Grinding is the most energy-intensive step in most mineral processing plants, and the energy required depends on ore hardness. The Bond work index is the industry's standard measure of that hardness: the higher the value, the more energy each tonne requires. The standard laboratory test is accurate but takes most of a working day per sample, which is too slow for measuring hardness on every shift.
The evaluation of any grinding circuit change faces the same difficulty. New liners, a different ball size or a new control system are installed, throughput changes, and the ore changes at the same time. Without a measure of ore hardness, the change attributable to the plant cannot be separated from the change attributable to the ore. A quick, repeatable hardness measurement on plant samples resolves this. This article describes one such method, published by William Horst and John Bassarear in 1976, and the application of its approach to grinding circuit evaluation today.
The Bond work index
The Bond work index, Wi, expresses an ore's resistance to grinding as the energy needed to reduce it from a very coarse size to 80% passing 100 µm. Bond's equation relates it to the energy W needed for any reduction from a feed 80% passing size F to a product 80% passing size P, both in microns:
where
- specific grinding energy (kWh/t)
- Bond work index (kWh/t)
- product size, 80% passing (µm)
- feed size, 80% passing (µm)
Bond defined Wi per short ton. Values quoted per metric tonne are 1.102 times larger, so the units should be checked before figures from different sources are compared.
The standard Bond ball mill grindability test simulates a closed grinding circuit in a laboratory mill until it reaches steady state. It is accurate and widely used for design, but Horst and Bassarear note that it takes a trained technician about an eight-hour shift per sample. That is acceptable for a design programme but too slow for a plant that requires hardness data on several streams every week.
The comparative method
Berry and Bruce proposed a shortcut in 1966. Equal masses of a reference ore of known work index and of the unknown ore are ground in the same laboratory mill under identical conditions. Because both samples receive the same energy, Bond's equation can be equated for the two, giving:
where
- work index of the unknown (plant) sample (kWh/t)
- work index of the reference ore, from a standard Bond test (kWh/t)
- ,
- 80% passing sizes of the reference ore product and feed (µm)
- ,
- 80% passing sizes of the unknown ore product and feed (µm)
The weakness of the method lies in the feed. It works well only when the two feeds have similar size distributions, and samples taken from a plant, such as SAG mill feed or screen undersize, rarely match the reference ore after crushing. Horst and Bassarear's comparison shows the consequence: on two of six copper ores, the uncorrected comparative result was 2 to 3 kWh/t lower than the standard Bond value.
Breakage rates and the feed size correction
Horst and Bassarear's improvement was to describe grinding in the laboratory mill with first-order breakage kinetics, in which the mass fraction coarser than each test sieve falls exponentially with grinding time:
where
- mass fraction coarser than test sieve i after grinding time t
- mass fraction coarser than sieve i in the mill feed
- first-order breakage rate for sieve i (per minute)
- grinding time (min)
Here ci is the fraction of material coarser than sieve i and ki is its breakage rate per minute. Grinding the unknown ore for two or three times and sizing each product gives enough points to fit ki for every sieve by least squares. With those rates, the product that the unknown ore would have given from the reference ore's feed size distribution, ground for the reference ore's time, can be calculated. The two ores are then compared on equal terms, regardless of the actual feed size of the unknown sample.
Procedure
- Crush the reference ore and the plant sample to minus 10 mesh (1.7 mm) and size both.
- Batch grind the reference ore for several times to find the time that gives the target product, then size that product.
- Batch grind the plant sample for at least two, preferably three, times that bracket the reference grind time, and size each product.
- Fit a breakage rate for each sieve from the plant sample's results.
- Apply those rates to the reference ore's feed size distribution for the reference grind time, giving the plant sample's calculated product.
- Read the 80% passing sizes and calculate the work index with the comparative equation.
Every condition other than the ore must remain identical between grinds: charge mass, pulp density, ball charge, mill speed and the mill itself. In the paper, each grind used 1,000 g of minus 10 mesh ore at 60% solids in a 203 mm by 203 mm laboratory ball mill, with 8.1 kg of 31.8 mm balls, at 75 rpm.
Worked example from the paper
The reference ore had a work index of 14.3 kWh/t. Both feeds were taken as 1,090 µm at 80% passing. After the reference grind time of 8 minutes, the reference product was 202 µm and the plant sample's calculated product was 261 µm:
The breakage rates can be checked directly. On the 35 mesh sieve, the reference feed had 58.0% retained and the fitted rate was 0.348 per minute. After 8 minutes:
where
- percentage retained on the 35 mesh sieve after 8 minutes
This leaves 3.6% retained, or 96.4% passing, which matches the paper's calculated product.
Agreement with the standard test
The paper compared the method with the full Bond test on six copper ores ranging from about 10 to 20 kWh/t. The two agreed within 0.4 kWh/t on every ore, whereas the uncorrected comparative method differed by up to 2.8 kWh/t. Samples of SAG mill feed and of screen undersize from the same plant, crushed to minus 10 mesh, also gave work indices within 0.4 kWh/t of each other.
Work index as a plant performance baseline
The plant application in the paper remains the most useful part of it. At the Cyprus Pima copper concentrator in Arizona, SAG mill throughput was plotted against SAG mill specific energy, with laboratory work index measured on samples taken during the same periods. As ore hardness rose, throughput fell along a consistent curve, which became the operating baseline for the circuit.
When a computer-based control system was installed on the grinding circuits in 1974, the new operating data could be compared with the baseline for ore of the same hardness. The scatter about the curve fell noticeably and throughput per grinding section rose, with the largest gains on the softer, faster-milling ores. Because the comparison was made at equal ore hardness, the improvement could be attributed to the control system rather than to easier ore.
The same approach applies to any circuit change today. The baseline should be established first, from throughput, mill power and a hardness measurement collected over enough shifts to cover the normal range of ore. The hardness test must remain consistent, with one laboratory method, one mill and one reference ore throughout, because a change in test method is indistinguishable from a change in ore.
After the change, performance is judged against the baseline at the same work index, not against the previous month's average. Product size should be checked as well as tonnage, because a throughput gain obtained with a coarser grind may cost more in recovery than it gains in tonnes.
Limitations
The method assumes first-order breakage. The authors describe it as suitable where linear breakage kinetics apply, as they did for the ores tested, and timed grinds should be confirmed to plot as straight lines before the rates are relied on.
Results are most reliable near the reference ore's grind. The further the target product is from that size, the more the work index may differ, because many ores break differently at different sizes.
A Bond ball mill work index describes ball milling. It is a useful hardness indicator for a SAG circuit, as the paper shows, but SAG design today also relies on tests directed at coarser breakage, such as drop weight testing.
The reference ore must be stable and well characterised. Enough of it should be retained for years of testing, and its standard work index should be rechecked periodically.
In Saldora, hardness results can be recorded in a daily measurement set alongside throughput and mill power, so that the baseline builds up as the plant operates. Ajax includes crushing and grinding unit models for testing a circuit change on a plant model first.
Glossary
- Bond work index
- A measure of how hard an ore is to grind, in kWh per tonne.
- Grindability
- How easily a material breaks in a grinding mill.
- Reference ore
- A well-tested ore with a known work index, used for comparison.
- Breakage rate
- How fast material coarser than a given size disappears during grinding, per minute.
- First-order kinetics
- A process in which the rate of change is proportional to the amount remaining, giving an exponential decay.
- SAG mill
- A semi-autogenous grinding mill, which uses the ore itself plus some steel balls as grinding media.
- Specific energy
- Energy used per tonne of ore, in kWh/t.
- Operating baseline
- The normal relationship between throughput and ore hardness, used to judge later changes.
Frequently asked questions
How accurate is the rapid method?
In the original study it agreed with the full Bond test within 0.4 kWh/t on six copper ores.
Can it be used for SAG mill design?
It is a useful hardness indicator for monitoring, but SAG design today also relies on tests directed at coarser breakage, such as drop weight tests.
Why not compare throughput before and after a change directly?
Ore hardness changes as well. Comparing at equal hardness separates the effect of the change from the effect of the ore.
What makes a suitable reference ore?
One with a well-established standard work index, available in sufficient quantity to last years of testing.
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
- Horst, W.E. and Bassarear, J.H. (1976). Use of simplified ore grindability technique to evaluate plant performance. SME Transactions, Vol. 260, December 1976, pp. 348–351.
- Berry, T.F. and Bruce, R.W. (1966). A simple method of determining the grindability of ores. Canadian Mining Journal, July 1966, pp. 63–65.
- Bond, F.C. (1952). The third theory of comminution. Transactions AIME, Vol. 193, pp. 484–494.
- Wills, B.A. and Finch, J.A. (2016). Wills' Mineral Processing Technology, 8th edition. Butterworth-Heinemann. Chapter 5, Comminution.
Worked examples and charts marked illustrative use invented figures, not data from any operation.