In flotation, a collector renders the valuable mineral hydrophobic, air bubbles introduced into the slurry attach to the coated particles, and the particles rise into a froth that overflows as concentrate. Particles do not all float at the same rate: well-liberated grains of the valuable mineral float within the first minute or two, while composite, very fine and very coarse grains float slowly or not at all. A flotation kinetics test measures this distribution of rates and yields two parameters, the maximum recovery achievable with extended flotation and the rate constant describing how quickly it is approached. Together they determine the flotation time, and therefore the cell volume, that a plant requires.
The kinetics test is the standard laboratory test for designing and checking a flotation circuit. This article covers the procedure, the balancing of the test, the rate models used to describe the results, and the conversion of laboratory flotation time to plant residence time.
Test procedure
- Prepare a representative charge, often 1 to 2 kg for a standard laboratory cell, split from a well-mixed, crushed composite. An unrepresentative charge gives a precise answer to the wrong question.
- Grind in a laboratory rod or ball mill to the plant's target P80, with the grind time set from a prior grind calibration on the same ore.
- Wash the ground pulp into the flotation cell, dilute to the planned pulp density, set the impeller speed, and adjust the pH with lime or acid as in the plant.
- Condition with the collector and any activator or depressant, followed by the frother, each with a set conditioning time to allow the reagents to reach the mineral surfaces.
- Start the air and scrape the froth at a steady rate into separate pans over fixed intervals, for example 0 to 0.5, 0.5 to 1, 1 to 2, 2 to 4, 4 to 8 and 8 to 15 minutes. Keep the froth depth and scraping rate constant throughout, adding make-up water to maintain the pulp level.
- At the end of the test, the pulp remaining in the cell is the tailings. Filter, dry and weigh every concentrate and the tailings.
- Assay each product for the valuable element and for any gangue element, such as iron or silica, that affects concentrate grade.
All test conditions should be recorded, including reagent doses in grams per tonne, pH at each stage, air rate, impeller speed, temperature, water quality and any unusual froth behaviour, since a kinetics test is only useful if it can be repeated.
Test balance
The first calculation is a metal balance. The head grade is back-calculated from the products, all of which have been weighed and assayed, rather than taken from a separate assay of the feed. A large difference between the calculated and assayed heads indicates lost material or an incorrect assay.
where
- ,
- mass (g) and assay (%) of timed concentrate i
- ,
- mass (g) and assay (%) of the final tailings
- mass of ore floated (g)
- calculated (back-calculated) head assay (%)
The worked example is a copper rougher test on 1,000 g of ore.
| Product | Time (min) | Mass (g) | Cu (%) | Cu (g) | Cumulative mass (g) | Cumulative grade (% Cu) | Cumulative recovery (%) |
|---|---|---|---|---|---|---|---|
| Concentrate 1 | 0 to 0.5 | 13.1 | 24.0 | 3.144 | 13.1 | 24.0 | 37.0 |
| Concentrate 2 | 0.5 to 1 | 8.8 | 20.0 | 1.760 | 21.9 | 22.4 | 57.7 |
| Concentrate 3 | 1 to 2 | 11.1 | 14.0 | 1.554 | 33.0 | 19.6 | 76.0 |
| Concentrate 4 | 2 to 4 | 9.5 | 8.0 | 0.760 | 42.5 | 17.0 | 84.9 |
| Concentrate 5 | 4 to 8 | 7.7 | 4.0 | 0.308 | 50.2 | 15.0 | 88.5 |
| Concentrate 6 | 8 to 15 | 9.8 | 2.0 | 0.196 | 60.0 | 12.9 | 90.8 |
| Tailings | 940.0 | 0.083 | 0.780 | ||||
| Calculated head | 1,000.0 | 0.850 | 8.502 |
Copper in each product is mass × grade ÷ 100; for concentrate 1, 13.1 × 24.0 ÷ 100 = 3.144 g. The cumulative recovery after concentrate 2 is (3.144 + 1.760) ÷ 8.502 = 57.7%, and the cumulative grade is 4.904 g ÷ 21.9 g = 22.4% Cu. The pattern is typical: the first concentrates are high grade, and the grade falls as slower, composite particles begin to float.
Rate models
A plot of cumulative recovery against time rises steeply and then levels off. A rate model reduces the curve to parameters that can be compared between ores, reagent schemes or grinds, and used for scale-up.
First-order model
The simplest and most widely used model assumes that every floatable particle has the same probability of floating per unit time:
where
- cumulative recovery after flotation time t (%)
- maximum recovery reached at very long time (%)
- flotation rate constant (1/min): how fast the floatable mineral comes off
- flotation time (min)
Klimpel model
Real ores contain both fast- and slow-floating particles. The Klimpel model assumes that rate constants are distributed uniformly from zero to a maximum k, and it usually represents the slow tail of the curve more closely:
where
- the largest rate constant in a rectangular spread of rate constants from 0 to k (1/min)
- ,
- as above
Both models are fitted by least squares, using a spreadsheet solver or any curve-fitting tool to adjust R∞ and k until the sum of squared differences between measured and predicted recovery is minimised. For the worked test:
| Model | R∞ (%) | k (1/min) | Largest misfit (percentage points) |
|---|---|---|---|
| First order | 88.4 | 1.04 | 2.4 (at 15 min) |
| Klimpel | 94.4 | 2.32 | 1.8 (at 2 min) |
The first-order model levels off too early and misses the slow copper that continues to float after 8 minutes, whereas the Klimpel model follows that tail. Both are descriptions rather than mechanisms, so the same model should be used consistently when tests are compared, bearing in mind that R∞ is an extrapolation beyond the last concentrate rather than a measured value.
A quick check without a fitting tool is available for a first-order ore: a plot of ln[R∞ ÷ (R∞ − R)] against time is a straight line through the origin with slope k. With R∞ = 88.4%, the first four points give 0.54, 1.06, 1.97 and 3.24 at 0.5, 1, 2 and 4 minutes, close to a straight line with a slope of about 0.8 to 1.1, bending as the slow fraction becomes dominant.
Scale-up to plant residence time
A laboratory cell is a batch process in which every particle spends the same time. A plant bank is continuous, with pulp flowing through a series of well-mixed cells, and in each cell some pulp leaves almost immediately while some remains much longer. Because of this residence time distribution, a plant requires more total time than the batch test to reach the same recovery, even with identical chemistry.
where
- number of equal, well-mixed cells in series
- total nominal residence time of the bank (min)
- ,
- rate constant and maximum recovery, as above
Using k = 1.04 per minute from the first-order fit, the batch test reaches 90% of R∞ in 2.2 minutes. The chart shows the total residence time a bank requires for the same result:
A single large cell would need almost four times the batch time, six cells about 1.2 times, and twenty cells approach the batch result. For this reason flotation banks consist of several cells in series rather than one large tank.
Further differences between laboratory and plant widen the gap: froth recovery is lower in a large cell, air dispersion and froth removal differ, ore and water vary, and reagent additions are less precise. Designers therefore apply a single empirical scale-up factor to the laboratory time:
where
- design residence time of the plant bank (min)
- laboratory flotation time that gives the target recovery (min)
- scale-up factor from laboratory to plant (dimensionless)
Published factors for conventional cells mostly fall between about 2 and 4, depending on the ore, the laboratory cell size and the circuit stage. The laboratory's and the equipment supplier's figure for the duty should be used, and checked against a similar operating plant where possible. In the worked example, the target is the 88.5% recovery reached at 8 minutes, and with a scale-up factor of 2.5 the plant rougher requires 2.5 × 8 = 20 minutes. Our companion article on flotation cell sizing converts that time into cell volume.
Interpreting kinetics results
Results should be compared at equal grade rather than equal time, because a reagent that floats more mass may increase recovery while reducing grade; plotting cumulative grade against cumulative recovery allows a fair comparison. Duplicate tests are essential, since the difference between duplicates is the noise that a reagent or grind change must exceed before it can be considered an improvement.
Sizing the concentrates and tailings shows which size fractions float slowly. Slow coarse particles point to liberation or froth problems, while slow fines point to entrainment, slime coating or reagent problems. Fitting a rate model to a significant gangue mineral, such as pyrite in a copper circuit, identifies the window in which most of the valuable mineral has floated but the gangue has not yet begun to.
Common errors
- Changing the scraping rate or froth depth during the test, so that the rate constant reflects the operator rather than the ore.
- Allowing the ground pulp to stand and oxidise before flotation, particularly with sulphide ores.
- Calculating recovery from an assayed head rather than the calculated head, which can give recoveries above 100% or conceal a loss.
- Over-interpreting R∞, which is a fitted value; a test that stops at 8 minutes provides little information about recovery at 30 minutes.
Glossary
- Collector
- A chemical that coats the valuable mineral and makes it water-repellent so it attaches to air bubbles.
- Frother
- A chemical that makes small, stable bubbles and a froth that holds the floated mineral.
- Rougher
- The first flotation stage, which aims to recover as much of the valuable mineral as possible at a moderate grade.
- Rate constant, k
- How fast the floatable mineral comes off, in units of 1/min. A larger k means faster flotation.
- Maximum recovery, R∞
- The recovery the model predicts after a very long flotation time.
- Calculated head
- The feed grade worked out by adding up the metal in all products, rather than assaying the feed directly.
- Scale-up factor
- The number laboratory time is multiplied by to give the design residence time of a plant bank.
- Short-circuiting
- Pulp leaving a well-mixed cell sooner than average, so it has less chance to float.
- Entrainment
- Fine particles carried into the froth with water rather than attached to bubbles.
Frequently asked questions
How long should a flotation kinetics test run?
Long enough for the recovery curve to level off. Rougher tests commonly run for 10 to 20 minutes with five to seven timed concentrates, collected more frequently early in the test when flotation is fastest.
Should the first-order or the Klimpel model be used?
Fit both, select the one that represents the data more closely, and use that model for all comparisons. The Klimpel model often fits ores with a slow-floating tail better.
What scale-up factor should be used for plant design?
Published factors for conventional cells mostly fall between about 2 and 4. The appropriate value depends on the ore, the laboratory cell and the circuit stage, and should be obtained from the laboratory and equipment supplier for the duty and compared with similar operations.
Why does the calculated head differ from the assayed head?
Small differences arise from sampling and assay error. A large difference suggests lost material, a weighing error or an incorrect assay, and the test should be checked before the results are used.
Can kinetics tests be used on an operating plant?
Yes. Kinetics tests on plant feed, or on samples from each cell, show whether the plant is short of residence time, whether the reagents are effective, and how the ore is changing.
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 metallurgistReferences and further reading
- Wills, B.A. and Finch, J.A. (2016). Wills' Mineral Processing Technology, 8th edition. Butterworth-Heinemann. Chapter 12, Froth flotation.
- Klimpel, R.R. (1980). Selection of chemical reagents for flotation. In Mular, A.L. and Bhappu, R.B. (eds), Mineral Processing Plant Design, 2nd edition. Society for Mining, Metallurgy and Exploration.
- Napier-Munn, T.J. (2014). Statistical Methods for Mineral Engineers: How to Design Experiments and Analyse Data. JKMRC Monograph Series, University of Queensland.
Worked examples and charts marked illustrative use invented figures, not data from any operation.