You are on page 1of 12

UPGRADING OF AUSTRALIAN MAGNETITE

OREBODIES

BY
B C POVEY AND K G LEATHER

Brian C Povey
Fellow graduated in 1971 from the Royal School of Mines in London with a BSc(Eng) in
metallurgy. He worked in a variety of metallurgical positions throughout Australia before
joining Savage River Mines. At Savage River he filled a number of roles in the technical and
managerial field over a fifteen year period and is currently a Metallurgical Consultant with
Kvaerner Metals in Perth.

Keith Leather
Fellow graduated from Sheffield University in 1964 and worked for ten years in Southern
Africa in a variety of operating and managerial roles before emigrating to Australia. He has
worked on numerous engineering projects and is currently Metallurgical Manager for Kvaerner
Metals (Australia).

Address: Tel:
Kvaerner Metals (08) 9429 5000
PO Box 7226
Cloisters Square
Perth WA 6000

1
A number of magnetite deposits are currently being investigated for potential development to
produce DRI/HBI products. These deposits are relatively low in iron content but are easily
upgraded by low intensity magnetic separation to a concentrate low in silica and alumina and
producing pellets which exceed 67.5% iron. The metallurgical investigation of these deposits
uses a Davis Tube apparatus to assess the quantity of magnetite. The Davis Tube effectively
models the expected plant performance for each determination giving both a yield, grade and
impurity estimate for each sample. To gain the full value of such a procedure the technique
has to be customised for the ore body being tested and calibrated with other laboratory
techniques. Results from a case study are presented showing a practical application of the
data.

In world terms the well known magnetite deposits are the taconite deposits in Minnesota and
Michigan, which provide 80% of the raw material input to the American Steel industry. In
Australia the major exploitation has been at Savage River in Tasmania which produced more
than 50Mt of pellets over a 30 year life and was the only Australian pellet plant which survived
the competition in the international market and the oil price crises of the seventies and early
eighties.

A number of significant magnetite deposits are known in Western Australia and two of these,
the Mt Gibson and the Fortesque deposits, are being studied as the ore source for HBI/DRI
production units. The cost of developing these projects to produce blast furnace grade iron ore
pellets is prohibitive due to the low added revenue value for the pellets in the iron ore market
and the distance of the projects from an ocean going transport system. However, the potential
for the cost effective upgrading of the concentrate to a level which will make high grade
HBI/DRI adds considerably to the value of the ore body and, linked to such plants, these
projects now appear promising.

The typical iron grade of a magnetite deposit is 25 - 35% Fe. The deposits are rarely discussed
in terms of % Fe but more commonly in terms of the weight recovery (Wt Rec) which is the
ratio of final concentrate to initial feed. An exploitable deposit would vary in Wt Rec from 35 -
50%. The Wt Rec is normally measured by the use of the Davis Tube apparatus shown in
Fig 1.

2
Fig 1 - Davis Tube Apparatus

A sample of pulverised ore is washed by a flowing stream of water in a medium strength


magnetic field of about 5000Gauss. The tube is oscillated in the field for five minutes to
improve the washing and the non magnetic fraction is flushed from the ore. The magnetic field
is then switched off and the concentrate is collected for drying and assaying. The weight of
concentrate compared to the feed weight gives the weight recovery and the iron assay of
concentrate is the potential plant product grade. Gangue minerals are also measured on the
concentrate since the level in the original ore bears no relation to the expected level in the final
product.

The Davis Tube measurement is thus a very important part of the analysis of a magnetite ore
body and it has the tremendous advantage of being a mini pilot plant run on each
determination. This in turn means that the preparation of the sample for the Davis Tube has to
be more controlled than would normally be the case during the pulverisation of a sample for
conventional assaying. The grinding of the sample must produce a size distribution equivalent
to that expected in a full scale plant. A sample which was pulverised for thirty seconds might
yield an Fe assay in concentrate of 56%Fe with a Davis Tube recovery (DTR) of 53.3% but
the same sample ground for 60 seconds might yield a grade of 64.7%Fe with a DTR of 46% -

3
quite a significant difference in terms of analysing the ore body and leading to quite different
conclusions about the viability of the ore body.

Since the Davis Tube does simulate the plant it is important therefore that the procedure is
customised for a particular ore body and that the adopted procedure is rigorously followed.
The laboratory must understand the reasons for the procedure and ensure they don't streamline
or alter the process at some stage which may save laboratory time but fundamentally alters the
size distribution of the sample. If this occurs then a step change will result in the grade and
weight recovery of the ore body which renders one of the sets of data unusable.

Development of the DTR Method

This paper will now show the typical development of the DTR method and its application with
reference to an actual study. The ore body that was being investigated had been categorised as
containing four ore types - 1, 2, 3, and 4 mainly based on a geological interpretation of the %
of chert in the bands between magnetite layers. Each of these ore types was to be defined in the
ore body and in the early stages of investigation it was decided to test them separately.
Subsequently types 1 and 2 were combined for testing due to their similar nature and test
results. The project was targeting to achieve a concentrate assaying 69.5%Fe which would
ultimately produce a fired pellet with an Fe assay of 67.5%Fe. From previous sighter testwork
it was known that the material was fine grained with the magnetite grain sizes between 20 and
40 micron and the pulverising technique would be critical in determining the grade and
accurate weight recovery of the ore.

During the first part of the DTR work a method was developed which could give the required
grade consistently. The ore was ground using a laboratory pulveriser and it was recognised
that the developed method must be suitable for use on a large number of drill hole samples as
this was to be used as the standard method for the analysis of the whole orebody. A testwork
programme was set up with the laboratory comparing a single stage pulverising technique with
a two stage technique. In the two stage technique the pulverisation is done for a fixed time but
then the sample is dry brushed through a 75micron screen. The undersize is set to one side
while the oversize is repulverised for a period of time depending on the weight of oversize

4
material. The two streams are then recombined and mixed before sample selection for the
Davis Tube test.

The results for the two techniques on one of the ore samples is shown in the tables below with
Table 1 the data for the single stage pulverising and Table 2 the data for the two stage
pulverising

Table 1 - Single Stage Pulverisation


Pulverising Time secs %-45µm DTR Wt Rec %Fe

30 69.0 48.3 59.1


45 74.0 45.9 61.7
60 77.0 43.1 64.5
70 78.0 43.4 64.9
90 82.0 42.4 65.7
120 81.4
150 80.8
180 79.6

Table 2- Two Stage Pulverisation


Initial Pulverising %-45µm after 2 DTR Wt Rec %Fe
Time secs stages

30 82.0 40.4 68.0


45 84.0 40.5 68.0
67 85.0 41.0 67.0
70 87.0 39.3 67.6

The data shows that extending the grinding time on the single stage technique does not
necessarily increase the % -45µm but clearly leaves coarse unground particles which lower the
grade and increase the weight recovery of the ore. By contrast the two stage technique
produces more stable results with lower DTRs and higher grades suggesting a more reliable
means of eliminating the coarse oversize. This would be particularly important where the waste
rock was significantly harder than the magnetite and where the hardness of the waste rock
varied.

5
In the case study the 70 second two stage grinding technique was chosen since that gave
acceptable results for the predominant ore types of 1 and 2 as shown in Table 3

Table 3 - Grades for all Ore types


Ore Type DTR Wt Rec %Fe

1+2 64.3 70.0


3 40.5 67.7
4 41.0 67.6

The technique chosen did not allow the type 3 and 4 ores to reach the required grade due to
the finer grain size of these ore types. However, the use of the same DTR method for each ore
type avoided any bias from the geological interpretation which was subjectively decided in the
field. To compensate for the low grades and slightly high weight recoveries for the type 3 and
4 ores a number of DTRs were carried out on these ores using a 45 micron screen rather than a
75 micron screen. This gave an adjustment factor to the DTR results in terms of yield and
grade to compensate for the fact that these ore types would need a finer grind in practise to
achieve the required grade. This technique allowed the laboratory to use the 75micron screen
for routine samples and avoided the problems of dry screening at 45micron on a large number
of samples.

Development of the Flowsheet and Optimum Grinding Conditions

A typical magnetite concentrator flow sheet starts with a coarse grinding section which reduces
the ore to a size where the magnetite is concentrated in rougher magnetic separators. This
concentrate is then reground to the optimum grind size for the final achievement of grade. The
reground concentrate would be deslimed and reconcentrated in a cleaner style counter current
magnetic concentrator. This may be the final concentrate or a further flotation stage may be
introduced to lower silica levels to maximise the quality of the magnetite concentrate. These
levels are typically 2 - 2.5% SiO2. The development of the practical circuit grinding conditions
requires the use of laboratory grinding mills to find both the initial release size for rougher
concentration and the optimum regrind size for concentrate grade. The yield and grade at
various grinds is measured for the samples and results related back to the ore body using the
data from the same samples tested by the Davis Tube method as a fixed reference.

6
Indicative Grind Target

The first step in the development of the process circuit uses the grind data from the DTR
weight recovery testwork plotted against the %Fe. This gives an indication of the expected
grind required. Figure 2 shows the grinding data developed during the DTR testwork in the
study and the effect of grind on the grade of concentrate. While these data refer only to the
type of grinding which occurs in the pulverising unit rather than a ball mill, it is indicative of
the expected behaviour in a more practical situation.

Fig 2 - Change of Grade with Grind - DTR Work

Type 1 and 2 Ore

70.0
% Fe in Concentrate

Type 4 Ore

Type 1 + 2

60.0 Type 3

Type 4

Type 3 Ore Type 1 and 2 Trend

Type 3 Trend

Type 4 Trend

50.0
60 70 80 90 100
% -45micron

This graph shows the grade of iron increasing as the grind increased for the type 3 and 4 ores.
In contrast the type 1 and 2 ore are consistently at grade over a range of pulverisation sizings.

Initial Concentration Size

The initial concentrating grind size can be developed by taking a sample of ore and grinding it
for various times to assess how the magnetite yield varies in relation to the weight loss. In this
study the ore was known to have a preconcentration size of 75 micron from earlier work but it

7
was decided to test this further. Three grinds were chosen 80% passing 200 micron, 80%
passing 100 micron and 80% passing 75 micron. The ore was then passed through a laboratory
electromagnetic separator with the magnetic field strength set to 1000 Gauss. The concentrate
and tails products then were tested by the two stage pulverisation DTR test which
approximated the final grind size. The DTR test in this case could be interpreted as a means of
measuring the % of magnetite in each product and hence a recovery of magnetite depending on
grind. A measurement of the % magnetite can also be made by using a Satmagan balance
which measures the magnetic susceptibility of a sample. A drawback of this technique is that it
can give inconsistent results if the ore is at various levels of oxidation - a common problem in
practical ore samples.

From the results the loss of magnetite yield at the various grinds can be found. This data is
combined with a knowledge of the power costs of the potential circuits and the cost of yield
losses to determine the best initial concentration size. In many cases the impact of yield losses
swamps the higher power costs of grinding finer.

Optimum Grind Size

The optimum grind size is the predicted grind required to give the concentrate grade required
for the project and is determined in a similar manner to the initial concentration size. The ore is
ground to a variety of sizes and is then concentrated in a laboratory scale magnetic separator.
In this study the grind sizes chosen were 80% passing 53µm, 80% passing 45µm, 80% passing
38µm and 80% passing 32µm. The testwork was carried out at a magnetic field strength of
1000 Gauss and, while the issue is not discussed in this paper, it is frequently useful to repeat
the work at varying Gauss levels. This allows the development of useable grade/recovery
curves in a similar manner to those that may be produced in investigative flotation work.

In this study and in several similar studies the concentrates from a single stage magnetic
concentrator are quite dirty and not indicative of what would be achieved by a full flow sheet
with proper desliming and concentrate washing. This can lead to an inconsistency in results
depending on the entrainment of the non magnetic portions in the concentrate. To simulate the
benefit of normal concentrate washing the material is therefore washed in a Davis Tube. The
sample is taken as collected without further pulverising and placed in the Tube. Water is

8
flushed through while the tube is oscillated in the magnetic field in the normal manner. In effect
the Tube is just used to provide a consistent means of holding the concentrate sample for
washing without risk of material loss likely to occur with simple decantation.

As well as the issue of contamination from dirty water there also the problem of the accuracy
of the chemical analyses at the high iron levels encountered in concentrate streams. While the
iron assay has a low absolute error relative to the gangue minerals the relative error is higher.
Consequently the interpretation of results from this point has to change from using the %Fe
assay to using the %SiO2 assay since the analysis of high iron levels in concentrate can not
provide the level of accuracy required. Once the % Fe is above 65 %Fe the assay accuracy
quoted as 0.5% absolute error means an error of +/-0.3%Fe can be expected. By contrast the
SiO2 assay at 2% absolute error is accurate to +/- 0.04% SiO2 - much more accurate for the
purpose.

To find the optimum grind the 80% passing size is plotted against the equivalent %SiO2 in the
washed concentrate. It is frequently possible to extend the curves by combining the results of
both the preconcentration grind and the optimum grind testwork into a single data series for
plotting. For the study in question the data fitted a exponential trend curve with a correlation
coefficient of greater than 0.97 - sufficient to allow it to be used to extrapolate the data to the
required grind size. Figure 3 shows the correlation curves and the close fit using an exponential
curve can be seen. Plotting the grind size on a log scale would give a straight line.

9
Fig 3 - 80% Passing Size Comparing Lab Grind, and DTR

200
Type 1 and 2

Type 3

Type 4
Type 1 and 2 Ore
150
80% Passing - micron

Type 1 + 2 Trend

Type 3 Trend Type 4 Ore

Type 4 Trend

100

50
Type 3 Ore

0
0 5 10 15 20
% Silica in Concentrate

Relationship to the DTR Results

The next stage of development is to relate the DTR results to the grind data so that the
required grind size for any given DTR result can be predicted. In this study although the ore
had been treated as three ore types it was realised that the actual DTR data showed that the
ore body was a continuum and not several discreet ore bodies with different concentrating
behaviours. The ore types were thus a convenient grouping for a range of common behaviour
rather than discreet individual responses. This meant that the results from the nominal ore
types could be linked and the behaviour of intermediate material predicted. The level of silica
in the DTR concentrate was sufficient to characterise the sample and allow either a
determination of the silica level for a given grind or vice versa. Figure 4 shows that this has
been developed from the data in Figure 3 to give a set of grind curves which can be used for
plant design.

10
Fig 4 - Expected % Silica in Concentrate based on %Silica in DTR Test for Various
Grinds
14
80% -53u
12
% SiO2 expected in Concentrate

80%-45u
10
80%-35u

8 80%-30u

80%-25u
6
DTR
4

0
0 1 2 3 4 5 6 7 8
% SiO2 in DTR Concentrate

Thus if the %SiO2 in a particular sample DTR is 5% then to achieve a 3%SiO2 in concentrate it
would need to be ground to 80% passing 30 micron. This graph also shows that the DTR
method itself could be regarded as a grind equivalent to a ball mill grind of 80% passing
35micron since this is the grind size where the % SiO2 in DTR is the same as the % SiO2 in
concentrate.

Conclusion

The paper has outlined the importance of establishing a consistent DTR method at the start of
a project and how the results of this procedure can be used as a reference for any of the ores in
the series. By using the data and then correlating it to actual laboratory grinds it allows the
development of design criteria for the plant design and ultimately a performance yardstick
when the plant is fully operational.

11
Acknowledgments

This work was supported by Mt Gibson Iron NL and Kvaerner Metals and their support is
acknowledged. The authors are grateful to Mt Gibson Iron for the use of some of the test data
results from their testwork programme.

References

1. Langenburg JT - Savage River Mines Private Correspondence


2. SME - Mineral Processing Handbook - N.L Weiss - Editor

12

You might also like