Chapter 14 · Classification and Partition

Hydrocyclone hydraulics and classification

Operating point, curve and closure

A hydrocyclone model typically has three linked tasks. It converts per-cyclone slurry flow into a pressure or head. It estimates a corrected partition using geometry, operating conditions and material parameters. It reconciles the actual solids split and liquid recovery. The term “split” can refer to the slurry underflow/overflow volume ratio, the fraction of feed slurry volume to underflow or the fraction of feed water to underflow. Each uses a different denominator.

For slurry volume Q, underflow volume Qᵤ and overflow volume Qₒ, define S = Qᵤ/Qₒ and Rᵥ = Qᵤ/Q. Then Rᵥ = S/(1+S). At the feed, compute solid volumes from each component's mass and density before adding water. Take one correctly assigned operating count before calculating per-cyclone volume. In a manifold where units share the same head and geometry, Q_each = Q_total/N. Unequal wear, valve positions or cyclone sizes make that approximation less reliable.

Underflow grade has a size and density dependence. A dense component can have smaller corrected cut size at the same pressure because its buoyant density differs. A model that uses one bulk curve for all components cannot represent that selective response except through component differences already present in the feed size distribution. A component-resolved model should use survey data capable of distinguishing component/size effects from feed mineralogy.

Plitt and Plitt–Flintoff forms

The Original Plitt representation calculates pressure, slurry volume split, corrected cut size and sharpness, then reconciles fines bypass subject to the modeled split. One form of the corrected coarse curve uses

E(x) = 1−exp[−ln 2 (x/d₅₀꜀)ᵐ].

Where calibrated component curves are available, calculate corrected recovered solids volume over both sizes and components. The Alpha form additionally allows component adjustments to the cut, sharpness and split and a coarse bypass to overflow. With water/fines bypass r and effective coarse bypass c, one representation is p = r + (1−r−c)E. The bounds are p(0) = r and p(∞) = 1−c. Bravo adds explicit alternative cut-size branches, component/density and throughput corrections, and limits on calculated split and class partition. A limit can be useful for control simulation. Inspect when a result sits on a limit, because the underlying relation can still be extrapolated beyond calibration. Original, Alpha and Bravo have pressure-based active-count options relevant to a circuit model with transient flow.

For the 1987 Plitt–Flintoff form, use cyclone diameters in cm and slurry flow q in L/min per operating cyclone. Define D꜀ as body diameter, Dᵢ as effective inlet diameter, Dₒ as vortex finder diameter, Dᵤ as spigot diameter, H as effective vortex height, T = Dᵤ²+Dₒ² and Cᵥ as solids percentage by volume. With calibration coefficients f and pulp density ρ in t/m³, one correlation package is

P = fₚ 1.88 q¹·⁸⁰ exp(0.0055Cᵥ)/(D꜀⁰·³⁷ Dᵢ⁰·⁹⁴ H⁰·²⁸ T⁰·⁸⁷).
S = fₛ 18.62 ρ⁰·²⁴ (Dᵤ/Dₒ)³·³¹ H⁰·⁵⁴ T⁰·³⁶ exp(0.0054Cᵥ)/(D꜀¹·¹¹ P⁰·²⁴).
m = fₘ 1.94 exp(−1.58Rᵥ)(D꜀²H/q)⁰·¹⁵.

Pressure P is in kPa for these coefficients. With apparent viscosity μ in the calibrated correlation units and bulk solids buoyant density Δρ on the t/m³ scale, corrected cut in µm is

d₅₀꜀ = f_d 39.7 D꜀⁰·⁴⁶ Dᵢ⁰·⁶⁰ Dₒ¹·²¹ √μ exp(0.063Cᵥ) / [Dᵤ⁰·⁷¹ H⁰·³⁸ q⁰·⁴⁵ (Δρ/1.6)ᵏ].

The constants include their empirical unit conventions. Moving from centimetres to millimetres without transforming the coefficients can materially change all four outputs. This model uses a bulk corrected curve. Its liquid bypass follows from the underflow volume closure rather than a separate water correlation. A large required change of fₚ can mask an inconsistent pressure head conversion. Check the measurement basis first.

Worked hydraulic closure from a corrected partition

Let the cyclone receive 40 t/h solids of density 2.50 t/m³ and 70 m³/h water at density 1.00 t/m³. Then Vₛ = 16, V𝑤 = 70 and Q = 86 m³/h. Suppose the hydraulic correlation gives Rᵥ = 30/86 and integration of the corrected partition gives K = 8 m³/h corrected solids recovered to underflow. For liquid bypass b, actual solids volume is K+b(Vₛ−K), and actual water volume is bV𝑤. Therefore

b = (Q Rᵥ−K)/(Q−K) = (30−8)/(86−8) = 0.282051.

Underflow receives 10.2564 m³/h solids and 19.7436 m³/h water. Their total is 30.000 m³/h. On a mass basis these are 25.641 t/h solids and 19.744 t/h water, or 56.50% solids by mass. If the same corrected curve were paired with an underflow hydraulic target of 7 m³/h, the inferred b would be −1/78. The target is below the corrected solids volume itself. Recalibrate the hydraulic and partition inputs using the same survey. Simply bounding b to zero hides a residual of 1 m³/h at minimum.

Nageswararao and Narasimha–Mainza

The Nageswararao family expresses pressure through a capacity term based on cyclone body size and geometric ratios. For one unit, representative volume flow Q is in m³/h and the pressure relation is P = ρ(Q/K_Q)². With body diameter D in metres, define Πₚ = P/(ρgD), using consistent pressure and density scales, and volumetric solids fraction c. A hindered-settling group is

H꜀ = 10^(1.82c)/[8.05(1−c)²].

Holding geometry fixed, corrected cut varies as Πₚ⁻⁰·²²H꜀⁰·⁹³ in this package. Water recovery varies as Πₚ⁻⁰·⁵³H꜀⁰·²⁷, with distinct factors for spigot, vortex finder, inlet, cone and effective length. The independent slurry volume correlation is a diagnostic value. The simulated volume results from the water recovery and size-partition mass balance. A measured discrepancy between those two predictions constrains calibration of the pressure, cut and volume terms separately.

Narasimha–Mainza introduces semi-mechanistic flow groups from feed rheology and the inlet. Given flow Q in m³/s and inlet diameter Dᵢ in m, it computes inlet velocity uᵢ = 4Q/(πDᵢ²) and a tangential velocity estimate uₜ = 4.5(Dᵢ/D)¹·¹³uᵢ. The centrifugal number is uₜ²/(gD/2) and the inlet Reynolds group is Re = ρuᵢD/μ with kg/m³ and Pa·s. A viscosity estimate uses μ = μ𝑤(1−c/0.62)⁻¹·⁵⁵ on its valid concentration range, or a measured apparent viscosity can be used. Recirculating slurries with strong rheological variability benefit from viscosity data at a stated measurement condition. The family correlates pressure, corrected cut, sharpness and water recovery through these flow and geometry groups. At c = 0.31, the fitted viscosity formula alone gives μ/μ𝑤 = (1−0.31/0.62)⁻¹·⁵⁵ ≈ 2.93.

Both forms may use a Whiten–Beta reduced overflow curve. Let z = β x/d₅₀꜀, where β is the descending root that gives G(β*) = 0.5. The curve is

G(z) = (1+βz)[exp(α)−1]/[exp(αz)+exp(α)−2].

The actual underflow partition is 1−(1−b)G(z), bounded to [0,1]. For β = 0, β* = 1. A nonzero β permits a hook at the fine end. A sparse fine tail may leave β uninterpretable. Confirm whether fine recovery residual is reproducible before treating the hook as a transferable property of the classifier.

Surveys and exercises

A useful sequence is to compare pressure against flow at known active counts, then fit water and volumetric splits, then estimate corrected cut and sharpness from reconciled products. For components, check both the total and individual curve agreement. A single product P₈₀ can be matched by several combinations of d₅₀꜀, sharpness and water bypass. Slurry volume, water flow and more than one size percentile add independent constraints. Check survey consistency at a second pressure or flow before using the result for prediction.

1. Pressure sensitivity to active count

At fixed geometry and slurry condition, pressure scales with q¹·⁸ in a group of four identical cyclones. What is the pressure factor if a fifth opens and total flow stays fixed?

Solution. Per-unit flow is 4/5 of its previous value. The new pressure is (0.8)¹·⁸ ≈ 0.669 times the old value. In a pump-fed circuit the total flow can change with pressure, so the fixed-flow estimate must be combined with the pump and piping balance.

2. Fines and coarse leakage

An Alpha-type coarse partition has fitted r = 0.22 and c = 0.07. What is actual recovery at the corrected midpoint?

Solution. p = 0.22 + (1−0.22−0.07) × 0.5 = 0.575. At the fine and coarse limits the recoveries approach 0.22 and 0.93.

3. Verify the flow field

With Q = 0.04 m³/s, Dᵢ = 0.08 m and D = 0.40 m, estimate the inlet velocity and approximate tangential velocity in the semi-mechanistic relation.

Solution. uᵢ = 4 × 0.04/(π × 0.08²) = 7.96 m/s. uₜ = 4.5(0.08/0.40)¹·¹³ × 7.96 ≈ 5.81 m/s. The centrifugal number is approximately 5.81²/(9.80665 × 0.20) = 17.2. A calculation based on flow in m³/h without conversion would raise the velocity by 3,600.

Reading

  • Plitt (1976) and Flintoff, Plitt and Turak (1987), empirical hydrocyclone classification and calibration. Nageswararao et al. (2004) present the formulation and discuss its relationship to the Nageswararao family. Two empirical hydrocyclone models revisited.
  • Narasimha et al. (2014), A semi-mechanistic model of hydrocyclones — Developed from industrial data and inputs from CFD. Journal article.