Classifiers and Separators: Complete Guide
Classifiers and separators divide a grinding-circuit stream into a fine product and a coarser reject stream. Their performance influences product fineness, particle-size distribution, circulating load and specific energy, but those outcomes depend on the complete mill–separator circuit rather than the classifier alone. A useful evaluation therefore combines the separator curve, material balance, mill discharge PSD, airflow, throughput and product quality.
The Complete Cement Technical Package includes 931 cement-industry files covering books, courses, Excel tools and presentations. The $249 one-time package includes this classifier/separator guide with design references, separation-curve examples, operating parameters and performance calculators, with instant download access immediately after payment.
The separator deserves the engineer’s attention because it concentrates leverage: a few percent of the bypass waste, or a blunt separation curve, costs the output and the energy of the whole circuit, while the corrected setting recovers both at zero capital cost: this page follows the file: the principles of the air classification first, the generations of the machines second, the tuning and the troubleshooting last.
1. The Principle of the Air Classification: The Physics of the Fines Split
The air separator is a wind machine: the material falls into a classifying zone where the air stream lifts the fine particles while the coarse particles fall back: the balance of the forces decides the cut: the drag of the air on the particle against the inertia and the gravity of the particle: the fine and light particles follow the air to the product, the coarse and heavy particles leave the air stream to the rejects: the classification is not a sieve: it is a balance of the aerodynamic forces, and that is why the same machine can cut at 30 or 90 micrometers by the simple change of the settings.
- The particle size at the cut: the cut size (d50) is the particle size at which half the mass goes to the fines and half to the rejects: the modern separators cut in the range of 10 to 100 micrometers depending on the rotor speed and the airflow;
- The sharpness: the ideal separator would split the feed at one exact size; the real machines split over a size band, and the sharpness index measures the width of the band: the sharper the curve, the less the fines lost with the rejects and the less the coarse particles leaking to the product;
- The bypass: the non-ideal fraction that reports to the wrong stream—for example, fines returning with rejects or coarse particles escaping with product. The exact definition depends on the Tromp-curve convention used, so the convention must be stated before values are compared.
- The fish-hook effect: the finest particles (below 10 micrometers) sometimes follow the coarse fraction due to the agglomeration and the wall effects: the shape of the real separation curve deviates from the ideal S-curve, and the instruments measure the deviation;
The physics of the classification is the language of the file, and the second section of the file resolves the equations of the particle motion with the worked examples: the settling velocity of the particle in the air, the drag coefficient, the rotor flow field: the engineer who reads the physics section understands why the speed changes cut the curve, and that understanding is the foundation of the tuning that the later sections describe.
| Force | Proportional to | Effect on the cut |
|---|---|---|
| Air drag on the particle | air velocity, particle diameter, density | carries the fines to the product |
| Centrifugal force (rotor type) | rotor speed squared, particle mass | throws the coarse back to the rejects |
| Gravity of the particle | particle mass | settles the coarse in the classifying zone |
Rotor speed, airflow, guide-vane setting, feed loading and particle properties interact to determine the cut and sharpness. Higher rotor speed commonly shifts the cut finer, while the effect of airflow depends on separator geometry and the complete aerodynamic system. Operators should therefore use the OEM performance map and plant response data rather than a universal direction table.
2. The Generations of the Separators: From the Static Cones to the High-Efficiency Rotors
The separator family grew in three generations, and the cement plants of the world still run all three: the generation defines the sharpness, the bypass and the specific power, and the file’s comparison table is the reference of the upgrade decisions:
| Generation | Machine type | Typical classification character | Best comparison basis |
|---|---|---|---|
| First | Static air separator, conical | Limited cut control and generally broader separation | Compare with the unit’s clean baseline and product duty |
| Second | Mechanical dynamic / whizzer type | Adjustable cut with moderate selectivity | Compare Tromp curve, airflow, loading and product quality |
| Third | Dynamic cage rotor with guide vanes | Higher potential selectivity with improved dispersion and control | Compare against OEM design and verified plant performance |
The first generation machines separate by the rising air inside the cone, with the coarse particles falling back through the rising stream: their curves are blunt and their output of fines is limited, but their mechanics are simple and they survive in the small plants: the second generation added the rotating selector blades (the whizzer) that throw the coarse particles outward while the air passes inward: the sharper cut and the higher capacity followed: the third generation enclosed the rotor in the cascade of the guide vanes, creating the uniform classification zone where the feed disperses into the full air curtain: the sharpest curves, the lowest bypass and the compact footprint: the third-generation machines are the standard of the modern mills, and their operating parameters fill the tuning chapters of the file.
3. The Parameters of the Third-Generation Separator: The Complete Setting List
The modern dynamic separator is tuned by a short list of parameters, and the file’s parameter table is the working document of the tuning sessions: each parameter has its typical range, its effect on the separation and its interaction with the circuit:
- The rotor speed: a primary control variable for cut size. The applicable rpm and tip-speed range depend on separator diameter, blade design and OEM limits. Small changes can materially affect the cut, but trial step size should be based on the actual machine response and operating margin.
- The guide-vane position: vanes shape the incoming swirl and flow distribution. The correct angle/opening is separator-specific and should be set from OEM guidance and plant testing rather than one universal percentage range.
- The classification airflow: airflow controls drag, transport capacity and pressure drop. Its effect on cut size depends on rotor speed, separator geometry, guide-vane setting and feed loading, so airflow should be evaluated from measured flow/pressure and the machine’s operating map rather than a universal normalized flow range.
- The feed distribution: the material enters the machine through the feed cone and the distribution plate: the uniform dispersion of the feed into the classification zone is the condition of the sharp curve, and the worn or the mis-set distribution parts are the classic cause of the lost sharpness;
- The solids loading: excessive material loading in the classification zone can reduce dispersion and separation quality. The acceptable loading is design-specific and should be judged from separator feed, airflow, pressure behavior and product/reject PSD.
The interaction table of the file pairs the parameters against each other: the rotor speed and the airflow both move the cut, the vanes and the feed distribution both move the sharpness, and the changes must be trialed one at a time to be understood: the file’s tuning protocol is the experiment design of the classifier: the baseline, the single-variable trials, the verification: the same discipline the whole optimization campaign follows.
4. The Separation Curve: The Signature of the Machine
The separation curve is the fingerprint of the classifier: the plot of the probability of the rejection (or the separation efficiency) against the particle size, measured from the size analyses of the separator feed, the fines and the rejects: the curve tells the engineer everything: the cut size (d50), the sharpness (the ratio of d75 to d25), the bypass (the efficiency at the finest sizes) and the fish-hook (the hump at the coarse end): the file teaches the measurement and the reading of the curve with the worked examples:
| Curve feature | Definition | How to interpret it |
|---|---|---|
| Cut size d50 | size at 50% separation probability | Compare with product duty, feed PSD and OEM/design baseline |
| Sharpness | defined from d25/d50/d75 or another stated method | Use one consistent formula and compare with the same machine baseline |
| Bypass | misplaced fraction at the fine/coarse end, depending on curve convention | State the convention and compare with clean/design performance |
| Fish-hook behavior | non-monotonic fine-size response | Investigate sampling, agglomeration, wall effects and true classification behavior |
The reading of the curve is the diagnosis: the high bypass with the correct d50 points to the seals and the feed distribution; the blunt curve with the low bypass points to the airflow distribution and the vane geometry; the fish-hook points to the dispersion quality of the feed: the file’s case studies show the measured curves of the well-tuned and the degraded machines side by side, so the engineer recognizes the states of his own separator: the curve is the instrument of the tuning, and the plant that measures it quarterly inherits the visible state of its classification.
5. The Circulating Load: The Rhythm of the Closed Circuit
The separator sends rejects back to the mill, creating the circulating load. A common definition is reject flow divided by fresh feed flow; some plants also track a circulation factor equal to separator feed divided by fresh feed. The terminology should be stated explicitly, and the optimum value is circuit-specific rather than a universal percentage.
- The benefits of the circulation: the coarse particles return to the mill again and again until they reach the fineness of the product: the mill always grinds a feed that matches its charge, and the product leaves the circuit only when ready: the closed circuit’s efficiency comes exactly from this screening of the product;
- The costs of circulation: higher recycle increases the duty on transport equipment and the separator and changes the material load seen by the mill. Whether a given recycle level is excessive depends on the circuit design, product target and measured specific energy.
- The measurement: the circulating load is computed from the finenesses of the streams (the algebraic balance of the feed, the fines and the rejects) or measured by the weigh scale of the rejects belt: the file’s calculator derives the load from the sieve analyses without the scale;
- The optimum: the operating optimum of the load sits where the mill and the separator work in balance: the mill under-loaded (the low circulation) leaves the separator starved, and the mill over-loaded (the very high circulation) floods with the fines: the tuning of the rotor speed and the airflow finds the balance point, and the file’s trial examples show the load curve against the specific energy;
The circulating load is the bridge between the classification and the grinding: the separator tuning changes the load, the load changes the mill’s feed size, and the mill’s response changes the separator feed: the circuit is one organism, and the file’s system section shows the full loop with the measured numbers of a real mill: the reader closes the section understanding why the separator speed trial must be judged by the circuit’s total specific energy, and not by the fineness alone.
6. The Tuning Practice: The Trial Matrix and the Verification
The tuning of the separator is a controlled experiment, and the file’s protocol is the one the optimization campaigns use: the trial matrix changes one parameter at a time around the operating point, measures the circuit response at every setting, and selects the setting that satisfies the quality window at the minimum specific energy:
| Trial variable | What to monitor | Decision basis |
|---|---|---|
| Rotor speed | Product PSD, rejects, circulating load, separator power | Use a small OEM-approved step and keep only changes that improve the required product/circuit KPIs |
| Guide-vane setting | Airflow distribution, pressure drop, Tromp curve, vibration | Compare against the validated baseline and machine limits |
| Airflow | Pressure, cut behavior, transport stability, fan power | Change within the aerodynamic operating envelope and judge complete-circuit energy |
| Feed rate | Separator loading, rejects, mill power, product quality | Accept only if the complete circuit remains stable and product specification is maintained |
After a promising setting is identified, verify it over a period long enough to capture normal process variability and representative laboratory samples. Compare Blaine/residue or PSD, circulating load, throughput and specific energy with the baseline at the same product quality. The required duration and sampling frequency are plant-specific.
7. The Classifiers of the Other Circuits: The Raw Meal, the Coal and the VRM
The classifier family serves every grinding circuit of the plant, and the file dedicates its sections to the cousins of the finish-mill machine: the raw meal classifier of the vertical roller mills, the coal classifiers with their explosion safety, and the air separators of the finish mills of the older lines: the principles are the same and the parameters differ:
- The VRM classifiers: the cage rotors inside the roller mill body, integrated with the drying gas: they cut the raw meal at 10 to 12% R90 with the tip speeds and the airflow of the mill’s internal loop: their tuning is the mill’s classifier speed and the gas flow, and their curve is measured with the same sampling of the mill product and the rejects;
- The coal classifiers: coal-mill classification has additional combustible-dust and fire/explosion hazards. Fineness, gas composition, temperature, inerting and protection systems must follow the coal-mill OEM design, the plant’s fuel properties and the site’s approved safety procedures; do not apply generic temperature or oxygen limits from a general separator guide.
- The raw mill ball circuit: the raw meal separators cut at the coarser retention than the finish mills (the raw meal needs 10 to 14% R90 only), and their tuning follows the same trial matrix: the raw circuit’s classification is often the forgotten relative, and the file points to the raw classifiers as the cheap energy gains of the plants that neglected them;
- The grit separators and the de-dusting: beyond the classifying machines, the cyclone arrays and the bag filters separate the powder from the gas: the file covers them as the dust-handling partners of the classifiers, with the pressure drops and the efficiencies that the circuits of the plant depend on;
The classifier knowledge transfers across the circuits: the engineer who masters the finish-mill separator inherits the raw classifier, the coal classifier and the VRM classifier, because the physics, the curve and the tuning protocol are the same family: the file’s structure teaches the family, not the single machine, and the plant benefits from the transferable skill at every grinding station.
8. The Upgrade Decisions: When the Old Separator Must Go
The old first and second-generation separators have their limits, and the file’s upgrade section is the honest economics of the replacements: the decision to replace the separator is an investment decision, and the numbers that drive it are the bypass, the sharpness, the capacity limit and the specific energy of the circuit:
- The capacity ceiling: the old machines reach the aerodynamic limits of their design (the maximum airflow and the fines loading), and the mill circuit cannot exceed the separator capacity no matter how well the mill grinds: the upgrade unlocks the next capacity step;
- The energy penalty: an inefficient separator can increase recirculation, overgrinding and fan duty. Quantify the penalty from measured separator/fan power, circulating load and complete-circuit kWh/t rather than assigning fixed values to separator generations.
- The quality effect: a different separator curve changes the product PSD and coarse tail. Its effect on strength, water demand and workability is cement-specific and must be confirmed by laboratory and concrete-performance testing.
- The payback: calculate payback from the actual installed project cost, verified kWh/t reduction, annual tonnage, electricity price, maintenance impact and any demonstrated production gain. Generic payback periods are not reliable enough for an investment decision.
The upgrade section closes with the retrofit options between the extremes: the modern rotors installed in the old casings, the improved vanes and the dispersion plates: the file’s decision matrix pairs the machine’s measured curve with the appropriate remedy (the tuning, the retrofit or the replacement), so the plant spends the minimum that the measured state requires: the upgrade decision of the file is a numbers document, never a fashion statement.
9. The Monitoring Plan of the Classifiers: The KPIs that Keep the Tuning Alive
The tuned separator drifts back toward the mediocrity unless the plant watches it, and the file’s monitoring section defines the small set of the KPIs that catch the drift in time: the monitoring plan is deliberately light, because the discipline survives only when the weekly effort is small: the plants that measure the four numbers below every week inherit the stable classification, and the file provides the blank monitoring sheet as the template of the routine:
- Bypass and sharpness: repeat the separation-curve test often enough to establish baseline and detect meaningful drift, and after major maintenance or unexplained performance change. The interval should reflect wear rate, test repeatability and production impact.
- Circulating load: trend it at a frequency appropriate to the available flow/PSD measurements. Persistent drift can arise from either mill or separator changes and should trigger a circuit-level diagnosis.
- Specific energy: trend complete-circuit kWh/t using a clearly defined equipment boundary and compare like-for-like products. Rising energy can reflect mill, separator, fan, filter or material changes, so use it as a trigger for diagnosis rather than proof of one cause.
- The product quality, daily: the Blaine and the R45 of the product against the window: the daily numbers release the product and the weekly averages feed the tuning review: the quality is the boundary of the monitoring, as it is of the tuning itself;
The monitoring closes the loop of the file: the tuning campaign sets the operating point, the monitoring holds it, and the quarterly curve review re-tunes the drift: the file’s monitoring section integrates the four KPIs into the mill’s existing reports (the daily log, the weekly energy meeting, the monthly quality review), so the classifier discipline rides on the plant’s routine instead of competing with it: the separator that is measured survives the seasons, and the monitoring plan of the file is the insurance premium of the tuning investment: small, regular and cheap, against the drift that would otherwise cost the output silently.
10. The Troubleshooting of the Separator: The Symptoms and the Causes
The separator wears and misbehaves, and the file’s troubleshooting table is the daily diagnostic of the mill floor: the symptoms of the classification problems are easy to see (the coarse product, the sagging output, the rising load) and their causes are multiple: the table discriminates them with the measurements:
| Symptom | Measurements to take | Likely causes |
|---|---|---|
| Product coarser at constant speed | curve, bypass, rotor clearance | Worn rotor tips, vane wear, sealing bypass up |
| Product finer at constant speed | airflow, fan curve, vane position | Airflow risen, vanes shifted, feed distribution changed |
| Circulating load climbing | stream finenesses, mill outlet residue | Mill grinding worse, separator cut too fine, feed coarser |
| Vibration of the separator | rotor balance, build-up inspection | Rotor imbalance, internal build-up, bearing wear |
| Bypass rising slowly | bypass from the curve | Seal wear, feed cone wear, internal leaks |
| Fan power out of envelope | fan curve, duct pressures | Damper drift, bag filter clogging, duct build-up |
The discipline of the troubleshooting matches the discipline of the tuning: measure first, then act: the file’s full matrix carries the second-level checks (the step tests, the visual inspections at the stops) that confirm every cause, and the remedy table maps each confirmed cause to its repair and its prevention: the separator problems are the most preventable problems of the circuit, because their symptoms appear in the curve weeks before they appear in the output: the plant that measures the curve quarterly inherits the early warning system of its classification.
11. Practical Classifier and Separator Audit Checklist
- Define the product target, separator type and stable operating baseline before changing settings.
- Confirm the Tromp-curve convention and use one consistent definition for bypass, sharpness and circulating load.
- Measure fresh feed, separator feed, product and rejects—or reconstruct the balance from validated stream measurements.
- Record rotor speed, airflow, pressure, feed loading, separator power and product/reject PSD over the same stable period.
- Compare d50, bypass and sharpness with the same machine’s clean/design baseline rather than universal good/bad numbers.
- Inspect feed distribution, guide vanes, rotor condition, seals, airlocks and duct build-up if classification has degraded.
- Test one OEM-approved operating change at a time and allow the circuit and laboratory sample to stabilize before the next change.
- Judge optimization or retrofit value from verified throughput, complete-circuit kWh/t, product quality, maintenance impact and installed cost.
12. Frequently Asked Questions
What is the difference between the classifier and the separator?
In the cement industry the two words name the same family: the machines that split the mill product into the fines and the rejects by the air classification: the “separator” is the classical name of the air machines of the ball mill circuits, and the “classifier” is the modern name of the dynamic rotors (and of the internals of the VRMs): this guide uses both words for the same physics and the same parameter tables, exactly as the plants do.
Why does the separator matter more than the mill for the quality of the cement?
Because the mill grinds whatever the separator returns, while the separator chooses what the product is: the mill can produce the full size distribution of the ground material, but only the separator decides which particles leave as the cement: the sharpness of the curve and the bypass appear directly in the product’s coarse tail and its particle size distribution, and with them in the strength, the water demand and the workability of the concrete: the separator is the quality gate, and the mill is the engine behind it.
How is the separation curve measured in practice?
With the synchronized sampling of the three streams (the separator feed, the fines and the rejects), the sieve or the laser size analysis of each, and the calculation of the separation probability per size class from the mass balance: the modern plants run the curve on the particle size analyzer with the spreadsheet of the package, and the whole measurement takes a shift including the sampling: the frequency of the measurement is quarterly for the routine monitoring and per-trial during the tuning campaigns.
What causes the bypass in a third-generation separator?
Common contributors include seal or clearance problems, uneven feed distribution, poor dispersion, vane wear, air leakage and airflow maldistribution. Confirm the mechanism from the Tromp curve, airflow/pressure data and internal inspection before repair. Do not translate each bypass percentage point directly into an equal output gain; the production effect depends on the complete circuit.
Can the separator be tuned for the different cement types quickly?
Yes: the operating recipes of the modern plants (the rotor speed, the vanes, the airflow and the control setpoints per product) switch with the product change within the minutes: the charge and the liners of the mill stay, the separator follows the recipe: the file documents the recipe tables and the switch procedure, including the transition sampling that releases the new product: the classification is the most flexible layer of the circuit, and the tuned plants exploit that flexibility every day.
13. Conclusion
The classifiers and the separators are the instruments of the grinding circuits: they shape the product, they set the circulating load, and they convert the enterprise of the mill into the quality of the cement: the modern dynamic machines with their sharp curves and the low bypass are the standard, and the tuning discipline of the file (the curve measurement, the trial matrix, the verification) is the practice that keeps them there: the separator is the layer of the circuit where the careful work pays the quickest, and the plants that treat it as an instrument, not as a fan, collect the dividend every shift.
The Complete Cement Technical Package includes this classifier guide together with design references, separation-curve examples, tuning material and Excel calculators. The complete 931-file library is offered for $249 as a one-time purchase with instant download access immediately after payment.
Related Classifier and Grinding Guides
- Tromp Curve for Cement Separators: Efficiency Guide
- High Efficiency Separator: Complete Technical Guide
- Separator Operation in Cement Grinding
- Cement Grinding Systems: Complete Technical Guide
Get this classifiers and separators guide + the full 931-file package
$249 — one-time purchase, instant download, lifetime access
This file is part of the Complete Cement Technical Package (931 files) available from cementequipment.org. Respective rights holders; library copy for the licensed single user.
