Chapter 20 · Concentration and Separation

Froth transport and column flotation

Why a wet froth requires two balances

A concentrate launder receives solids attached to bubbles, entrained solids transported from the pulp and liquid that drains only partly before overflow. Attached solids can detach and return to the pulp. Fine suspended gangue can enter the concentrate with interstitial water even when its surfaces have little affinity for bubbles. A useful flotation model therefore needs both a solids balance by particle class and a water balance. Under simple entrainment, an increase in concentrate water creates a second route for fine gangue and may lower concentrate grade. A decline in froth recovery lowers true-flotation performance even when collection in the pulp is unchanged.

A sampling plan separates these effects by measuring concentrate water and solids concurrently. A series of tests varying water recovery, with air rate and feed conditions tracked, can help identify a water-entrainment relation. Deepening froth or increasing wash water can also change collection and froth drainage, so a grade change alone cannot identify which mechanism changed. A concentrate with solids flow Cₛ and water flow C𝓌 has a wet mass flow Cₛ + C𝓌 and a dry-solids mass fraction Cₛ/(Cₛ + C𝓌). Froth aeration can affect its apparent volumetric flow and pump duty independently of these two conserved masses.

Compartment odds for a mechanical cell

Savassi (2005) analyzed true flotation and entrainment as simultaneous mechanisms originating in distinct pulp regions. The implemented Savassi Compartment Flotation Cell uses a reduced steady-state form with measured or specified feed-water recovery W, class collection rate k in min⁻¹, effective residence time τ in minutes, combined froth survival F and entrainment coefficient E. With W below one, the water-conditioned entrainment odds increase nonlinearly.

Oflot = kτF.
Oent = E W/(1 − W).
R = (Oflot + Oent)/(1 + Oflot + Oent).

Combined recovery is derived from combined odds. For independent pulp loss routes with fixed relative rates competing with discharge, odds are natural because each route is compared to the same uncollected outlet. The active equations are a reduced model of the publication. They represent the three compartments through fitted parameters, without resolving recirculating water or froth drainage as independent streams. Use class-specific rates, froth survival and E if supported by the data. Class E can be nearly zero for coarse suspended particles and high for ultrafines. When W approaches one, the odds approach infinity for positive E, and the model becomes very sensitive to that water setting.

Consider 15 t/h of a particular mineral in the feed and 80 t/h of water in total. The class has k = 0.45 min⁻¹, τ = 4 min, F = 0.80, E = 0.35 and W = 0.25.

Oflot = 1.4400, Oent = 0.11667, R = 1.55667/2.55667 = 0.60887.

The concentrate contains 9.1330 t/h of that class and 20 t/h water. Its complete solids grade requires the sum of all other floated classes. The flotation-only recovery for the same k, τ and F is 1.44/2.44 = 0.59016. The effect of entrainment on the combined recovery is 1.870 percentage points. In the odds balance, 0.11667/2.55667 of the feed can be associated with the entrainment route and 1.44/2.55667 with true flotation. The increase on enabling entrainment is smaller than the assigned entrainment mass because both routes draw on the same remaining feed.

Column hydraulics and axial dispersion

Tall columns have an extended bubble–particle contact zone and a froth zone through which gas and recovered solids rise. Column diameter D sets area A = πD²/4. For n identical columns in parallel, feed to each column is divided by n. Use metres and seconds in hydraulic equations. Convert gas superficial velocity from cm/s by multiplying by 0.01 and bubble diameter from mm by dividing by 1000. With downward bias uᵦ and gas volume fraction εg, the active advanced model uses the following hydraulic approximations. Q𝓌 is feed-water volume to one column in m³/h, taking a reference water density of 1 t/m³. Hc is effective collection height and uₛ is downward particle slip.

uₗ = Q𝓌/(3600A) + max(0, uᵦ).
τₗ = Hc(1 − εg)/uₗ.
τₚ = τₗ uₗ/(uₗ + uₛ).

The slip estimate is an iterative particle force balance with hindered settling and a Reynolds correction. Using d in metres, liquid viscosity μ in Pa·s, solids volume fraction φ, density difference Δρ in kg/m³ and particle density ρₚ, the advanced expression is as follows.

uₛ = g d² max(0, Δρ)(1−φ)²·⁷/[18μ(1+0.15 Re⁰·⁶⁸⁷)].
Re = d uₛ ρₚ(1−φ)/μ.

The calculation uses a fixed number of iterations. Inspect τₚ and Re for unusual inputs. A higher downward superficial velocity or faster particle settling reduces contact time. Regrind may increase fine entrainment while lengthening the available time for fine-particle collection.

For gas velocity ug and bubble diameter db, the advanced model clips a size-dependent collection probability for d below dmax.

P = clip{α(d/db)²ⁿ[1 − (d/dmax)¹·⁵], 0, 1}.
k = 1.5 (ug/db) P.

Set P to zero at or above the maximum floatable diameter and at nonpositive size. Dimensionless α and n tune the collection probability. Increasing ug increases encounter frequency at fixed db. A real sparger may change db and gas holdup with the air rate, so the independent sensitivities in the model are conditional.

A dimensionless vessel number N is calculated from the dispersion adjustment β and hydraulic state. With c = √(1+4kτₚN), the axial-dispersion survival term S and advanced collection recovery are as follows.

N = (β/0.6)(D/Hc)⁰·⁶³ ug(1−εg)/uₗ.
S = 4c exp[1/(2N)] / {(1+c)² exp[c/(2N)] − (1−c)² exp[−c/(2N)]}.
Rcollection = 1 − S.

The model uses a numerically stabilized form and the exponential recovery limit for effectively zero dispersion. At k = 0, c = 1 and S = 1, which is a direct check of the transformation. The active Finch Column Flotation compatibility version also calculates S with its own input conventions and fixed viscosity. In that version the implemented function returns S directly to the Floated solids allocation. Increasing k can therefore decrease reported Floated recovery in a sensitivity test with other inputs fixed. Verify model response with a single known class before interpreting legacy simulations physically. The Finch-Dobby-Modified Advanced Column Flotation model assigns the collection complement 1−S as described above and models interface and froth effects explicitly.

Froth survival, wash water and capacity

In the advanced column, two rate multipliers generate kfast and kslow from k. The collection fraction R₁ is F∞ times the weighted sum of their separate dispersion recoveries. An interface survival R₂ is specified. Froth residence τf is depth Hf times froth gas holdup εf divided by gas superficial velocity. Froth survival R₃ is a fitted factor G times exp(−λτf), with λ in s⁻¹. The model treats interface rejection and froth detachment as possible returns to the collection sequence.

τf = Hf εf/ug.
R₃ = G exp(−λτf).
Rtrue = R₁R₂R₃/[1 − R₁(1−R₂) − R₁R₂(1−R₃)].

To derive the denominator, sum the probability of all repeated attempts after return, with a first-attempt return probability of R₁(1−R₂) + R₁R₂(1−R₃). This geometric-series argument applies in the admissible range below certain return. Degenerate endpoints are guarded in the implementation. The fraction of remaining solids recovered by entrainment is clipped E W/[1 + (d/dₑ)ⁿᵉ], where W is feed-water recovery to Floated and dₑ is the half-effect size. The class’s true and entrained masses then contribute to total froth solids load.

In direct water mode, W is the specified fraction W₀. In wash-water mode, the model reduces W₀ using a wash ratio X = added wash water / (feed water × W₀), subject to a small-denominator guard, and downward bias in m/s.

W = W₀ exp(−ηX)/[1 + 10 max(0, uᵦ)].

Total outlet water is feed water plus added wash water. The Floated water remains W times feed water and Bottom receives the residual of the total. Track wash water as an explicit circuit input when reviewing plant closure. The optional carrying-capacity estimate is Fcap A n ug, using ug in cm/s in this expression. When estimated capacity is below total floated solids before the limit, both true and entrained masses in all classes are reduced by the same capacity-to-load ratio. The redistribution preserves the prelimit concentrate solids composition under uniform scaling. It changes the total mass pull, and water routing is computed separately. An operating column can lose capacity preferentially by particle class, a behavior that this common factor cannot resolve.

Worked column case

One column is 1.2 m in diameter, with a 6 m collection zone and 2.4 m froth depth within an 8.4 m height. Use collection gas holdup 0.20, froth gas holdup 0.80, bubble diameter 1.5 mm, gas velocity 2.4 cm/s and downward bias 0.08 cm/s. Feed contains 15 t/h water and 4 t/h solids, all in one class at d = 55 µm and ρₚ = 3200 kg/m³. Its solids volume fraction is 1.25/16.25 = 0.076923. Set μ = 0.00125 Pa·s and β = 0.8. Set α = 0.36, n = 0.8, dmax = 240 µm, F∞ = 0.85, fast share = 0.60, fast and slow rate multipliers 1.3 and 0.3, R₂ = 0.90, G = 0.80 and λ = 0.006 s⁻¹.

CalculationValue
Area and downward liquid velocity1.13097 m² and 0.00448414 m/s
Liquid residence and vessel number1070.44 s and 2.07114
Slip velocity and particle Reynolds number0.00219820 m/s and 0.28570
Particle residence718.31 s
Collection probability and base rate0.00161685 and 0.0388045 s⁻¹
Fast and slow collection recovery0.99603 and 0.93522
Weighted collection fraction R₁0.82595
Froth residence and froth survival R₃80.0 s and 0.49503
True recovery before carrying limit0.67889

Set W₀ = 0.32, η = 0.70 and 2 t/h wash water. The initial wash ratio is 2/(15 × 0.32) = 0.41667. W falls to 0.23715, leaving 3.55723 t/h water in the Floated stream and 13.44277 t/h in Bottom. With E = 0.55, dₑ = 25 µm and nₑ = 2, entrainment recovers 0.022334 of the remaining class. Before the carrying limit, true mass is 2.71556 t/h and entrained mass is 0.02869 t/h. An enabled limit with Fcap = 0.6 gives 0.6 × 1.13097 × 2.4 = 1.62860 t/h. It scales both mechanisms by 0.59346. Floated solids remaining are 1.61158 t/h true and 0.01702 t/h entrained. Bottom has 2.37140 t/h solids, and the Floated slurry has 31.405% solids by mass.

Holding the carrying limit fixed can conceal strong sensitivity of collection. In this example the reported floated solids stay at 1.62860 t/h whenever the predicted uncapped load remains above the limit. Review both the prelimit and delivered load when fitting. Use tracer and hydraulic data to fix plausible residence and dispersion, fit collection on data below capacity, examine froth and wash conditions for entrainment, then fit the capacity factor using runs near the observed operating limit.

Exercises and answers

  1. For a class with kτF = 1.8 and E = 0.2, compare Savassi recovery at W = 0 and W = 0.5.
  2. In the column case, suppress the carrying limit and compute concentrate solids concentration. Explain which stream water changes.
  3. Consider equal-size hydrophobic and hydrophilic classes with no shared entrainment. Which parameter combinations can be fixed using recovery measured only at a single air rate and froth depth?

Answers

  1. The recoveries are 1.8/2.8 = 0.64286 and 2.0/3.0 = 0.66667.
  2. Solids rise to 2.74425 t/h at unchanged 3.55723 t/h floated water, giving 43.549% solids by mass. Water in both streams is unchanged because its equations do not depend on capacity.
  3. The measured total constrains the combined collection-to-froth response at that state. An infinity of collection, interface and froth survival combinations can yield that response. Hydraulic measurements and targeted changes in air rate or froth depth are required to constrain the stages.

Savassi (2005), A compartment model for the mass transfer inside a conventional flotation cell, describes the compartment concept. Finch and Dobby (1990), Column Flotation, discuss collection, dispersion and froth phenomena. The equations and compatibility behavior described here were checked against SavassiCompartmentFlotation, ColumnFlotationFinch and FinchDobbyColumnFlotationAdvanced.