Tiegel home Talk to a metallurgist

Hydrocyclone performance: partition curves, cut size and bypass from a cyclone survey

Calculating a hydrocyclone partition curve from survey data, determining the cut size d50c and water bypass, and estimating cyclone size with Arterburn's method.

· 7 minute read · By

A hydrocyclone is a cone with no moving parts. Slurry is pumped tangentially into the top, so that it rotates. The rotation throws coarse and dense particles outward to the wall, where they spiral down and leave through the apex as the underflow. Fine and light particles remain near the axis and are carried up through the vortex finder with most of the water as the overflow.

In a grinding circuit the cyclone determines which particles proceed to flotation or leaching and which return to the mill for further grinding. The separation is never perfect: some coarse particles escape to the overflow, and some fines are carried into the underflow with the water. The partition curve quantifies the classification.

Hydrocyclones control the product size of almost every grinding circuit, so their classification performance is a basic plant measurement. This article covers the cyclone survey, calculation of the partition curve, correction for water bypass, determination of cut size and sharpness, and estimation of the cyclone size needed for a target grind.

Hydrocyclone streamsFeed enters tangentially; fine particles leave with most of the water through the vortex finder as overflow; coarse particles leave through the apex as underflow.Feed (under pressure)Overflow: fines + most waterUnderflow: coarse + some waterSpinning flow throwscoarse, heavy particlesto the wall and down;fines stay in the coreand go up
How a hydrocyclone splits a slurry by size. The fraction of water that goes to the underflow carries some fines with it unclassified.

Cyclone survey

  1. Survey only when the circuit has been stable for at least an hour, with steady feed rate, mill power, cyclone pressure and number of cyclones operating, and record these conditions.
  2. Sample the feed, underflow and overflow over the same period, with several increments of each. Cut the full stream where possible; a partial cut of a stream that is segregated by size gives biased sizing.
  3. Determine the percent solids of each sample by weighing it wet, drying it and weighing it again.
  4. Size each stream by wet screening at the finest sieve, then drying and dry screening the coarser material through a √2 sieve series, for example 850, 600, 425, 300, 212, 150, 106, 75, 53 and 38 µm.

Underflow mass split

The fraction of feed solids reporting to the underflow, U/F, is obtained from a mass balance. With percent solids by mass of 55% in the feed, 75% in the underflow and 35% in the overflow, the circulating load method in the circulating load calculator gives an underflow-to-overflow ratio of 2.14, so U/F = 2.14 ÷ 3.14 = 0.682. The sizing data can give the same ratio, and reconciling the two by mass balance reconciliation is good practice.

Partition numbers by size class

Ei=U·uiF·fi

where

Ei
partition number: fraction of the feed in size class i that reports to the underflow
UF
fraction of the feed solids that reports to the underflow (from % solids or sizing data)
ui, fi
mass % of size class i in the underflow and in the feed
Size class (µm)Feed (%)Underflow (%)Overflow (%)Partition E (%)Corrected Ec (%)
850 to 6004.05.90.0100100
600 to 4256.59.50.099.699.5
425 to 3009.013.20.0100100
300 to 21211.015.80.697.997.1
212 to 15012.015.93.790.386.6
150 to 10611.512.59.574.164.1
106 to 7510.08.313.656.639.9
75 to 538.05.313.845.224.1
53 to 386.53.712.638.815.3
Below 3821.510.046.231.75.4

For the 150 to 106 µm class, E = 0.682 × 12.5 ÷ 11.5 = 0.741, or 74.1%. Coarse classes can calculate slightly above 100% because of sampling and sieving error and should be read as 100%.

Water bypass and the corrected curve

The actual partition curve does not fall to zero at the finest sizes. In the example, almost a third of the material below 38 µm still reports to the underflow. These fines were not classified; they followed the water leaving through the apex. The fraction of feed water reporting to the underflow is the bypass, Rf:

Rf=water in underflowwater in feedEc,i=Ei−Rf1−Rf

where

Rf
water split (bypass) to underflow: the fraction of fine particles that follow the water to the underflow unclassified
Ec,i
corrected partition number, the classification the cyclone actually did

From the percent solids, the feed carries 45 ÷ 55 = 0.818 t of water per tonne of solids. The underflow carries 0.682 t of solids with 25 ÷ 75 × 0.682 = 0.227 t of water, so Rf = 0.227 ÷ 0.818 = 27.8%. For the 106 to 75 µm class, Ec = (56.6 − 27.8) ÷ (100 − 27.8) = 39.9%.

Hydrocyclone partition curveFraction of each size to underflow. The actual curve levels off at about 28% for fine sizes because of water bypass; the corrected curve passes 50% at 104 micrometres.02550751002050100200500Particle size (µm, log scale)To underflow (%)water bypass Rf = 27.8%50%Actual partitionCorrected partitionLynch–Rao modelActual partition: 714 → 100Actual partition: 505 → 100Actual partition: 357 → 100Actual partition: 252 → 98Actual partition: 178 → 90Actual partition: 126 → 74Actual partition: 89 → 57Actual partition: 63 → 45Actual partition: 45 → 39Actual partition: 20 → 32ActualCorrected partition: 714 → 100Corrected partition: 505 → 100Corrected partition: 357 → 100Corrected partition: 252 → 97Corrected partition: 178 → 87Corrected partition: 126 → 64Corrected partition: 89 → 40Corrected partition: 63 → 24Corrected partition: 45 → 15Corrected partition: 20 → 5CorrectedModel
Partition curve from the worked survey. Water bypass lifts the actual curve at fine sizes; correcting for it gives the cut size d50c of about 104 µm. Illustrative figures.

Cut size and sharpness

The corrected cut size, d50c, is the size at which the corrected curve crosses 50%. Interpolating on a log scale between the 89 µm (39.9%) and 126 µm (64.1%) class mid-points gives d50c ≈ 103 µm. The actual, uncorrected d50 is finer, at about 73 µm, because the bypassed fines raise the curve.

Sharpness describes the steepness of the curve. The Lynch and Rao reduced efficiency curve is widely used:

Ec=eαx−1eαx+eα−2withx=dd50c

where

d
particle size (µm)
d50c
corrected cut size: the size with an equal chance of going either way (µm)
α
sharpness of separation: a higher value means a sharper cut

Fitting this curve to the corrected values gives α ≈ 2.5 for the example, which is a typical value. A simpler measure of sharpness is the ratio d25c ÷ d75c read from the corrected curve; values closer to 1 indicate a sharper cut.

Interpreting bypass, cut size and roping

High bypass returns unclassified fines to the mill, where they are over-ground and occupy mill capacity. Causes include an apex too large for the duty, low feed pressure and excessive water reporting to the underflow. A denser underflow, within the limit of roping, reduces bypass.

A coarse d50c produces a coarse overflow. The cut size rises with feed density and cyclone diameter, and with lower operating pressure.

Roping, in which the underflow discharges as a thick rope instead of a spray, indicates that the apex is overloaded. Coarse particles are then forced into the overflow and the classification deteriorates sharply.

Cyclone sizing by Arterburn's method

For a first estimate of cyclone diameter, Arterburn's widely used method starts from a base cut size for a standard cyclone and corrects it for the feed conditions:

d50c,base=2.84×D0.66d50c=d50c,base×C1×C2×C3C1=(53−V53)−1.43C2=3.27ΔP−0.28C3=(1.65Gs−Gl)0.5

where

D
cyclone diameter (cm)
d50c
corrected cut size (µm)
V
feed solids concentration by volume (%)
ΔP
pressure drop across the cyclone (kPa)
Gs, Gl
specific gravity of the solids and of the liquid

Worked example: the survey feed is 55% solids by mass with a solids SG of 2.7, which is 31.2% solids by volume, and the pressure drop is 70 kPa. The correction factors are C1 = ((53 − 31.2) ÷ 53)−1.43 = 3.55, reflecting the marked coarsening of the cut in a dense feed; C2 = 3.27 × 70−0.28 = 1.00; and C3 = (1.65 ÷ 1.7)0.5 = 0.99.

Arterburn related the required d50c to the product size through a set of multipliers for a typical overflow size distribution: for a product 80% passing a given size, d50c is 1.25 times that size. A target overflow P80 of 106 µm therefore requires d50c ≈ 1.25 × 106 = 133 µm. The base cut size must be 133 ÷ (3.55 × 1.00 × 0.99) = 38 µm, so 2.84 × D0.66 = 38 and D ≈ 51 cm. A cyclone of about 500 mm diameter is the starting point. The number of cyclones then follows from the feed flow and the capacity of one cyclone at the selected pressure, taken from the supplier's data.

These are design estimates for a typical ore. Plant surveys such as the one above check the estimate, and the cyclones are then tuned through the apex and vortex finder sizes.

Glossary

Underflow
The coarse product leaving the bottom (apex) of the cyclone; in a grinding circuit it returns to the mill.
Overflow
The fine product leaving through the vortex finder at the top.
Partition curve
A graph of the fraction of each particle size that reports to the underflow.
Cut size, d50c
The particle size with a 50% chance of going to the underflow on the corrected curve.
Water bypass, Rf
The fraction of feed water that goes to the underflow, carrying unclassified fines with it.
Sharpness, α
How steep the partition curve is. A sharper cut sends fewer wrong-sized particles to each product.
Roping
An overloaded apex discharging a thick rope of slurry, which forces coarse particles into the overflow.
Apex and vortex finder
The bottom outlet and the top outlet tube of the cyclone. Their sizes set the split of water and solids.

Frequently asked questions

What is the difference between d50 and d50c?

d50 is read from the actual partition curve, which includes the bypassed fines. d50c is read from the curve corrected for bypass and describes the classification itself. d50c is used for comparing cyclones and in design.

How can bypass in a hydrocyclone be reduced?

By raising underflow density, within the limit of roping, through a smaller apex or adjusted operating conditions, and by keeping feed pressure in the design range. The cyclone should also be checked for wear, which changes the apex and spigot dimensions.

Why does feed density affect the cut size so strongly?

A dense slurry slows the movement of particles through the fluid, so more of them remain in the inner flow and leave with the overflow. Arterburn's C1 factor represents this effect and rises steeply as the volume percent solids increases.

Is Arterburn's method accurate enough for design?

It is a widely used first estimate for a typical ore. Final selection uses supplier data, models calibrated on surveys of similar plants, or testwork.

Discuss your plant with a metallurgist

Tiegel is a Perth metallurgy company. We help with testwork, plant optimisation, metal accounting and investigations, and we build the Tiegel Suite software.

Talk to a metallurgist

References and further reading

  1. Arterburn, R.A. (1982). The sizing and selection of hydrocyclones. In Mular, A.L. and Jergensen, G.V. (eds), Design and Installation of Comminution Circuits. Society for Mining, Metallurgy and Exploration. Link
  2. Lynch, A.J. and Rao, T.C. (1975). Modelling and scale-up of hydrocyclone classifiers. Proceedings of the 11th International Mineral Processing Congress, Cagliari.
  3. Napier-Munn, T.J., Morrell, S., Morrison, R.D. and Kojovic, T. (1996). Mineral Comminution Circuits: Their Operation and Optimisation. JKMRC Monograph Series, University of Queensland.
  4. Wills, B.A. and Finch, J.A. (2016). Wills' Mineral Processing Technology, 8th edition. Butterworth-Heinemann. Chapter 9, Classification.

Worked examples and charts marked illustrative use invented figures, not data from any operation.