-generation separator does the same job with a sharper cut. |
| Transport and auxiliaries | ~1 | Feeders, bucket elevators, airslides, rotary valves. |
| Section total | 18–24 | Per tonne of raw meal. At 1.55 t of raw meal per tonne of clinker and a 0.68 clinker factor, that is roughly 19–25 kWh per tonne of cement — between a quarter and a third of plant power. |
§ 03 First principles — the drying duty nobody costs
Evaporating water is expensive and the number is fixed by physics. The latent heat of vaporisation of water is 539 kcal per kg, and heating that water from ambient to boiling adds about 75 more. Call it 614 kcal for every kilogram of water removed.
Now scale it. A tonne of raw meal at 6% moisture carries 60 kg of water, so drying it needs 36,840 kcal per tonne of raw meal. At 1.55 tonnes of raw meal per tonne of clinker that is 57 kcal per kilogram of clinker — a number of the same order as an entire multi-year energy programme, spent invisibly on evaporation.
WHRS proposals are normally sized on preheater exhaust temperature and flow. If the mill’s drying duty is not deducted, the boiler is oversized and the payback is fiction. And the deduction is not a constant: it moves with monsoon moisture, with which part of the quarry is being worked, and with mill availability.
Three questions to ask of any WHRS proposal: what feed moisture was assumed, over what period, and what mill running-hour percentage? A proposal built on dry-season moisture and 100% mill availability is a proposal built on the best fortnight of the year.
And the reverse also holds. A plant that reduces feed moisture — covered stockpiles, better drainage, a covered conveyor — is not only saving mill fan power. It is freeing preheater exhaust heat for power generation, and that second benefit is usually larger and almost never counted.
The electrical side of moisture is smaller but real: each additional percentage point of feed moisture costs roughly 1–2 kWh per tonne, because more gas has to be moved to carry the water away, and the fan pays for it.
§ 04 What must be true
A raw mill does three jobs at once — it grinds, it dries and it classifies — and it does them inside a gas circuit it shares with the kiln. Almost every raw mill problem presented as a grinding problem is one of the other two. These six conditions do not depend on the mill brand, the country or the decade.
| What must be true | Why | Target | How you verify it |
|---|---|---|---|
| 1. The bed is stable | A vertical mill grinds a bed, not particles. Too thin and the rollers hit the table — vibration, then a trip. Too thick and the rollers ride on uncrushed material. Everything else in the mill is downstream of bed stability. | vibration <1.0 mm/s |
Trend mill vibration against feed rate and water injection for one week. A mill that trips on vibration more than twice a shift is not a grinding problem — it is a feed-consistency or dam-ring problem. |
| 2. The gas circuit is closed | The mill fan moves the whole circuit. False air leaking into the mill, the cyclone or the duct is gas the fan pays for and the mill never uses. The fan is roughly 37% of raw mill power at best-practice plants, so this is the largest single lever on the electrical side. | O₂ rise across the circuit minimal |
O₂ at the mill inlet against O₂ at the fan inlet, and the fan’s absorbed kW against its design curve. A fan running far from its design point is almost always being asked to move air nobody ordered. |
| 3. There is enough heat to dry the feed, and no more | Water must leave the material before the material will grind. Drying normally runs on preheater exit gas or cooler vent air, which is free. The moment a hot-gas generator fires, the plant is buying fuel to dry rock — and that cost appears in kcal/kg, not kWh/t. | feed 3–8% H₂O mill exit ~0.5% |
Feed moisture by component, across a monsoon and a dry month. Then ask how many hours the hot-gas generator ran. A plant that cannot answer the second question is not measuring its real drying cost. |
| 4. The classifier returns the right material, not merely a lot of material | Circulating load is not a virtue. Over-grinding costs power and buys nothing, because the kiln is burning a coarse fraction, not a mean. Burnability is set by the quartz above 45 µm (A3 §06), which a residue on 90 µm does not see. | 10–15% R 90 µm and a 45 µm check |
Ask for the full particle size distribution, not the residue. Two mills at the same R 90 µm can differ by several points of free lime because one has a tail of coarse silica and the other does not. |
| 5. The mill is fed what it was designed for | Capacity is set by grindability, moisture and product fineness — not by the mill. Installed drive power across one mill model varies by nearly two to one for exactly this reason. | — | Compare the current feed’s Bond index and moisture against the commissioning guarantee. A mill that will not make its nameplate is often meeting the spec it was sold against and failing the rock it now gets. |
| 6. The wear parts are on a planned cycle, not an event cycle | Roller and table liner wear raises specific power continuously and invisibly before it causes a stoppage. The cost shows as a creeping kWh/t, which is attributed to everything else. | — | Plot mill kWh/t against hours since the last reline, over two or three cycles. The slope is the real wear cost and no plant publishes it — it has to be built from the client’s own spares ledger. |
Rows 1 to 3 are conditions for the mill to run at all. Rows 4 to 6 are conditions for it to run well. A consultant who starts at row 4 on a mill that is failing row 1 will produce a correct recommendation that cannot be implemented, because the mill will not hold a steady enough operating point to show the benefit. The sequence is not a preference; it is the dependency.
Note also what is not on this list: the mill drive. Drive efficiency is real but it is fixed at purchase and it is a small number against rows 2 and 3. There is no gearless mill drive on a cement raw mill anywhere in the world, and a conversation that starts there has started in the wrong place.
§ 05 The quantitative anatomy
| Metric | Typical | Note |
|---|---|---|
| Raw meal fineness | 12–15% R90 | Residue on a 90-micron sieve. Finer costs mill power and buys burnability — the direct trade against A3. |
| Feed moisture in | 5–10% | Deposit- and season-dependent. The single most under-monitored input in this section. |
| Product moisture out | 0.5–1.0% | Higher than this and the meal will not flow in the silo or the preheater. |
| Drying heat | 614 kcal/kg water | 539 latent plus about 75 sensible. Physics, not a benchmark. |
| Drying duty at 6% moisture | 57 kcal/kg cl | Derived. Roughly half the recoverable preheater exhaust heat. |
| Moisture penalty, electrical | 1–2 kWh/t per % | Fan power, from moving more gas. |
| Mill running hours | 85–92% of kiln hours | The mill must make more than the kiln eats, in less time. Ask for this number — a mill at 95% has no headroom for a stoppage. |
A widely circulated Indian best-in-class figure for the raw mill is 10.43–10.64 kWh per tonne of raw meal. The component breakdown in §02 sums to 18–24. Both cannot be the same measurement.
Almost certainly the 10.5 figure is the mill main motor only and the 18–24 range is the whole department including fan, separator and transport. But the source does not say, and this is exactly the basis trap F4 §06 is about. Until the boundary is stated, the two numbers are not comparable and neither should be used to benchmark a client.
What to do instead: ask the plant for its raw-mill section consumption and the measurement boundary — motor, MCC or section incomer — and compare it against its own history rather than against a figure whose definition you cannot verify.
§ 06 What goes wrong here
| The failure | The tell, and the usual misdiagnosis |
|---|---|
| Moisture is the variable that hides. It is seasonal, it is rarely trended, and its cost is thermal rather than electrical — so it does not appear where anyone is looking. | Tell: ask for twelve months of feed moisture. Most plants have spot values and no trend. Usual misdiagnosis: “mill capacity has fallen”, treated as a grinding problem. |
| The fan is in the wrong duct. Inlet-box geometry, bad transitions, restrictive dampers — resistance that has nothing to do with the fan itself. | Tell: a pressure survey across the circuit. 30–40 mmWC of avoidable drop is common and worth about 0.5 kWh/t. Usual misdiagnosis: the fan is undersized, so a bigger fan is proposed (§07). |
| Grinding finer than the kiln needs. R90 held at 10% when 14% would burn perfectly well, because nobody has tested it. | Every point of R90 costs mill power. The right target depends on the quartz size distribution (A3 §06), not on a standard. Two months of controlled trial settles it and almost nobody runs one. |
| False air in the mill circuit. Leaking rotary valves, expansion joints, inspection doors. | Dilutes and cools the drying gas, so the fan moves more of it and the mill dries less well. It shows up as both a power problem and a capacity problem, which is why it is misread as either. |
| Sizing a WHRS boiler without deducting the drying duty. | The proposal assumes exhaust heat the mill is already using. Payback is fiction and the boiler underperforms from commissioning, at which point it is blamed on the kiln. |
| Benchmarking against a figure with no stated boundary. | See the box in §05. 10.5 and 21 are not the same measurement, and a plant told it should be at 10.5 will be told by its own engineers that the number is nonsense — correctly. |
§ 07 Scenarios from the field — two plants, the same problem, two methods
Maratha enlarged the raw-mill fan inlet box and redesigned the damper, taking 30–40 mmWC out of the circuit. The results: 0.5 kWh per tonne of material, feed rate up from 580 to 625 tph (+7.8%), and fan power slightly down from 3,150 to 3,100 kW. Cost ₹8.75 lakh, payback three months.
Read the third number carefully. The fan moved more gas on less power. That is not possible if the fan was the constraint; it is only possible if the duct was. The fan had been fighting a restriction someone built in at commissioning, and had been fighting it for years.
Reddipalayam did the same job with computational fluid dynamics rather than by inspection. The inlet box went from 3.8 m² to 5.1 m², gas velocity from 28 m/s to 21 m/s, and fan efficiency from 71% to 74%.
Check the physics yourself: mass continuity says velocity should fall by the area ratio, and 28 × 3.8 ÷ 5.1 = 20.9 m/s. The reported figures are internally consistent to within a tenth, which is a good sign that the numbers are measured rather than modelled optimistically.
Neither fan was the wrong fan. Both were in the wrong duct. A fan is a machine that trades pressure for volume against a system curve, and the system curve is drawn by the ductwork. When a plant tells you the mill fan is undersized, the first question is not what the fan is rated at — it is where the pressure drop is, and whether anyone has surveyed it since commissioning.
This is also a cheap class of work: ₹8.75 lakh and a three-month payback at Maratha, against a new fan at ten to twenty times that. And it is a class of work that only gets found by measurement, which is why it sits in the CII registers and not in capital plans.
§ 08 The numbers that matter
| Metric | Reference | Basis — state it every time | Your plant |
|---|---|---|---|
| Raw mill section power | 18–24 kWh/t | Per tonne of raw meal, whole section. See the basis warning in §05. | — |
| Mill main motor only | 10–14 kWh/t | Motor basis. Not comparable to a section figure. | — |
| Mill fan | 5–7 kWh/t | Nearly a third of the section, and the part that responds to ductwork. | — |
| Raw meal fineness | 12–15% R90 | Residue on 90 µm. Trade against burnability (A3). | — |
| Feed moisture | 5–10% | Ask for twelve months, not a spot value. | — |
| Drying duty | 9.6 kcal/kg cl per % moisture | Derived from 614 kcal/kg water and 1.55 t raw meal per t clinker. | — |
| Moisture, electrical penalty | 1–2 kWh/t per % | Fan power. | — |
| Separator upgrade | −5 to −10 kWh/t cement +15–25% output | ECRA. LBNL notes payback exceeds ten years on energy alone. | — |
§ 09 Build the model — the raw-mill power and heat bridge
What the model deliberately does not do. It does not tell you whether a WHRS is viable — that needs the boiler’s pinch temperature, the mill’s actual running hours and a twelve-month moisture profile, none of which is a single number. It tells you whether the heat a proposal assumes is already spoken for, which is the question that most often goes unasked. Exhaust heat is computed as gas volume × 0.33 kcal/Nm³/°C × (exhaust temperature − 100 °C).
§ 10 Upgrade paths
| Tier | Intervention | What it preserves, and the honest assessment |
|---|---|---|
| 0 | Trend feed moisture for twelve months | Free. It is the input that hides, and everything on this list is mis-sized without it — including any WHRS proposal. |
| 0 | Test a coarser R90 | A controlled trial, two months. If the kiln burns 14% R90 as well as 12%, the mill power saved is permanent and free. The answer depends on quartz size, so pair it with A3’s petrographic work. |
| 1 | Pressure survey and duct rework | The best-evidenced item in this lesson. Maratha: ₹8.75 lakh, three-month payback, 0.5 kWh/t and 7.8% more throughput. Preserves the fan, the mill and the circuit — it changes geometry, not machinery. |
| 1 | False air survey and sealing | Normal shutdown work. Cuts fan load and improves drying at the same time. |
| 2 | Cover the stockpile and conveyors | Civil work, no process risk. Cuts moisture, which cuts fan power and frees preheater exhaust for WHRS. The second benefit is usually larger and almost never counted. |
| 3 | Separator upgrade | ECRA: −5 to −10 kWh/t of cement with 15–25% more mill output. LBNL notes payback exceeds ten years on energy alone — so the case is made on throughput, and a consultant who argues it on kWh/t will lose. It wins when the mill is the plant’s bottleneck, and loses when it is not. |
| 4 | New mill | Only with a capacity project. Modern VRMs are being specified at very high power density — Gebr. Pfeiffer’s MVR 6000 R-6, ordered for JK Cement’s Jaisalmer greenfield in December 2025, is 1,200 tph at 7,400 kW. That is a layout decision as much as an efficiency one. |
§ 11 The frontier, and who is running it
Raw milling is a mature technology with an unusually empty public record. The machines have got larger and the control has not moved at all. No plant anywhere publishes its raw mill specific power consumption under its own name — every number in circulation comes from an anonymised benchmarking pool, a vendor acceptance test or a vendor design value. Knowing which of the three you are being handed is most of the skill in this section.
| Status | What | Plant, company, country | The number |
|---|---|---|---|
| ORDERED | The largest raw mill in the world | Jaisalmer — JK Cement, Rajasthan, India. Gebr. Pfeiffer MVR 6000 R-6, SLS 6300 VR classifier. Greenfield; commissioning stated as end 2026. | 1,200 t/h raw meal, 7,400 kW drive. Announced December 2025. Read the fineness before you use the capacity: it is quoted at 1.5% residue on 212 µm, not the conventional 90 µm. A t/h at 212 µm is not comparable with a t/h at 90 µm, and most “largest mill” comparisons ignore this. |
| ORDERED | The same mill, the same district, a different number | Jaisalmer — Wonder Cement, Rajasthan, India. Pfeiffer MVR 6000 R-6, SLS 6000 VR. Ordered December 2025, sixteen days after JK’s. | 1,000 t/h, 6,725 kW. Fineness not disclosed. Two identical mills on the same limestone, ordered a fortnight apart, rated 20% apart. That gap is the lesson: the mill does not set the capacity. |
| OPERATING | Achieved against guaranteed — the only such pair published | Touk Meas — Chip Mong INSEE Cement, Kampot, Cambodia. Pfeiffer MVR 5000 R-4, conventional gearbox. Commissioned October 2017. | Acceptance test 453 t/h achieved against 410 t/h guaranteed, at 12.6% R 90 µm against 15% guaranteed; vibration 0.5 mm/s. A finer product and 10% more of it than the contract required. Vendor-reported, but it is the only place in the public record where a named raw mill’s achieved figure sits next to its guaranteed one. |
| OPERATING | Roller press instead of a vertical mill | Siam City Cement (INSEE), Thailand. KHD Comflex, replacing two ball mill circuits. 72-hour performance test, early 2020. | 350 t/h at 13.36 kWh/t raw meal. The vendor describes the saving as “more than 35%” in one sentence and “almost 40%” in another, in the same release. Use the absolute. Distrust the percentage. 13.36 sits between CII’s roller-press best and its tenth-best vertical mill, which is the sanity check that makes it credible. |
| OPERATING | Benchmark, plants anonymous | CII Energy Benchmarking for the Indian Cement Industry, Version 6.0, May 2023, more than 110 plants. | Raw mill VRM circuit best 10.64 kWh/t raw meal; top-ten band 10.64–14.40. Best plant split: mill drive 5.1, mill fan 3.9, auxiliaries 1.6. About 20% better than the same table in 2014. The plants are “Plant 1” to “Plant 10”. This is the best public raw-mill data in the world and it names nobody. |
| ANNOUNCEMENT | Analyser and control software on a raw mill | Turaif — Northern Region Cement, Saudi Arabia. SpectraFlow airslide analyser with raw-mix control software, on the white cement line. Reported August 2026. | No performance figure of any kind. The only number in the announcement is the line’s 0.85 Mt/yr capacity. Worth carrying as the worked example of an announcement that reads like evidence and is not. |
| NOT A THING | Gearless mill drives on raw mills | — | ABB’s own reference literature names one cement gearless drive ever: a 6.4 MW unit on a cement mill at Le Havre in 1969. Every other gearless mill drive is a mining SAG or ball mill. If a vendor or a client raises gearless drives for a raw mill, they have imported a mining concept. |
1. No named-plant raw mill kWh/t, anywhere. If you quote “best in class is 10.6”, you are quoting an anonymous Indian plant in a 2023 CII table, and you should say so in the same breath.
2. No named-plant classifier or separator upgrade result on a raw mill. These are sold constantly and documented never. Demand the plant’s own mill fan kW and product residue before and after.
3. No raw mill availability percentage or cost-of-wear figure at any named plant. Build it from the spares ledger.
4. No named plant publishes feed moisture alongside the drying heat penalty — the number that decides whether a hot-gas generator is a cost or a convenience. Vendors quote a moisture capability; none quotes its cost.
5. No named-plant AI or advanced-control result on the raw mill, closed loop or advisory, despite plentiful kiln and finish-mill cases. ABB’s “around 6% less energy” for grinding optimisation carries no plant, no date and no measurement basis. The Tokuyama Nanyo result frequently cited here — +3% throughput, −3% specific power — is on finish mills, not the raw mill.
Installed drive power per t/h of raw meal, one mill family, four plants. The arithmetic is ours; the inputs are the published orders above.
| Plant | Mill | kW / (t/h) | Fineness basis |
|---|---|---|---|
| Al Abraj Cement, Libya | MVR 3750 R-4 | 8.5 | 10% R 90 µm |
| Shree Cement, Raipur, India | MVR 6000 R-6 | 10.9 | not stated |
| Wonder Cement, Jaisalmer | MVR 6000 R-6 | 6.7 | not stated |
| JK Cement, Jaisalmer | MVR 6000 R-6 | 6.2 | 1.5% R 212 µm |
A spread of 1.8 to 1 across one mill model, set by grindability, moisture and product fineness. Against this, CII’s best plant absorbs 5.1 kWh/t on the mill drive. Installed kW/t is a procurement number; absorbed kWh/t is an operating number. Comparing a client’s installed power with somebody else’s absorbed power is the most common unit error in this part of the plant.
§ 12 What’s changing now
Current as of August 2026 · refresh every six monthsGebr. Pfeiffer’s MVR 6000 R-6 was ordered for JK Cement’s Jaisalmer greenfield in December 2025 — 1,200 tph at 7,400 kW, billed as India’s largest raw mill. UltraTech ordered seven VRMs from the same supplier; Ramco ordered three Loesche mills.
What it means: high power density is being chosen deliberately. One very large mill instead of two means less civil work, less ducting and a smaller footprint — but also no redundancy. Ask what the plan is when it stops, because a single-mill plant has a different risk profile from a two-mill plant at the same capacity.The deposits coming on stream under A1 §12 are lower grade and often need more corrective additive — and correctives are frequently the wettest component in the mix.
What it means: a mill sized for the original deposit can lose capacity without anything about the mill changing. Establish whether a capacity complaint is new rock or an old machine before proposing either.§ 13 Check yourself
§ 14 Mini case — the boiler that was sized on the dry season
A 2.6 Mt/yr plant installed a ₹62 crore preheater waste heat recovery system rated at 8.5 MW. In its first full year it averaged 5.9 MW — 69% of rating. The vendor pointed at kiln operation. The plant pointed at the vendor.
What the sizing had assumed. Preheater exhaust at 340 °C and 1.5 Nm³/kg of clinker gives about 119 kcal/kg recoverable to 100 °C. The feasibility study deducted a raw-mill drying duty based on 4.2% feed moisture — a figure taken from three spot samples in February.
What the moisture actually was. Twelve months of belt samples, collected during the review, averaged 7.1%, ranging from 3.8% in March to 11.4% in August. At 7.1% the drying duty is 68 kcal/kg against the assumed 40 — so the boiler had 51 kcal/kg to work with rather than the 79 it was sized on. That ratio is 65%, and the plant was achieving 69%. The machine was performing slightly better than the heat available to it.
Neither party was at fault in the way each believed. The kiln was fine. The boiler was fine. The feasibility study had sized a boiler on the driest fortnight of the year, and the error had been locked into ₹62 crore of steel before anyone with twelve months of moisture data saw it.
What came out of it. Covering the limestone stockpile and the main conveyor cut average moisture to 5.4%, releasing about 16 kcal/kg back to the boiler — roughly 1.1 MW of the missing 2.6, for ₹3.4 crore of civil work. The remainder was written off as a sizing error, which is what it was.
A composite scenario. The 614 kcal/kg water, the 1.55 raw meal ratio and the exhaust-heat arithmetic are physics; the plant is not a real one.
The transferable move. Ask any heat-recovery proposal what feed moisture it assumed, measured over what period. If the answer is a spot value, or a dry-season value, the boiler is oversized and the argument about why will run for years after the money is spent.
§ 15 Go deeper
§ 16 Carry forward
- “Your raw mill is a dryer as much as a grinder. At six percent feed moisture it’s taking about half the heat in your preheater exhaust — which is the same heat your waste-heat project wants.”
- “Six hundred and fourteen kilocalories per kilogram of water. That’s physics. Multiply it by your moisture and your raw meal ratio and you have the drying duty nobody costs.”
- “What feed moisture did the WHRS study assume, and over what period? If it’s a spot value you have an oversized boiler.”
- “Before we look at the fan — where’s the pressure drop? Two Indian plants got seven percent more throughput on less fan power by changing an inlet box. Neither fan was the wrong fan.”
- “A separator upgrade is a throughput argument, not an energy argument. On energy alone LBNL puts the payback past ten years, so if the mill isn’t your bottleneck this project doesn’t pay.”