Advanced Mineral Processing and Advanced Mineral Processing Techniques

Laboratory Report Cover Sheet

Group No.: 3

Laboratory No.: 1-5 labs

All forms of plagiarism, cheating and unauthorized collusion are regarded seriously by the
University and could result in penalties including failure in the Unit and possible exclusion from
the University. Please read the Curtin University Statement on Plagiarism. If you have any
doubts please contact the Unit Co‐ordinator.

Declaration

Except as indicated, this work being submitted is solely my/our own work.

Name (s) Student Id Experiment E-Signature

Barry Sanders 21216752 1 BS

Ray Adhitya Rachman 22170163 2

Qingzheng Zhang 20058584 3

Steven AndrewSondakh 22460974 4

Connor Vink 20378583 5

Due Date: 13/10/2024

Date Submitted: 11/10/2024

Office use only:

Date Received
Recorded by
Marked by
Date Marked
 Determination of the Bond Work Index
Introduction
Comminution is the first step in liberating the locked mineral from the gangue. This is achieved
by crushing and grinding to reduce the particle size and increase surface area. The particle size
reduction is directly proportional to the energy input and this relationship is studied in
comminution theory (Wills and Atkinson 1993). This experiment focuses on theory by Bond
(1952), which is based on the relationship between the energy input being proportional to the
particle breaking to produce a new crack tip length, this means it is comparing the product and
feed size reduction. This is shown in the bond work index for a ball mill:

Where:
Wi = Bond work index (kWh/t)
S = Test Sieve mesh size (µm)
G = Average gram per mill revolution for the last three cycles (g/rev)
F80 = The 80% passing size fraction of the feed (µm)
P80 = The 80% passing size fraction of the fine products (µm)

The theory was conducted under certain tests conditions, for instance the feed must be below 3.5
mm, the ball mill was tested at a 250% recirculating load at 70 RPM and the charge was
compacted into a 700 cc container to obtain the peak mass density (Wills and Napier-Munn,
2006). The test sieve mesh size (S) was the size that the sample was sufficiently liberated and
chosen at 180 µm. The mass of test sieve undersize produced per mill revolution (G) was the
average of the last three calculated mass of fines per mill revolution once the cycles for the
recirculating load became steady.
The particle analysis to determine the passing % size fraction over sieve sizes can be modelled
graphically to produce a straight line by Gates – Gaudin – Schuhmann (GGS) or Rosin –
Rammler (RR) models. Both the feed and product size analysis fit the GGS model which was
plotted cumulative percent passing against sieve sizes with log-log axis. The equation is (WASM
2024):

Where:
Y = The cumulative % mass passing
x = size of the sieve aperture (µm)
k = size parameter
a = distribution parameter
The log x against log Y graph produced a straight line on log-log axis while discounting values
near the 100 % passing. The slope of the line was the distribution parameter (a) and k was
calculated at 100 % passing using the equation of the line. The k value is different to
experimental data as it was taken from the linear extrapolation. With all known values, the x =
F80 and P80 can be determined using the GGS model, Eq. 2, when Y = 80 % cumulative mass
passing for the feed and product data.
The objective of the experiment was to determine the relationship between work input and
particle size reduction which will be analysed using the bond work index for a ball mill (Eq. 1).
Another objective is to find the cumulative 80 % passing using graphical linear models (GGS
and RR). It is hypothesised that the bond ball mill work index will be calculated from the
experimental data and to achieve a 250% re-circulating load, at least 3 attempts will be needed to
reach a constant rate.

Materials and methods

Material:
• 4 lots of - 3.35 mm ore sample
• Ball mill unit
• Weigh balance (+/- 0.001 g)
• A total of 285 steel balls for the ball mill that weighed 20.125 kg, detailed below Table 1.
Table 1: Ball mill experimental steel balls.
Ball Diameter (mm) 38.1 31.75 25.4 19.05 15.87
No. of Balls 43 67 10 71 94
• 1 L plastic measuring cylinder to pack 700 mL of sample.
• Feed sizing analysis using mesh sizes (μm): 2000, 850, 500, 180 and 75.
• Product sizing analysis using mesh sizes (μm): 180, 125, 75, 63 and 53.

Method: A scheme for the grindability of ore in a ball mill was conducted in 4 stages:
Sample Preparation → Locked Cycle Grinding → Product Screen → Calculate Wi.
Sample Preparation:
1) Ensure 285 steel ball charge of 20.125 kg, detailed above in Table 1.
2) Split the pre-crushed (-3.35 mm) sample into 4 even samples.
3) Use one of the samples to determine the bulk density of feed (g/ml), by filling a 1 L plastic
measuring cylinder to fill 700 cc with the sample to maximum density by shaking and
compacting. This 700 cc charge weighed 1334 g and will be used in Step 7.
a. Ideal Product Mass (IPM) for a 250% circulating load mass (commonly used in testing)
was determined by dividing the charge (1334 g) by 3.5.
4) Conduct a feed size analysis of one the samples using mesh sizes (μm): 2000, 850, 500, 180 and
75.
a. The cumulative passing (%) graph can be produced to obtain the 80% passing size value
F80 (microns).
b. Determine the percentage of fines in the feed (~18 % for this experiment) by dividing the
total mass of fines (-180 μm sieve size) by the total mass.
 c. The mesh grind size in this particular test was 180 μm (S, value in the Wi equation).
Locked cycle grinding:
5) Make another charge similar in weight to Step 3. Place into the ball mill with the steel ball
charges and secure the lid.
6) Grind the sample for 100 revolutions at 70 rpm. After, remove the contents and separate the
product and balls.
7) Use a 180 μm sieve while taking care not to overload it and determine the mass of oversize (O/S)
and undersized (U/S) fractions. Separate the fines into a separate container and replace the lost
weight with new feed from Step 3 for the next cycle.
a. Calculate the mass of fines in the charge by multiplying the charge mass by the
percentage of fines in the feed from Step 5 (~18 %).
b. The net fines produced was calculated by subtracting the calculated feed fines (Step 7a)
from the U/S fraction in Step 7.
c. Calculate the net mass of fines per revolution (g/rev), divide the net fines produced from
Step 7b by the number of revolutions.
d. Calculate the number of revolutions for next cycle, by subtracting the mass of fines in
charge (Step 7a) from the IPM (Step 3a) and all divided by the net mass per revolution
(Step 7c)
8) Repeat from Step 6 using the new revolutions calculated in Step 7d until reaching steady state,
when the net mass of fines per revolution in Step 8c becomes constant (+/- 0.02 g/rev). Usually,
it’s 6-8 more cycles but this setup only required 2 more cycles.
9) Take the average of the last 3 cycles for the net mass of fines per revolution (g/rev). This value is
the G value in the Wi equation.
Product Screen:
10) Conduct a product size analysis of the fines from the last 3 cycles using mesh sizes (μm): 180,
125, 75, 63 and 53.
a. The cumulative passing (%) graph can be produced to obtain the 80% passing size value
P80 (microns).
Calculate Bond Work Index:
11) Calculate the Work Index using the formula:

Results and Discussions

The results and discussions follow the scheme for the grindability of ore in a ball mill:
Sample Preparation → Locked Cycle Grinding → Product Screen → Calculate Wi (Eq.1).

The sample preparation consists of feed size analysis to determine the feed cumulative 80%
passing, F80. The locked cycle grinding determines the fines products from the ball mill at a
steady state. A size analysis was conducted on the milled products to determine the product
cumulative 80% passing, P80. Finally, the Work index was calculated to measure the grindability
of the sample ore.
Bulk Feed Size:

The bulk feed size analysis data in appendix Table 3, and the graphed cumulative percentages are
in Figures 1 and 2. Figure 1, illustrates the cumulative % passing in blue and retained in green.
They are direct opposites of each other and the Gates – Gaudin – Schuhmann (GGS) model uses
the cumulative % passing, while the Rosin – Rammler (RR) model analyses the cumulative %
retained. For easier interpretation of the size analysis, the experimental (blue and green) and a
linear line (green) is formed on the log-log scale for cumulative % passing vs. sieve sizes, shown
in Figure 2.

Figure 1: Homogenous bulk feed cumulative % passing / retained for values in Table 3.

Figure 2: The bulk feed cumulative passing (%) using Log Scale and trendline of the curve.

In Figure 2, the linear green line represents the distribution modulus, the a-value in GGS model
Eq. 2. This distribution parameter indicates the degree of size distribution and is 0.0378 as seen
from the slope in Figure 2. The slope was extrapolated to Y = 100 % passing, achieving a
maximum size parameter (k-value in Eq. 2) of 2390 µm from the modified linear equation:

The 80% cumulative passing F80 was calculated using the GGS model (Eq. 2) with the known a
and k values:

The Feed 80% cumulative passing F80 was 2410 µm.

The fines in the feed was recorded as below 180 µm, this is the S value in Eq. 1. The % of fines
in the feed was calculated using the linear equation in Figure 2:

The amount of fines in the bulk feed was approximately 18 %.

Locked Cycle Grinding:

The 3 ball mill cycles needed to achieve 250% load at steady state are shown in Table 2. The
weight of the feed charge (compacted sample of 700 cc) was 1334 g, was initially run at 100
revolutions in the ball mill. The First cycle mass fraction of oversize (O/S) and undersize (U/S)
was 859 and 473 g respectively in Columns C and D of Table 2.

Table 2: Ball mill load cycle grinding steady state

A B C D E F G H

Weight of Weight of Weight of Expected Weight Net Fines

Test No. of Revs new O/S in U/S in weight U/S U/S in feed Produced Gram/ Rev
feed(g) Product (g) Product (g) (g) (g) (g)

1 100 1334 859 473 381.14 240.00 233.00 2.33

2 127 1330 1070 260 381.14 85.14 174.86 1.38
3 200 1325 1007 318 381.14 46.80 271.2 1.36
Average Net Grams/Rev(G) 1.69
The expected weight of U/S (IPM) was calculated in Column E of Table 2, based on 250% re-
circulating loads:

The U/S of the feed was calculated in Column F of Table 2, using the % of fines in the bulk feed
(~18%), then applied to the charge weight of 1334 g to give an expected weight of U/S 240.00 g
for the first test.

The net mass of fines produced by the ball mill was calculated in Column G of Table 2, by
subtracting the initial feed fines (Column F) from the overall (Column D):

The mass of fines produced per revolution is an important metric in grinding efficiency, and
calculated in Column H of Table 2:

In Test 1, a total of 2.33 g/rev was produced in the 100 revolutions at 70 RPM grinding from the
1334 g charge.

New tests with a modified number of revolutions are needed to achieve a constant product when
the gram/rev (column H) is constant. The new revolutions were calculated in subsequent tests in
Column A of Table 2, using:

Tests 2 and 3 repeated the steps from Test 1 with a revised revolutions and the removal of fines
was replaced with new feed to keep a constant weight of charge. Test 2 was conducted at 127
Revs (Column A). The third test had a similar gram/Rev produced to the previous test, thus
reaching steady state. The average of the last 3 cycles is considered the steady state condition,
representing the G value in the Work Index (Eq. 1) that gave 1.69 g/rev.

Average Fine Product Sizing:

The final 3 cycle fines were size analysed in the appendix Table 4, and graphed in log-log scale
for cumulative % passing vs. sieve sizes in Figure 3. The linear line (green) formed the linear
portion of the size distribution and the last point (blue) deviated from the trend.
 Figure 3: Average fines ball mill product cumulative passing (%).

The distribution parameter, the a-value in the GGS model (Eq. 2), was represented graphically,
Figure 3, as the slope of the linear portion, green line to get 0.6381. The slope was extrapolated
to Y = 100 % passing, achieving a maximum size parameter (k-value in Eq. 2) of 192.4 µm from
the modified linear equation:

The 80% cumulative passing P80 was calculated, using the GGS model (Eq. 2) with the known
‘a’ and ‘k’ values:

The Product 80% cumulative passing P80 was 135.6 µm.

Work Index:

The bond work index can be determined by finding the Bond’s Ball Mill Work Index Equation
(Eq. 1) from experimental data at steady state. The S value was the maximum sieve size for
acceptable fine particle size and was found in the feed analysis to be 180 µm. The 80% passing
for feed and products used the GGS model to get 2410 and 135.6 µm, respectively. The average
g/rev, G, for the last 3 cycles was determined to be 2.33 g/rev. These values were substituted into
Eq. 1:
The Bond Work Index, Wi was calculated to be 14.8 kWh/t, which is higher than quartz, 14
kWh/t. The higher the Wi, then there is larger resistance to grinding for the sample and need for
more energy. The sample was assumed to be quartz or quartz similar due to availability and low
costs.

Conclusions
The objective, was to determine the relationship between work input and particle size reduction
was analysed using the bond work index in a ball mill which was 14.8 kWh/t. This was achieved
by the size analysis of feed and products using the GGS model to determine the F80 of 2410 µm
and P80 of 135.6 µm. The RR model was not necessary as the results were linear. The grind only
took three tests to achieve steady-state for a fine product of 2.33 g/rev. The grindability was
higher than quartz, which is the likely sample due to availability and cost. The higher
grindability index could be due to differences in ball mills, efficiency and a larger and stronger
sample. In the future, more sieve sizes for analysis to better understand the size distributions, this
however takes up time and is not recommended for teaching labs.

References
- Bond, F.C. 1952. “The third theory of comminution”, Trans. AIMF, 193, 484.
- Wills, B.A. and Atkinson, K. 1993. “Some observations on the fracture and liberation of mineral
assemblies”, Minerals Eng., 6(7), 697.
- Wills, B.A., and Napier-Munn, T.J. 2006. “Wills' Mineral Processing Technology: An
Introduction to the Practical Aspects of Ore Treatment and Mineral Recovery.” 7th Ed.
Oxford: Butterworth-Heinemann.
- WASM Curtin University. 2024. Lab Manual. Blackboard.
Appendices
The size analysis for bulk feed and fine products are shown in Table 3 and 4, respectively.

Table 3: Ball mill feed size distribution and calculated cumulative (%) retained and passing.
Mass Retained Cumulative % Cumulative %
Sieve range (µm) Mass (%)
(g) Retained Passing
+3350 0 0 0 100
-3350 to -2000 229 38 38 62
-2000 to +800 143 24 62 38
-800 to +500 47 8 69 31
-500 to +180 71 12 81 19
-180 to +63 57 9 91 9
-63 56 9 100 0
Total 603

Table 4: Ball mill product sieve size distribution and calculated cumulative (%) retained and passing.
Mass Retained Cumulative % Cumulative %
Sieve range (µm) Mass (%)
(g) Retained Passing
180 76 14 14 86
125 161 30 44 56
75 135 25 70 30
63 82 15 85 15
53 31 6 91 9
0 50 9 100 0
Total 535
 Experiment 2
Fine Particle Sizing by Cyclosizer and Laser Sizer

Part I - Cyclosizer
Introduction
In mineral processing, accurately classifying particles by size is essential for optimizing
processes such as flotation, leaching, and separation. The size of the particles affects recovery
rates, efficiency, and the overall effectiveness of these processes. Traditional methods, such as
sieving, become less effective for particles smaller than 63 micrometers. Therefore, alternative
methods like elutriation, which separates particles based on their settling velocity in a fluid, are
commonly used to handle finer particle sizes.

A cyclosizer is a specialized instrument that uses elutriation in water to classify fine particles by
size, using centrifugal force in reverse cyclones. The cyclosizer is particularly effective for
classifying particles in the range of more or less 11 to 44 micrometers, making it one of the ideal
ways for fine particle separation in mineral processing.

Hypotheses:
1. The cyclosizer will effectively separate the quartz sample into distinct size fractions, with
most particles falling between 33 and 15 micrometers.
2. The particle size distribution will show that the larger fractions contain more mass than
the smaller fractions.
Objectives:
1. To analyze the particle size distribution of a quartz sample using the cyclosizer and
determine the mass of each size fraction.
2. To evaluate the effectiveness of the cyclosizer in separating fine particles and its
relevance to mineral processing applications.

*Due to significant errors encountered during the data collection process, the analysis in this
report will primarily rely on the ideal data provided through the course's website. Our
experimental data collection was affected by factors such as incomplete drying, improper
handling, and miscalculations of filter paper weight, leading to a total mass measurement that
exceeded the original 30-gram sample. Additionally, the laser sizer analysis was not performed
during our experiment. As a result, the provided data has been used for a more accurate analysis,
while the discrepancies in our collected data are discussed to highlight potential sources of
error.*
Materials and Methods
Equipment:

- Cyclosizer (Warman)
- Quartz Sample
- Thermometer
- Beakers
- Aluminum Trays
- Filter Papers
- Digital Balance
- Drying Oven

Methods:

1. Before starting the experiment, the temperature of the water from the water system was
measured. The water pressure was then adjusted to the right level for the cyclosizer to
work properly. Once the water system was set, 30 grams of quartz was prepared. The
quartz was sieved to make sure all the particles were smaller than 63 micrometers.

2. The quartz sample was split into smaller portions and put into the cyclosizer. The
cyclosizer, which has five cyclones, was used to separate the particles into size ranges of
about 44, 33, 23, 15, and 11 micrometers. Water was used to carry the particles through
the cyclosizer, where they were sorted based on size.

3. After the separation, the particles collected from each cyclone were placed into beakers.
The water was allowed to settle, and then it was poured off, leaving the solid particles
behind. These particles were then transferred to aluminum trays and dried in an oven to
remove any remaining moisture.

4. Once dry, each fraction was weighed using a digital balance with a precision of 0.01
grams. The weights were recorded, and the particle size distribution was analyzed using a
model called the Gaudin-Schuhmann distribution, which helped calculate the cumulative
mass percentage for each size group.
Results and Discussion

Results
Particle Size Mass of each fraction Cumulative Mass Cumulative Mass Percentage
no. (μm) (g) (g) (%)
1 44 2.5 2.5 8.34
2 33 4.3 6.8 22.67
3 23 3.6 10.4 34.67
4 16 2.5 12.9 43
5 12 0.5 13.4 44.67
Table 1.

figure 1. Cumulative Mass v Particle Size for Cyclosizer

The cumulative mass percentage increases sharply from 8.33% at 44 µm to 22.67% for particles
between 33 and 44 µm, indicating a significant portion of the sample in this range. As the
particle size decreases, the increase in cumulative mass slows, reaching 43.00% for particles
between 16 and 23 µm, suggesting fewer particles were recovered in the finer fractions. The
section at 44.67% for particles between 11 and 16 µm indicates that part of the total mass was
either not recovered or comprised particles too fine for the cyclosizer to capture effectively.
Calculations :

Cumulative Mass of Fractions Up to a Size ⬚

Cumulative Mass Percentage = ( )
𝑇𝑜𝑡𝑎𝑙 𝑀𝑎𝑠𝑠

Starting Mass : 30 gr

Cumulative mass percentages:

1. For the first fraction (44 µm):

Cumulative Mass = 2.5 g

2.5 g ⬚
Cumulative Mass Percentage = ( 30 g ) x 100 = 8.33 %

2. For the second fraction (33 µm):

Cumulative Mass = 2.5 g + 4.3 g = 6.8 g

6.8 g ⬚
Cumulative Mass Percentage = ( 30 g ) x 100 = 22.67 %

3. For the third fraction (23 µm):

Cumulative Mass = 6.8 g + 3.6 g = 10.4 g

10.4 g ⬚
Cumulative Mass Percentage = ( ) x 100 = 34.67 %
30 g

4. For the fourth fraction (16 µm):

Cumulative Mass = 10.4 g + 2.5 g = 12.9 g

12.9 g ⬚
Cumulative Mass Percentage = ( ) x 100 = 43 %
30 g

5. For the fifth fraction (12 µm):

Cumulative Mass = 12.9 g + 0.5 g = 13.4 g

13.4 g ⬚
Cumulative Mass Percentage = ( ) x 100 = 44.67 %
30 g
Gaudin-Schuhmann Distribution Discussion

Gaudin-Schuhmann’s equation :

𝑃 𝑑 𝑛
=( )
𝑃0 𝑑 0

Where 𝑃 represents the cumulative mass percentage of particles passing through a sieve of size
𝑑, and 𝑃0 is the total mass percentage. The value of 𝑛 describes the distribution of particle sizes,
with a larger negative 𝑛 indicating a higher proportion of finer particles. In this experiment,
deviations from the linear trend in the Gaudin-Schuhmann plot suggest potential limitations in
the cyclosizer’s ability to capture finer particles, particularly below 11 µm.

figure 2. Gaudin-Schuhmann plot for Cyclosizer

The figure 2 above shows a Gaudin-Schuhmann plot for the cyclosizer. The trend on the log-log
scale is curved rather than strictly linear, showing that the particle size distribution is not
perfectly consistent across the size fractions. This suggests that while the material is generally
distributed across a range of particle sizes, there are irregularities, particularly in the finer
fractions.
The slope reflects the distribution of the quartz sample, where most of the mass is concentrated
in the intermediate size fractions (33-44 µm), as shown by the gradual rise in cumulative mass
percentage. However, the decline in slope at the finer size ranges (11-16 µm) indicates that fewer
particles were recovered in these fractions, which could be attributed to the limitations of the
cyclosizer in capturing very fine particles or potential losses during the experimental process.

Part II – Laser Sizer

Laser Sizer Data :

no. Particle Size (μm) % Passing

1 30.8 99.95
2 13.2 94.35
3 5.64 77.25
4 2.41 58.1
5 1.03 38.6
6 0.44 14.05
Table 2.

figure 3. Particle size vs % Passing for Laser Sizer

The laser sizer results, as presented in table 2 and figure 3, tells a significant portion of the
sample (nearly 100%) passing through the 30.8 µm range. The data also indicates that over 77%
of the particles are smaller than 5.64 µm, demonstrating that laser diffraction provides a more
detailed analysis of finer particles compared to the cyclosizer.

The graph shows a steady decrease in the percentage of particles passing as the particle size
decreases. This is consistent with the expectation that finer particles dominate the lower size
ranges, providing valuable insight into the sample's full particle size distribution, particularly
below the 10 µm range where the cyclosizer has limitations.

Comparison :

no. Particle Size (μm) Cumulative Mass % no. Particle Size (μm) % Passing
1 44 8.34 1 30.8 99.95
2 33 22.67 2 13.2 94.35
3 23 34.67 3 5.64 77.25
4 16 43 4 2.41 58.1
5 12 44.67 5 1.03 38.6
6 0.44 14.05

Table 2. Table 3.

Cumulative Mass Distribution

100
90
80
70
60
%

50
40
30
20
10
0
0 5 10 15 20 25 30 35 40 45 50
Particle Size (μm)

Cumulative Mass Percentage (%) % Passing

figure 4. Cumulative Mass Distribution

The cyclosizer effectively captures particles down to 11-16 µm, but it is less capable of
analyzing particles smaller than this, as seen in the lower cumulative mass percentage at larger
particle sizes. The laser sizer captures the finer particles below 10 µm, with the percentage
passing reaching nearly 100% at the coarser sizes.

Advantages and Disadvantages :

The cyclosizer is particularly effective for separating and analyzing coarser particles, generally
in the range of 11 to 44 µm. One of its main advantages is that it provides clear, discrete size
fractions, which makes it straightforward to calculate cumulative mass percentages and analyze
the size distribution for coarser particles. However, its major disadvantage lies in its limited
ability to classify particles smaller than 10 µm, making it less suitable for analyzing fine
particles. Additionally, the cyclosizer is more labor intensive and time consuming process,
requiring careful handling to avoid errors, such as improper drying or inaccurate mass
measurements, just like how we did.

On the other side, the laser sizer offers a more efficient method for analyzing finer particles,
especially those smaller than 10 µm. Its main advantage is the speed and ease with which it
provides a continuous size distribution across a wide range of particle sizes, making it highly
effective for measuring fine particles which the cyclosizer cannot handle. The laser sizer can
analyze many particles simultaneously and does not rely on manual separation processes,
reducing the potential for human error. However, its resolution at higher particle sizes is more
limited, making it less effective at analyzing coarser fractions compared to the cyclosizer. This
limitation means that laser diffraction alone may not provide a complete picture of the particle
size distribution for coarser materials.

Error Analysis :

Throughout the experiment, several factors have been of effect with the results that we had
gotten. One of the big issues was the inconsistency on sample recovery, particularly in the finer
size fractions. We speculated that it was mainly due to human error during handling the sample.
Losses during sample handling, during filtration, or transferring between containers, could have
led to inconsistencies.

Also, in the report we primarily used the data given on the blackboard to overcome these
limitations. We acknowledged that our results have deviated quite significantly from the
expected values, likely due to the sampling errors mentioned above, which is why we have
leaned towards the ideal data from the blackboard. These issues signifies the need for more
careful procedural control and data verification in future experiments.
Conclusion :

In conclusion, both the cyclosizer and laser sizer complement each other nicely. the cyclosizer
effectively classified the quartz sample into distinct size fractions, with most of the mass
concentrated in the intermediate size ranges of 33-44 µm. The Gaudin-Schuhmann distribution
shows a curved trend rather than a strictly linear one, particularly at the finer particle sizes. This
indicates a slower rate of particle recovery in the smaller size fractions, which could suggest
limitations in the cyclosizer’s ability to effectively capture particles smaller than 11 µm. This
curvature highlights the importance of using complementary methods like laser sizer analysis to
gain a more complete understanding of the full particle size distribution.

Laser diffraction provides complementary analysis, particularly for particles smaller than 10 µm.
While this analysis was not performed entirely, the data given highlights how laser diffraction
enhances the resolution of particle size distribution when combined with cyclosizer data.

References

1. MARC Technologies RSS. (n.d.). M16 cyclosizer – classic (superceded). Available at:
https://www.marctech.com.au/laboratory-products-solutions/particle-sizing-systems/the-
cyclosizer/warman-cyclosizer/.

2. Labcompare. (n.d.). The dual light nano particle Sizer’s impact on Quality Control.
Available at: https://www.labcompare.com/Laboratory-Analytical-Instruments/247-
Laser-Particle-Size-Analyzer-Laser-Diffraction/.

3. WASM Curtin University. (2024). Lab Manual. Blackboard.

4. WASM Curtin University. (2024). Warman Cyclosizer Instruction Manual. Blackboard.

5. Albijanic, B. (2024). Structure of a laboratory report. Kalgoorlie: Boris Albijanic.

 Particle Classification by Hydrocyclone

Introduction

Mineral beneficiation is the process of concentrating the valuable minerals in the raw ore through
a series of physical or chemical processing techniques. Particle size classification is a crucial step
in the beneficiation process since the efficiency of subsequent separation processes depends on
particle sizes. Hydrocyclones are widely used for classification of fine particles in mineral
processing plants. Wills (2016) states that hydrocyclones utilize centrifugal forces produced
within a conical structure to segregate particles based on size and density, as well. This approach
is particularly useful for particles less than 1 mm in size, which are not well handled with
traditional screens (Bosman & Muller, 2019). A tangential inlet stream impinges on a typical
hydrocyclone here, with particles in the feed slurry swirling together to be separated based on
mass – larger (heavier) or denser particles gravitate toward the underflow and smaller (lighter) or
finer particles exit via the overflow. This trial is directed at determining the performance of a
hydrocyclone that has their classification efficiencies calculated under various inlet pressures.

The inlet pressure of a cyclone was previously set to 50 kPa, while it is expected that increasing
it to 80 kPa in the current experiment would increase classification efficiency, hence decreasing
d50 value which indicates that more fine particles will be classify into overflow. It was also
hypothesized that the bypass flow corrected results would affinitive categories, and that the
reduced data set after applying bypass flow correction would show a consistent machine
performance across pressures but primarily with movement of all curves in d50 (indicating cut
size is driven by pressure were as efficiency remains similar for each combination). The
experimental objectives were (1) to determine the degree of classification achieved by the
hydrocyclone at 50 kPa and 80 kPa by evaluating the size distributions in both the underflow and
overflow streams, with calculation of d50 under each condition; (2) to ascertain normalized
performance characteristics of the cyclone with data across both pressures so as to provide
insight into inherent efficiency and bypass flow.
Materials and Equipment

Materials required for the experiment include a hydrocyclone test rig (operating at 50 kPa and 80
Kpa), Marcey Gauge to measure pulp densities, and the solid sample that will be fed into the
cyclone for classification. Other ancillary features include a sieve stack for the analysis of the
underflow particle size (500 μm, 355 μm, 250 μm, 180 μm and 63μm), laser sizer to characterise
overflow particles and finer underflow ones; as well as buckets for collection and weighing
masses flows of overflow and underflow streams.

Method

Initially, hydrocyclone was operated at 50 kPa pressure and water circulated through this to
facilitate smooth flow. Pressure control was maintained, the cyclone feed hopper level was
progressively topped up with a clean solid sample. Composite samples of water and solids in
both the underflow and overflow streams were collected by volume over timed intervals, so mass
of slurry could then be determined. A Marcey gauge is a right circular cylinder modelled to
measure the pulp densities of all streams, feed, underflow and overflow when the relative density
of solids in slurry is assumed to be 2.60. It was used to calculate mass flow rates of each stream,
and the total feed mass flow rate as the sum of underflow plus overflow. This process was
repeated for the 80 kPa pressure level.

Experimental Setup

A cyclone feed hopper connected to the hydrocyclone unit, and a controlled inlet pressure (either
50 kPa or 80 kPa) are included in the experimental set up. The underflow and overflow streams
of the cyclone are collected in buckets for more detailed analysis, while pulp density
measurements with a Marcey Gauge are performed. Underflow is analyzed in a sieve stack and
laser sizer, whilst the overflow is analyzed directly on the laser sizer.

Mathematical Models
The performance of the hydrocyclone was evaluated using several key mathematical
models. First, the mass flow rate of the underflow and overflow streams was calculated
using the formula: Mass Flow Rate (tph)= Mass of Solids (kg) / Time (hours), with the total
feed mass flow rate being the sum of the two streams. The d50 value, representing the
particle size at which 50% of particles report to the underflow and 50% to the overflow,
was determined from the performance curve. To correct for bypass flow, the corrected
U/F % Passing was calculated using the formula: Corrected U/F % Passing = (U/F %
Passing – Fbypass) / (1 – Fbypass), accounting for inefficiencies in the classification
process. Finally, to compare the cyclone's performance across pressures, the normalized
size was calculated by dividing each particle size by the corresponding d50: Normalized
Size = Particle Size / d50, enabling the analysis of reduced performance curves. Detailed
calculations are performed in the Appendix.
Results and Discussion
1. Performance Curves for 50 kPa and 80 kPa
The performance curves shown in Figure 1 provide the relationship between particle size
and underflow recovery for both 50 kPa and 80 kPa inlet pressures. The d50 value (here
testing under these conditions as the particle size at which 50% report to the overflow and
50% report to the underflow) is an indication of classification efficiency. When the
hydrocyclone was operated at 50 kPa the d50 value was found to be 10.7 µm and for a
pressure of 80 kPa, d50 = 9.3 µm indicating that at higher pressures the cyclone will
classify finer particles into the overflow. One can expect a lower d50 value with the 80 kPa
classification pressure due to the higher centrifugal forces which will help separation of
fine particles from the slurry.

Figure 1: Performance Curve for 50kpa and 80kpa

2. Influence of Pressure on Underflow and Overflow Characteristics

These results are complimented by the slurry specific gravity (SG) and % solids data from table
1 in the appendix which reinforces those observations. At 50 kPa the underflow has SG of 1.403
with 46.8% solids, whereas the overflow is SG of 1.023 with 3.5% solids The SG and solids
content of the underflow at 80 kPa also drops markedly to 1.095 and 14.25% respectively
(compared to the overflow's remaining relatively unchanged values of an SG of 1.020 and solids
content of only ~3.5%). Therefore, the underflow SG and solids content are dropped at higher
pressure, which means that to some extent the cyclone overclassifies fine particles in the
underflow into the overflow.

The timed sample weights provided in Table 2 in the appendix indicated that the mass of slurry
processed in 10 s is greater at 80 kPa for both underflow (2647.1 g) and overflow (4686.1 g), as
compared with 50 kPa (2326.3 g for underflow; 4230.9 g for overflow). The higher mass flow
rate at 80 kPa counterbalances partially this improvement in classification, which would indicate
that high pressure not only improves cyclone classifying but also enhances its throughout. In
table 3 in the appendix, Dry Solids Mass also offers some additional information about which
solids enrichment solid of stands befitting the underflow, and which of them attend to the
overflow. On the other hand, at 80 kPa the underflow contains only 343.4 g of solids and
overflow tablet contains 126.3 g of cumulative solids; indicating further that at higher pressure
more fines are reporting to the overflow product than in the former case, and showing that
thereby cyclone was able separate finer particles from slurry much better at higher pressure
conditions.

3. Corrected Performance Curves for 50 kPa and 80 kPa

Corrected performance curves with bypass flow enable more accurate classification efficiency
evaluations of a hydrocyclone. The corrected U/F % Passing after applying the bypass correction
decreased to a small extent for most particle size fractions with more decrease observed at finer
sizes, representative of the effect of bypass flow. Nevertheless, there are no changes for the d50
values of 10.7 µm (50 kPa) and 9.3 µm (80 kPa), demonstrating that this correction does not
greatly affect the global efficiency of classification but shows slight misclassification in fine
particle separation. As in the performance curves, the corrected curves all have a higher gain for
80 kPa operation and another clear trend is now available, which is confirmation of our
conclusion: The best classification results are obtained with high pressure.
 Figure 2: Corrected performance Curve for 50kpa and 80kpa

4. Reduced Performance Curves

The reduced performance curve in Figure 3, shows the cyclone’s classification efficiency after
normalizing particle sizes by d50. The convergence of the curves for 50 kPa and 80 kPa at
smaller normalized sizes indicates that the cyclone operates similarly across both pressures when
particle sizes are normalized. This suggests that pressure primarily shifts the d50, while the
overall classification efficiency remains consistent. The sieve analysis data from table 4 in the
appendix supports this conclusion. At 50 kPa, the majority of solids in the underflow are in the
+125-250 µm range (81.1 g), while at 80 kPa, fewer solids are present in the underflow, with
only 54.6 g in the +125-250 µm range. This reduction in mass at higher pressure confirms that
finer particles are being classified into the overflow, leading to a steeper performance curve at 80
kPa.
The laser sizer analysis results for both the underflow and overflow, as presented in Table 5 in
the appendix, further illustrate the differences in particle size distribution between the two
pressures. For 50 kPa, the overflow shows a 50.3% U/F passing at 10.7 µm, while at 80 kPa,
50.3% U/F passing is observed at 9.3 µm, indicating that the higher-pressure results in finer
particles reporting to the overflow.
 Figure 3: Reduced performance Curve for 50kpa and 80kpa

Conclusions

The primary purpose of this study was to analyse the proper effectiveness for classification of a
hydrocyclone. This process would be undertaken at two different entrance pressures: 50 and 80
kPa. The particle size distribution was examined in both underflow and overflow of the
hydrocycle. The performance curves revealed that increasing the inlet pressure from 50 kPa to 80
kPa improved the cyclone’s classification efficiency, with the d50 decreasing from 10.7 µm to
9.3 µm, indicating finer particle separation at higher pressure. The corrected performance curve,
after accounting for bypass flow, confirmed these findings, demonstrating minimal impact from
bypass on the overall classification efficiency. The reduced performance curve, which
normalized particle size by d50, showed consistent classification behaviour across pressures,
with the key difference being the shift in d50 due to pressure variation. These results align with
existing literature, validating the hypothesis that higher pressure enhances classification
efficiency. Future research should investigate the effects of varying slurry properties and cyclone
geometries to further optimize classification across different operational conditions.

References

Wills, B. A. (2016). Wills' mineral processing technology: An introduction to the practical

aspects of ore treatment and mineral recovery (8th ed.). Butterworth-Heinemann.
Bosman, J., & Muller, D. (2019). Fine wet screening with the sieve bend. Mineral Processing
Journal, 36(2), 123-135.

Appendices

Table 1: Slurry SG of Different Streams

50 kPa 80 kPa
U/F O/F U/F O/F
SG 1.403 1.023 1.095 1.020
% Solids 46.8% 3.5% 14.25% 3.5%

Table 2: Timed (10 s) sample weight of the underflow and the overflow
50 kPa U/F slurry (g) 2326.3

O/F slurry (g) 4230.9

80 kPA U/F slurry (g) 2647.1

O/F slurry (g) 4686.1

Table 3: Mass of the Dry Mass

50 kPa U/F solids (g) 817.0

O/F solids (g) 100.3

80 kPa U/F Mass (g) 343.4

O/F Mass (g) 126.3

Table 4: Cyclone Underflow Sieve Analysis

Seive Size (μm) 50 kPa 80 kPa
+500 0.7 g 0.4 g
+355-500 2.1 g 0.6 g
+250-355 5.5 g 3.3 g
+125-250 81.1 g 54.6 g
+63-125 74.5 g 54.2 g
-63 34.2 g 27.9 g
Table 5: Laser sizer analysis of overflow and underflow at 50 kPa and 80kPa and calculations based on
the formula given.
O/F % U/F % Corrected nomalized Particle O/F % U/F % Corrected nomalized
Particle Passing Passing U/F % size 50 Size (µm) Passing Passing U/F % size 80
Size Sample Sample Passing Sample 1 Sample 1 Passing
(µm) 1 1 Sample 1 Sample 1

500 100 100 100 46.72897 500 100 100 100 53.76344
355 100 100 100 33.17757 355 100 100 100 38.17204
250 100 100 100 23.36449 250 100 100 100 26.88172
106 100 100 100 9.906542 106 100 100 100 11.39785
63 100 100 100 5.88785 63 100 100 100 6.774194
52.3 100 99.9 99.89474 4.88785 42.3 100 99.9 99.89474 4.548387
47.1 100 99.8 99.78947 4.401869 38.1 100 99.8 99.78947 4.096774
42.3 100 99.6 99.57895 3.953271 34.2 100 99.4 99.36842 3.677419
38.1 100 99.2 99.15789 3.560748 30.8 100 98.9 98.84211 3.311828
34.2 100 98.5 98.42105 3.196262 27.7 100 97.8 97.68421 2.978495
30.8 100 97.4 97.26316 2.878505 22.4 99.9 93.4 93.05263 2.408602
27.7 100 93.6 93.26316 2.588785 18.1 99.5 85 84.21053 1.946237
24.9 99.9 93 92.63158 2.327103 16.3 99.1 79.6 78.52632 1.752688
22.4 99.8 89.8 89.26316 2.093458 14.6 98.5 73.7 72.31579 1.569892
20.1 99.7 84.7 83.89474 1.878505 13.2 97.7 67.8 66.10526 1.419355
18.1 99.4 79.2 78.10526 1.691589 11.8 96.5 62.1 60.10526 1.268817
16.3 99 73.3 71.89474 1.523364 10.7 94.9 57 54.73684 1.150538
14.6 98.3 67.1 65.36842 1.364486 9.58 93 52.3 49.78947 1.030108
13.1 97.3 61.1 59.05263 1.224299 8.62 90.6 48.2 45.47368 0.926882
11.8 96.1 55.4 53.05263 1.102804 7.75 87.8 44.5 41.57895 0.833333
10.7 94.4 50.3 47.68421 1 6.97 84.6 41.9 38.84211 0.749462
9.58 92.4 45.7 42.84211 0.895327 6.27 81 38.2 34.94737 0.674194
8.62 89.9 41.7 38.63158 0.805607 5.64 77.1 35.4 32 0.606452
7.75 87 38.2 34.94737 0.724299 5.07 72.9 32.9 29.36842 0.545161
6.97 83.8 34.9 31.47368 0.651402 4.56 68.5 30.4 26.73684 0.490323
6.27 80.2 32 28.42105 0.585981 4.1 64.1 28.2 24.42105 0.44086
5.64 76.4 29.4 25.68421 0.527103 3.69 59.8 26.1 22.21053 0.396774
5.07 72.1 26.9 23.05263 0.473832 3.32 55.6 24.2 20.21053 0.356989
4.56 67.8 24.6 20.63158 0.426168 2.98 51.7 22.3 18.21053 0.32043
4.1 63.5 22.4 18.31579 0.383178 2.68 47.8 20.5 16.31579 0.288172
3.69 59.2 20.4 16.21053 0.34486 2.41 44.1 18.7 14.42105 0.25914
3.32 55.1 18.8 14.52632 0.31028 2.17 40.8 16.9 12.52632 0.233333
2.98 51.1 16.8 12.42105 0.278505 1.98 37.9 15.2 10.73684 0.212903
2.68 47.3 15.6 11.15789 0.250467 1.75 35.2 13.6 9.052632 0.188172
2.41 43.6 13.5 8.947368 0.225234 1.58 32.7 12.1 7.473684 0.169892
2.17 40.3 12 7.368421 0.202804 1.42 30.4 10.7 6 0.152688
1.98 37.4 10.6 5.894737 0.185047 1.28 28.1 9.4 4.631579 0.137634
1.75 34.8 9.3 4.526316 0.163551 1.15 25.8 8.3 3.473684 0.123656
1.58 32.3 8 3.157895 0.147664 1.03 23.6 7.3 2.421053 0.110753
1.42 30 6.9 2 0.13271 0.93 21.3 6.4 1.473684 0.1
1.28 27.8 6.3 1.368421 0.119626 0.83 19 5.5 0.526316 0.089247
1.15 25.6 5.1 0.105263 0.107477 0.75 16.8 4.7 -0.31579 0.080645
1.03 23.4 4.4 -0.63158 0.096262 0.68 14.6 4 -1.05263 0.073118
0.93 21.1 3.7 -1.36842 0.086916 0.61 12.5 3.3 -1.78947 0.065591
0.83 18.9 3.6 -1.47368 0.07757 0.55 10.4 2.8 -2.31579 0.05914
0.75 16.7 2.7 -2.42105 0.070093 0.49 8.8 2.2 -2.94737 0.052688
0.68 14.5 2.2 -2.94737 0.063551 0.44 6.9 1.8 -3.36842 0.047312
0.61 12.4 1.8 -3.36842 0.057009 0.4 5.4 1.4 -3.78947 0.043011
0.55 10.4 1.5 -3.68421 0.051402 0.36 4 1 -4.21053 0.03871
0.49 8.6 1.6 -3.57895 0.045794 0.32 2.8 0.7 -4.52632 0.034409
0.44 6.9 0.9 -4.31579 0.041121 0.17 0 0 -5.26316 0.01828
0.4 5.4 0.7 -4.52632 0.037383 0.15 0 0 -5.26316 0.016129
0.36 4.1 0.5 -4.73684 0.033645 0.14 0 0 -5.26316 0.015054
0.32 2.8 0.4 -4.84211 0.029907 0.12 0 0 -5.26316 0.012903
0.29 1.8 0.2 -5.05263 0.027103 0.11 0 0 -5.26316 0.011828
0.26 1 0.1 -5.15789 0.024299 0.1 0 0 -5.26316 0.010753
0.23 0.5 0.1 -5.15789 0.021495
0.21 0.1 0 -5.26316 0.019626
0.19 0 0 -5.26316 0.017757
0.17 0 0 -5.26316 0.015888
0.15 0 0 -5.26316 0.014019
0.14 0 0 -5.26316 0.013084
0.12 0 0 -5.26316 0.011215
0.11 0 0 -5.26316 0.01028
0.1 0 0 -5.26316 0.009346
 Experiment 4 - Flotation Kinetics

Introduction
Flotation utilizes bubbles to target minerals and separate them from the gangue. Flotation is
common to separating ores that have a low average particle size for liberation using gravity
separation or when the gravity difference between minerals is too close. In this experiment we
tested the effects of particle size on floatation (Gupta and Yan 2006).
We can hypothesize that the smaller would have a higher recoverability due to them being finer
and the higher likelihood of contacting the collector reagent. We would also see the effects of the
stirring duration on the sample. We can assume that the longer it takes to stir the more kinetic
energy the sample will experience, with that sample that is in the tank for a longer time may results
in increase of recovery, since the particle distribution would increase. Although, it can also be
argued that the longer the stirring duration the higher the chance the froth layer may become
unstable.

Materials and Methods

Equipment:
- 500 g of -250 µm quartz
- Wemco float cell
- Screens
- Trays
- Bucket
- Stopwatch
- pH meter
- Micro pipettes and syringes
Reagents:
- 0.01 M sodium oleate
- 0. 1 M and 1 M NaOH solutions
- 0. 1 M and 1 M HCl solutions
- 0.01 M and 1 M calcium chloride solution
- MIBC frother

Methods
1. The experiment is first done by sampling 500 gr of 250 µm quartz and divided into
portions of standard sieves
 2. Next, after preparing the float cell, the remaining portion of the quartz was activated
by conditioning it in a 10-3 mol/L calcium chloride solution for two minutes while
stirring continuously.
3. The pH adjustments were done afterwards by adding NaOH solution to increase the pH
to 12. Next, 500 gr/ton sodium oleate and 3 drops of MIBC frother are added. The pulp
was conditioned for 5 minutes with the air inlet valve closed.
4. After opening the air inlet samples are collected for drying. Samples are gathered at the
15-, 30-, 60-, 120-, 180-, and 300-seconds mark.

Results and Discussions

Results Data
The sieve size distribution is given in Table 2 in the appendix, and the visual graph is shown in
Figure 1. The data of Froth product weight after drying is in Table 1.

Figure 1: Sieve size vs weight graph

Table 1: Data of Froth product weight after drying

Dry Froth Product Weights
Time (s) Weight (g)
30 67.7
60 111
120 148.2
180 60.3
 300 41.2

Calculations
C0 = 450 gr/2.5 l
= 180 gr/l

Recoveries
R30 = 67.7 gr / 500 gr * 100% = 13.54%
R60 = (67.7 + 111) / 500 * 100% = 178.7/500 * 100% = 35.74%
R120 = (178.7 + 148.2) / 500 *100% = 326.9/500 * 100% = 65.38%
R180 = (326.9 + 60.3) / 500 * 100% = 387.2/500 * 100% = 77.44%
R300 = (387.2 + 41.2) / 500 * 100% = 428.4/500 * 100% = 85.68%

- This degree of recovery concerns us since the higher the recovery the lower the grade.

Figure 2: Recovery percentage vs time graph of the dry froth product

• Klimpel Model

R = R∞ (1 – e -kt)
We can assume that R∞ is around 99% of recovery. Re-arranging gives:

 -kt = Ln (1 – R/ R∞)

For t30:
Ln (1 – 0.1354/ 0.99) = Ln (1 – 0.136)
 Ln (0.864) = −0.147

For t60:
Ln (1 – 0.3574/ 0.99) = Ln (1 – 0.361)
 Ln (0.639) = −0.417

For t120:
Ln (1 – 0. 6538 / 0.99) = Ln (1 – 0.66)
 Ln (0.34) = −1.08

For t180:
Ln (1 – 0.7744/ 0.99) = Ln (1 – 0.781)
 Ln (0.219) = −1.52

For t300:
Ln (1 – 0.8568/ 0.99) = Ln (1 – 0.86)
 Ln (0.14) = -1.97

The Klimpel model can be re-arranged to find the negative rate constant, –k:
-kt = Ln (1 – R/ R∞)
 -k = Ln (1 – R/ R∞) / t
The Klimpel model, Ln (1 – R/ R∞) Vs. T, was plotted in Figure 3, and the value of –k was
determined from the slope. The slope of the graph was -0.00671, therefore the rate constant k =
0.00671

Figure 3: Ln (1 – R/ R∞) vs time

• Kelsall Model (two component model)

R30: Ln (1 - 0.1354) = -0.147

R60: Ln (1 - 0.3574) = -0.443
R120: Ln (1 - 0.6538) = -1.056
R180: Ln (1 - 0.7744) = -1.487
R300: Ln (1 - 0.8568) = -1.951
 Figure 4: Ln(1-R) in relation to time

- The graph shows us that the mineral has a combined (fast and slow) floating component.

Assuming a first order relationship and gas volume of 15% in the cell, calculate the number of 10
m3 cells required to recover 80% of the same feed at the rate 50 t/h. Consider the feed pulp density
and the reagent combinations being the same as in the batch test and the specific gravity of quartz
being 2.65 g/cm3

λ = 0.15
Ra = 0.80
N = number of cells required

Ra = 100[1−(1+kλ)-N]
0.8 = 100[1−(1+0.00671 * 0.15)-N]
0.8 = 100[1−(1+0.001)-N]
0.8 / 100 = [1 – (1.001)-N]
0.008 – 1 = (- 1.001)-N
- 0.992 = (-1.001)-N
0.992 = (1.001)-N
ln (0.992) = -N * ln (1.001)
N = ln (0.992) / - ln (1.001)
N = - 0.00803 / -0.00099
N = 8.11

Model Comparison
Both Klimpel and Kelsall models yielded similar results, we can see that the rates start relatively
fast before the rate reduced, which is visible at the 120 and 180 second mark. Klimpel and Kelsall’s
model are similar in calculation where they are Ln (1 – R/ R∞) and Ln(1-R) respectively. Given
that R∞ is near to the value of 1 we can see why both calculations produce identical graphs.

Conclusion
Overall flotation kinetics shows first-order kinetics, where the concentration of the mineral in the
froth determines the recovery rate. The flotation rate constant also has a large impact on the
recovery. Based on calculations if a higher flotation rate constant were to appear, the recovery
speed would be faster than a lower rate constant. The size of the particles would also affect the
amount of material recovered. It is absolute that smaller and finer particles are harder to recover
from the froth, but on the other hand particles that are too course will also be hard to float, Hence
it will reduce the flotation’s efficiency.

Reference:
- Gupta, A., & Yan, D. S. (2006). Mineral processing design and operation: An introduction.
Elsevier Science.
- Han, J., Chen, P., Liu, T., & Li, Y. (2023). Research and application of fluidized flotation units: A
review. Journal of Industrial and Engineering Chemistry, Vol 126.
- WASM Curtin University. 2024. Lab Manual. Blackboard.

Appendix

Table 2: Sieve data

Sieve Analysis
Sieve Size (Microns) Weight (gr)
180 9.7
125 12.1
90 4.6
75 9.1
53 19.5
 Flocculation
Introduction:

Flocculants play a crucial role in solid-liquid separation processes, aiding in the aggregation of
suspended particles to form larger flocs. These polymers can be classified based on their ionic
character: non-ionic, cationic, anionic, and amphoteric. In this report, we focus on non-ionic and
cationic flocculants and explore their respective advantages, focusing on why the results show
non-ionic performed slightly better (Dao and Kei 2016).

Non-ionic flocculants can be better in some cases compared to cationic flocculants for several
reasons:

pH Sensitivity: Non-ionic flocculants are less affected by pH changes compared to cationic

flocculants. Cationic flocculants can be significantly influenced by pH and ionic strength
because their charge is derived from either protonation or quaternary nitrogen groups. If the pH
or ionic strength deviates from the optimal range, cationic flocculants may lose their
effectiveness. Non-ionic flocculants, on the other hand, do not carry a charge and are generally
more stable across a range of pH values.

Ionic Strength: Non-ionic flocculants are less impacted by ionic strength variations in the
solution. Cationic flocculants can lose their effectiveness in high ionic strength environments
due to charge screening effects. Non-ionic flocculants are less affected by such ionic strength
changes.

Application Versatility: Non-ionic flocculants are often used in applications where the specific
charge characteristics of cationic or anionic flocculants are not ideal. For instance, non-ionic
flocculants are typically employed when the flocculation process requires the polymer to
function primarily through bridging rather than charge interactions.

Less Risk of Adverse Reactions: In certain processes, especially those involving sensitive
materials or where unintended reactions can occur, non-ionic flocculants can be preferable
because they do not introduce additional charge interactions that might lead to undesirable side
reactions or interfere with the process.

Bridging Mechanism: Non-ionic flocculants work primarily through the bridging mechanism,
where polymer chains extend and form bridges between particles. This can be advantageous in
certain scenarios where charge-based mechanisms are less effective or where the flocculant
needs to operate effectively regardless of the charge characteristics of the particles being treated.
In summary, non-ionic flocculants offer stability in varied conditions and are less influenced by
pH and ionic strength, making them suitable for applications where these factors might affect the
performance of cationic or anionic flocculants (Tarletton and Wakeman 2007).

Materials and Methods:

Setup: Figures 1 and 2, are the scheme of our setup, with Figure 1 being the setup for Part I and
II of the experiments and Figure 2 being the setup for Part III.

Figure 1: being the setup for Part I and II of the experiments

Figure 2: setup for Part III

Equipment:
− Four 250 ml. Measuring Cylinders.
− Ten 100 ml. Measuring Cylinders
− Stopwatch
− Anionic, cationic and non-ionic flocculants. 1 N HCl solution
− 1M. NaOH solution
− 300 g of tailings test sample.
− Paper strips, sticky tape, and marker pen Clarity gauge

Three methods are being explored in this experiment which are as follows:
1. Coe and Clevenger's method determines thickener area by balancing particle settling rates
against the upward flow of overflow water. The key equation A = t(F−D)/ρRA calculates
the required area A, where t is the solids feed rate, F and D are the liquid-to-solid mass
ratios for feed and underflow, and ρ is the fluid density. The equation simplifies to A* =
(F−D)/R for unit area per ton of dry solids per hour, with a common practice of adding a
25% safety factor.

2. Kynch's method, which requires only a single batch test, focuses on the critical point of a
settling curve to determine the settling rate R. The critical point is identified through
methods like using a condensed time scale or log-log paper. From there, the unit area A*
is calculated using A* = tu/Co*Ho, where tu is the time at the desired underflow
intercept, Co is the feed concentration, and Ho is the initial pulp height.

3. Oltmann's empirical method combines elements of both Coe and Clevenger's and
Kynch's methods. It starts by locating the critical point as in Kynch's approach, then
determines the gradient of the settling curve from Ho to the critical point to find R′. This
adjusted settling rate R′ is used in the Coe and Clevenger equation to estimate the
thickener area.

Results:

Table 1: Part 1 cationic and non-ionic experimental data

Flocculant ml pH Time (sec) Height Height from Settling Rate
(mm) (60 - top (mm) (mm/s)
80) Mudline
Cationic 0.5 9.34 5 32 79 4.00
Non-ionic 0.5 9.34 4.62 31 88 4.33
 Cationic 1 9.34 3.22 33 88 6.21
Non-ionic 1 9.34 2.75 32.5 88.5 7.27
Cationic 1.5 9.34 4.73 32 76 4.23
Non-ionic 1.5 9.34 4.12 32.5 92.5 4.85
Cationic 5 9.34 6.27 37 92 3.19
Non-ionic 5 9.34 3.66 37 104 5.46

Table 2: Part 2 non-ionic and pH experimental data

Flocculant ml pH Time (sec) Height Height from Settling
(mm) (60 - top (mm) Rate
80) Mudline (mm/s)
Non-ionic 1 9.34 2.75 32.5 88.5 7.27
Non-Ionic 1 6.48 5.92 40 89 3.38
Non-Ionic 1 3.3 4.87 33 90 4.11

Table 3: Part 3 non-ionic flocculant and 20% solids experimental data

Settling
Height Rate
Flocculant ml pH Time (sec) (mm) (mm/s)
0 0
Non- ionic 2.5 9.1 10 64 20% solids 6.40
20 89.5 2.55
30 99 0.95
40 104 0.50
50 107.5 0.35
60 111 0.35
5min 300 126 0.06

Table 4: Part 3 non-ionic flocculant and 30% solids experimental data

Settling
Height Rate
Flocculant ml pH Time (s) (mm) (mm/s)
Non- ionic 2.5 8.9 10 15 30% solids 0.38
20 28 1.30
30 40 1.20
40 52 1.20
50 60 0.80
60 66 0.60
70 72 0.60
80 78 0.60
90 81 0.30
 100 83 0.20
110 86 0.30
120 88 0.20
5min 300 116 0.16

Table 5: Part 3 non-ionic flocculant and 40% solids experimental data

Settling
Height Rate
Flocculant ml pH Time (s) (mm) (mm/s)
Non- ionic 2.5 8.98 10 3 40% solids 0.39
20 9 0.60
30 11 0.20
40 13 0.20
50 23 1.00
60 32 0.90
70 39 0.70
80 46 0.70
90 49 0.30
100 53 0.40
110 55 0.20
120 57 0.20
5min 300 72 0.08
Calculations:

Part I
 Figure 3: The settling rate vs flocculant addition graph for Part I

Comparing the data in Figure 3, the non – ionic had faster settling rate with every flocculation
addition.
Part II

Figure 4: The settling rate vs flocculant addition graph for Part II

The settling rates in Figure 4, show that as the pH increased, the settling rate also increased.

Part III

Figure 5: The settling rate vs flocculant addition graph for Part III

As observed from Figure 5, the higher the percent of solids, the more irregular the settling rate
becomes with time. These experiments are all done under similar pH and the same amount of
flocculant added (2.5 mL).
 Figure 6: The height of mudline vs time graph and calculated values

This graph is from the results of 20% solids as per the lab manual, to find tu, and ultimately
calculate A*. D is 50%, and thus the mudline is at 0.125 m (250 mm / 2 converted to m).

1. Coe and Clevenger

Where:
A = the area of thickener (m2) = ?
t = the solids feed rate (t/hr) = 1000 t/d
F = liquid : solid mass ratio of feed = 20 %
D = liquid : solid mass ratio of desired underflow = 50 %
ρ = Relative density of fluid (t/m3) = 1 t/m3
R = Particle (or floc) settling rate (m/hr) = 23.04 m/hr
 2. Kynch

Where:
A* = unit area, the area required per ton of dry solids per hour
tu = time at intercept, e.g. hours
Co = Feed concentration (mass of solids (tonnes) / m3 pulp
Ho = Initial pulp height (metres)

Assume density of solid is 2500 kg/m3 or 2.5 t/m3 and density of liquid is water which is 1 t/m3

Volume of pulp is 250 mL = 0.00025 m3

 3. Kynch and Coe and Clevenger Combination
R’ = slope from Ho,to which is (0.25,0) to the critical point which is (0.1605,0.005556)

Comparison:
Coe and Clevenger produced a smaller A* value (0.16 m²/(t.hr) with a safety factor) compared to
Kynch (0.18 m²/(t.hr)). While these values are quite close, Kynch's result indicates a slightly
larger thickener area required for the same process.

Kynch Formula considers the settling curve and the time intercept in Figure 6, while Coe and
Clevenger use the settling rate directly. This difference in approach can result in slight variations
in the calculated thickener area, depending on the accuracy of the intercept and settling rate
measurements.

Both methods are valid, but the slight discrepancy suggests that the Coe and Clevenger method
may underestimate the required area slightly if the settling rate (R) is not accurately measured.
Kynch’s formula, which factors in the actual settling curve, might provide a more realistic
estimation.

The Kynch formula and combination method yield higher areas, with the combination method
providing the largest estimate (7.75 m²). The combination method, which integrates both
approaches, suggests that a slightly larger area than the Kynch estimate might be necessary to
account for a more accurate balance between settling rates and underflow requirements.

Conclusion:

Both Coe and Clevenger and Kynch formulae provide reasonable estimates for thickener area
calculations. In practice, Coe and Clevenger might be more suitable for quick calculations,
especially when the settling rate is well understood, while Kynch provides a more precise
approach when detailed settling data is available. Given the relatively small difference in the
results, either method could be valid depending on the level of precision required for the
operation. In saying that, the combined Kynch and Coe and Clevenger method provides a more
accurate thickener area calculation than either method alone. It incorporates both the early
settling behaviour (Kynch) and the simpler flow rate and density considerations (Coe and
Clevenger), leading to a slightly larger recommended area for ensuring efficient thickening.

References

Tarleton, E.S., Wakeman, R.J. 2007. “Pretreatment of Suspensions” Solid/Liquid Separation

Equipment Selection and Process Design 126-151.
https://www.sciencedirect.com/topics/engineering/flocculant.

Dao Vu H., Cameron Neil R., and Saito Kei. 2016. “Synthesis, properties and performance of
organic polymers employed in flocculation applications” Polymer Chemistry 7 (11).
https://pubs.rsc.org/en/content/articlepdf/2016/py/c5py01572c.

WASM Curtin University. 2024. Lab Manual. Blackboard.