181
MODEL L I NG AND SI MUL ATI ON OF THE GRI NDI NG CI RCUI T AT ‘ EL
PI LON’ MI NE
Juan Luis ReyesBahena
Abstract
The grinding operation in a ball mill is widely used in comminution. This process is
extremely capital and energy intensive process and at the same time it is notoriously
inefficient. Therefore, simulation can be used as a convenient tool in studying the
behaviour and performance of grinding circuits by facilitating the evaluation of various
scenarios and identification of optimum operating conditions. This paper presents how
these modern simulation techniques are applicable to a grinding circuit using JKSimMet
computer package.
A grinding model developed at ‘El Pilon’ mine using JKSimMet (Version 4) was used to
simulate three different milling designs. The objective of these simulations was to
investigate a new design of the grinding circuit with an additional ball mill to increase
both grinding throughput and fineness. The simulation results highlight that two closed
grinding circuits in series are more appropriate to achieve these targets. Surveying the
implemented grinding circuit validated the JKSimMet simulation results. An increase of
grinding throughput from 600 tpd to 800 tpd was achieved in the grinding circuit in serie
as well as an increase of the grinding fineness from 57% to 75% passing 75 µm.
This case study shows that JKSimMet can be used to design grinding circuit accuracy
and highlights that this computer simulator is a powerful tool in the optimisation of
comminution circuits.
Key words: Grinding circuit; milling in serie; modelling; optimisation; simulation
INTRODUCTION
The grinding operation in a ball mill is widely used in comminution. However, this process is extremely
capital and energy intensive process and at the same time it is notoriously inefficient. Therefore, accurate
designing of both circuit and operating parameters are important to optimise the power consumption,
which can lead to substantial economic benefit. Simulation can be used as a convenient tool in studying
the behaviour and performance of grinding circuits by facilitating the evaluation of various scenarios and
identification of optimum operating conditions. Currently, the development of simulators for mineral
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
182
processing units has considerably increased in last decade; e.g. typical software packages are USIM PAC
(Broussand et al, 1988), JKSimMet (JKTech, 1989), MODSIM (Herbst et al, 1989), and SIMBAL
(CANMET, 1989).
JKSimMet (Version 4) computer package is used in this study to model and simulate the ‘El Pilon’
grinding and classification stages. This simulator was used to assist in the design of a new grinding circuit
with the addition of a new ball mill in order to increase both grinding throughput and fineness. Thus,
simulation results are implemented in the plant which were validated with a new sampling campaign.
‘EL PILON’ MINE
‘El Pilon’ mine mill is located in the town of San Martin de Bolaños, approximately 500 km east of the
Guadalajara State in Northern Mexico. The concentrator processes a sulphide orebody consisting of 1.64%
Pb, 0.10% Cu, 5.31% Fe and 2.64% Zn. Silver content ranges from 270 to 320 g/ton and gold from 0.20 to
0.34 g/ton. The ore is leached with cyanide to obtain the doré (i.e. AgAu alloy). Current milling rates are
approximately 600 ton/day (tpd).
Ore is delivered to the mill by trucks and sent directly to the primary crusher. The ore is crushed in a
primary jaw crusher and transported by conveyor to the secondary and tertiary crushing circuit. The
crushing circuit product contains material approximately 96% passing 9.53 mm (3/8 inches). This material
is conveyed via belt feeders to the 2.59m by 3.67m (8 ft ½ inches in diameter by 12.0 ft in length) 300
HP Marcy ball mill. The ball mill, with 40% of mill filling and 70% of its critical speed, is fitted with steel
liners. It uses steel balls in a ratio of 70% of 76.2 mm (3 inches) and 30% of 50.8 mm (2 inches) as
grinding media. The pulp density in the ball mill is maintained at 76% solids, and lime and cyanide are
added to the mill feed. Lime addition increases the circuit pH to reduce volatilization of cyanide and to
increase the settling of material in the thickener downstream. Cyanide is used to leach silver/gold bearing
particles throughout the circuit.
The grinding circuit uses a twostage cyclone process to classify mill discharge. The classification circuit
consists of two 0.508m (20 inches) diameter cyclone. Only one cyclone is operating at a time while the
other is on standby. Discharge from the ball mill flows to a pump box where it is diluted to a density
approximately 48% solids before being sent to the cyclone, for classification, by means of a 0.20 m by
0.15 m (DRL 8 inches by 6 inches) pump at 870 rpm. Underflow from the cyclone is fed back to the ball
mill while the overflow material (19% to 22% solid) with 58% passing 75 µm (200 mesh) is directed to a
19.81 m (65 foot) diameter thickener located outside the east wall of the mill complex. Underflow of the
thickener, having a pulp density from 50 to 60% solids, is sent to the leaching process. The leached pulp
and its silver/gold bearing solution are treated using the Merrill Crowe process to obtain the doré.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
183
GRINDING CIRCUIT SIMULATOR
JKSimMet is a powerful tool for analysis and simulation of mineral processing plant data. It gives
engineers the ability to design and optimise any crushing or grinding circuit including stages of
classification. It allows engineers to:
• Design a circuit on the graphics monitor,
• Enter model and plant data,
• Simulate the circuit, and
• Graph and print the results.
JKSimMet performs steady state simulation of a range of comminution and classification operations. The
principal application of JKSimMet is to carry out process analysis and optimisation of existing circuits and
also extremely useful for conducting conceptual design studies, where the purpose is to assess the
suitability of different flowsheets to achieve a desired performance objective, as presented in this paper.
Mathematical Models
Process models of the two main units at ‘El Pilon’ grinding circuit are presented:
• Ball mill, and
• Hydrocyclone
Ball Mill
The perfect mixed model (Whiten, 1976) is quite similar to the general population balance model. This
model assumes a perfectly mixed mill which removes most of the complexities in the general population
balance model. The perfect mixed model start with the same population balance, i.e.
out Breakage out oduct Pr in Breakage in Feed + = + (1)
However, two factors control the breakage of particles inside the mill as follows:
• Selection function which defines the breakage probability of particles,
i i i
s d p = (2)
where, d
i
, is the discharge function in the mill, thus
out) (breakage s r selection
i i
= (3)
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
184
Breakage function which define how the selected particles are broken (appearance function) in the mill
content:
in) (breakage s r a appearance
j j ij
= (4)
Then, the balance equation around each size fraction, which describes the mill behaviour, is:
i i i
i
1 j
j j ij i
s r p s r a f + = +
¿
=
(5)
Now substitute for the experimentally mill contents:
i
i
i i
i
1 j j
j j ij
i
d
p
r p
d
p r a
f + =
(
(
¸
(
¸
+
¿
=
(6)
This means that the ratio
i
i
d
r
can be calculated for each size fraction from a set of actual feed (f
i
) and
product (p
i
) measurements, subject to a reasonable form of the appearance function (a
ij
).
Hydrocyclone
The hydrocyclone model used in this study was that developed by Nageswararao under the supervision of
A. J. Lynch and L. Elber (Nageswararao, 1978). This model comprises empirical equations for the main
performance criteria, incorporating terms suggested by dimensional reasoning with exponents estimated
from data sets obtained by Nageswararao (1978) and Rao (Lynch and Rao, 1975). The predictive
equations are:
15 . 0
20 . 0
C
C
50 . 0
C
i
22 . 0
C p
93 . 0
47 . 0
C
U
52 . 0
C
O
1 D
c
c 50
D
L
D
D
gD
P
D
D
D
D
K
D
d
θ
ρ
λ


.

\



.

\



.

\



.

\



.

\

=
−
−
−
(7)
and
65 . 0
C 0 D 1 D
D K K
−
= (8)
where K
D0
depends on feed solids characteristics only.
20 . 0
C
C 1 . 0
45 . 0
C
i
68 . 0
C
O
50 . 0
p
2
C 1 Q f
D
L
D
D
D
D P
D K Q


.

\



.

\



.

\



.

\

=
−
θ
ρ
(9)
For cyclones of Krebs geometry treating identical feeds:
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
185
10 . 0
C 0 Q 1 Q
D K K
−
= (10)
where K
Q0
depends on feed solids characteristics only.
22 . 0
C
C 24 . 0
50 . 0
C
i 27 . 0
53 . 0
C p
40 . 2
C
U
19 . 1
C
O
1 W f
D
L
D
D
gD
P
D
D
D
D
K R


.

\



.

\



.

\



.

\



.

\

=
−
−
−
−
θ λ
ρ
(11)
22 . 0
C
C 24 . 0
25 . 0
C
i
31 . 0
C p
83 . 1
C
U
94 . 0
C
O
1 V v
D
L
D
D
gD
P
D
D
D
D
K R


.

\



.

\



.

\



.

\



.

\

=
−
−
−
−
θ
ρ
(12)
where D
i
is inlet diameter (diameter of circle of some area as cyclone inlet), m
D
O
is overflow (vortex finder) diameter, m
D
U
is underflow (apex) diameter, m
D
C
is cyclone cylinder diameter, m
L
C
is length of cylindrical section, m
θ is cone full angle, degrees
P is feed pressure at inlet, kPa
ρ
p
is feed slurry density, t/m
3
g is acceleration due to gravity, 9.81 m/s
2
Q
f
is flowrate, m
3
/h
R
f
is recovery of water to underflow, %
R
v
is volumetric recovery of feed slurry to underflow, %
K is constant to be estimated from data
λ is hindered settling correction term,
( )
3
v
Cv 1
C
−
C
v
is volumetric fraction of solid in feed slurry
The d
50c
predicted by Equation 7 is inserted in the Whiten function to determine the corrected efficiency
curve for the cyclone:
( ) ( ) ( )
( ) ( )
(
¸
(
¸
− +
− +
=
2 exp x exp
1 exp x 1
C E
*
*
oa
α αβ
α ββ
(13)
where α is an efficiency parameter
β is a fishhook factor
C is recovery of water to overflow, (C = 100R
f
), %
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
186
EXPERIMENTAL WORK
Metallurgical Sampling
In August 1998, two metallurgical sampling were conducted at ‘El Pilon’ grinding circuit (Figure 1)
assuring steady state at least two hours prior surveys. The surveys were conducted under normal
operations.
As illustrated in Figure 1, the sampling points in each survey were:
• New circuit feed,
• Ball mill discharge,
• Cyclone underflow, and
• Cyclone overflow.
CIRCUIT: CURRENT CIRCUIT WINDOWS: Top  TPH Solids Bottom  %Solids
1 Help 2 Operatn 3 Error 4 Test 5 Unit 6 DrawCot 7 Stream8 9 UntFeed 0 RunSIM
CYCLONE
D20
MARCY BALL MILL
8´5”x12´
FEED
TO LEACHING
PROCESS
F
P
SAMPLING POINTS
Figure 1 – Current grinding circuit at ‘El Pilon’ mine
Samples were cut at each of these sampling point every 15 minutes for two hours and then combined into
an overall sample for sieve analysis and solid content measurements. The tonnage in the new feed stream
was also recorded during each survey. Samples were weighed, dried and weighed again to calculate the
percent of solids. The percent of solid was then calculated by:
100
Wt
Wt
solid %
s
∗ = (14)
where Wt
s
is the weight of solids, and
Wt is the weight of slurry.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
187
All the samples from the survey where then sized at ‘El Pilon’ Chemical Laboratory. Each sample was wet
sieved at 45µm to remove any ultrafine particles. The +45mm material was then dry screened at 12500µm,
9520µm, 2000µm, 1700µm, 250µm, 180µm, 150µm, 106µm, 75µm, and 45µm.
RESULTS AND DISCUSSION
The results are presented and discussed into four parts (a) mass balancing, (b) model fitting, (c) simulation
and (d) validation and operation of the new grinding circuit at ‘El Pilon’ mine.
Mass Balancing
Mass balancing is the most useful method of data assessment. JKSimMet mass balancing subsystem is
used to assess data efficiently. The balancing process produces bestfit estimates of flowrates and a set of
adjusted size which is consistent with those flowrates. Overall, JKSimMet mass balancing technique
minimises the sum of weighed squared errors (JKTech 1989), that is:
¿¿ ¿
= = =


.

\
 −
+


.

\
 −
=
N
1 j
L
1 i
L
1 i i
i i
ij
ij ij
a A
x X
SSQ
σ σ
(15)
where N is the number of measurements,
L is the number of streams,
X is a measurement,
x is the adjusted measurement,
A is the measured flow,
a is the adjusted flow, and
σ
ij
and σ
i
are the weights or standard deviations for the measurements and flows respectively.
Based on experimental results and personal judgement, the standard deviations are assigned to the data as
follows:
• For fractions greater than 10%, a standard deviation of 1.0% is assumed.
• For fractions less than 1%, a standard deviation of 0.1% is assumed.
• For fractions between 1% and 10%, a standard deviation of 0.1% plus one tenth of the fraction is
assumed.
To mass balance the grinding circuit data, the new circuit feed is assumed to be accurate (i.e. contain zero
error). It should be noted that the mass balancing program tells how the experimental data (including both
the stream data and sizing data) meet the mass conservation criterion. It does not tell which data are
accurate or inaccurate. Neither can it give the true values. The accuracy of the experimental data depends
entirely on the experimental design and handling.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
188
Based on the criterion described above, the adjusted data were very similar to the experimental values. The
mass balancing results are presented in Figure 2 and the grinding performance is summarised in Table 1.
During the survey, the milling operation was observed to have a very poor performance. Under the
conditions of the survey, a large recycle load in the mill 697.4% is found and also the volume percent
solids in the ball mill discharge too high 52.7%. These numbers confirm the observation why a large
amount or coarse particles were rejected at the end of the mill. Using the Bond power model (Eq. 16), the
power draw for the mill is calculated to be 281.24 HP under the conditions of the survey. It should be
noted that the power draw is 6.3% lower than the installed maximum (300 HP) while a reasonable safety
margin for normal plant fluctuations may be 10%.
tph
80 80
rate feed Mill
F
10
P
10
Wi Power ⋅


.

\

− ⋅ = (16)
0
10
20
30
40
50
60
70
80
90
100
0.01 0.10 1.00 10.00 100.00
Particle Size, mm
C
u
m
u
l
a
t
i
v
e
%
P
a
s
s
i
n
g
Feed BM Discharge Cyc U/F Cyc O/F
marks ~ Experimental Data
lines ~ Adjusted Data
Figure 2 – Mass balancing of the grinding circuit at ‘El Pilon’ mine
Table 1 – Grinding circuit performance at ‘El Pilon’ mine (August 1998)
Stream Solids Water SG Pulp V Flowrate % Passing 75µm
New Circuit Feed 24.95 96.5 0.91 2.63 9.81 3.50
Ball Mill Discharge 173.99 75.6 56.16 1.95 118.30 11.09
Cyclone Underflow 149.04 73.7 53.19 1.90 106.42 3.40
Cyclone Overflow 24.95 19.0 106.36 1.14 115.27 57.00
Units [ton/h] [%] [ton/h] [ton/m
3
] [m
3
/h] [%]
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
189
Model Fitting
The mass balanced data were used to calibrate the mathematical models of ball mill and cyclone included
in the JKSimMet (Version 4) simulator. It is important to know the structure of the models and what
parameters have to be fitted. These subjects are presented in details in the JKSimMet manual (JKTech,
1989). The ore parameters should be determined from ore characterisation tests of representative samples.
Because the main objective of this study is simulation and the ore characterisation tests was not available,
the standard appearance function was chosen. Thus, the machine parameters (e.g. breakage rates) are
calculated from survey data.
No significant differences were found in the calibration of ‘El Pilon’ models. Thus, the ball mill and
cyclone parameters are assumed to represent the grinding circuit within a 30% error. Once the models
have been calibrated to ‘El Pilon’ grinding circuit, the behaviour of the circuit over a wide range of
operational conditions and designs can be predicted.
Simulation
The calibrated models were used to simulate three different flowsheets by addition of a new ball mill 9x9
ft, as described below. The new ball mill operating data are not available; however, it can be assumed that
the 8.5x12ft mill parameters can be used in the new ball mill because of the same ore treated. Overall, the
simulations described below were performed at two new feed rates 700 tpd (29.2 tph) and 800 tpd (33.3
tph) and large circulating loads keeping constant the percent solid (19.0%) in the final grinding product as
required by the metallurgical staff of ‘El Pilon’ mine. Therefore, the ball mill and cyclone design
parameters (e.g. %solids, vortex, apex, etc.) were adjusted to achieve the objectives of this study  an
increase in grinding throughput and fineness. Because ball mills are of the overflow discharge type, each
simulation was performed considering the volume percent solid not exceed 52% to avoid the slurry
transport problems.
Grinding Circuit in Parallel
Figure 3 shows the simulated circuit in which the two mills are operating in parallel each one with a
classification stage in closed circuit.
The simulation results are presented in Table 2 which highlight that this design in parallel increases the
grinding throughput. However, the particle size in the final product (% passing 75µm) increases only 8%
and it decreases 1% as increasing the grinding throughput from 700 tpd to 800 tpd.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
190
1 Help 2 Operatn 3 Error 4 Test 5 Unit 6 DrawCot 7 Stream8 9 UntFeed 0 RunSIM
CIRCUIT: GRINDING IN PARALLEL WINDOWS: Top  TPH Solids Bottom  % Solids
TO LEACHING
PROCESS
FEED
F
BALL MILL
9´x9´
BALL MILL
8´5”x12´
CYCLONE
D20
33.3
95.3
16.7
95.3
69.0
77.5
69.0
77.5
52.3
73.1
69.0
43.1
16.7
95.3 69.4
77.6
52.7
73.2
69.4
43.2
69.4
77.6
16.7
19.0
16.7
19.0
33.3
19.0
CYCLONE
D20
P
Figure 3 – Simulated grinding circuit in parallel at ‘El Pilon’ mine
Table 2 – Simulation results of the grinding circuit in parallel
700 tpd 800 tpd
Stream Solid Flow
Rate, tph
%Solids
% Passing
75µm
Solid Flow
Rate, tph
%Solids
% Passing
75µm
Ball Mill 8.5x12 Circuit
BM Discharge 58.9 74.2 23.52 69.0 77.5 21.94
Cyclone Overflow 14.6 19.0 67.01 16.9 19.0 66.05
Ball Mill 9x9 Circuit
BM Discharge 59.2 74.3 23.32 69.4 77.6 21.75
Cyclone Overflow 14.6 19.0 66.82 16.8 19.0 65.85
Grinding Circuit in SemiSerie
This simulated flowsheet is named ‘semiserie’ because the first mill (8.5x12 ft) is operating in open
circuit following by a classification stage and the second mill in inverse closed circuit with classification,
as illustrated in Figure 4.
The ball mill discharge is fed into a first classification stage (cyclone D20) which the underflow is mixed
with the second cyclone underflow and fed into the second mill (9x9 ft). The 9x9mill discharge is sent
into a secondary classification stage (cyclone D20). Then, the two cyclone overflows are sent to the
leaching process.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
191
1 Help 2 Operatn 3 Error 4 Test 5 Unit 6 DrawCot 7 Stream8 9 UntFeed 0 RunSIM
CIRCUIT: GRINDING IN SEMISERIE WINDOWS: Top  TPH Solids Bottom  % Solids
F 33.3
95.3
33.3
79.7
33.3
79.7
CYCLONES
D20
BALL MILL
8´5”x12´
BALL MILL
9´x9´
142.
70.3
23.3
51.3
142.
70.3
118.
75.9
142.
50.6
33.7
34.1
10.0.
19.0
23.3
19.0
33.3
19.0
142.
70.3
TO LEACHING
PROCESS
FEED
P
Figure 4 – Simulated grinding circuit in semiserie at ‘El Pilon’ mine
Simulation results of this design are presented in Table 3. As may be seen, the solid flow rate in the first
cyclone overflow is too low. The simulator predicts a very low operating pressure in this cyclone 5.34 kPa
at 700 tpd and 6.57 kPa at 800 tpd. Under these conditions, the first cyclone is operating as a simple
splitter increasing the final particle size in the cyclone overflow.
Table 3 – Simulation results of the grinding circuit in semiserie
700 tpd 800 tpd
Stream Solid Flow
Rate, tph
%Solids
% Passing
75µm
Solid Flow
Rate, tph
%Solids
% Passing
75µm
Ball Mill 8.5x12 Circuit
BM Discharge 29.2 79.7 38.92 33.3 79.7 36.63
Cyclone Overflow 8.7 19.0 76.94 10.0 19.0 76.41
Ball Mill 9x9 Circuit
BM Discharge 109.0 63.3 20.25 142.0 70.3 13.85
Cyclone Overflow 20.5 19.0 66.64 23.3 19.0 56.55
However, the particle size in the second cyclone overflow is coarse 66.6% passing 75µm at 700 tpd and
56.6% at 800 tpd. Overall, the particle size in the final grinding product is affected by the performance of
the second cyclone. These results were achieved using a top ball size of 70% at 76.2mm and 30% at 50.8
mm in both mills. Subsequent simulations were conducted by decreasing the top ball size (100% at 50.8
mm) in the second mill however 4.7 tph (112.8 tpd) of 2mm particles (> 2000µm) remains without being
breakage.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
192
Grinding Circuit in Serie
In this design the first mill (8.5x12 ft) is operating in closed circuit with a classification stage and the
second mill (9x9 ft) is in inverse closed circuit with classification (Figure 5). As showed in Figure 5, the
first cyclone overflow is combined with the second mill discharge and fed into a second classification
stage to avoid overgrinding. Then, the second cyclone overflow is sent to the leaching process.
CIRCUIT: GRINDING IN SERIE WINDOWS: Top  TPH Solids Bottom  % Solids
1 Help 2 Operatn 3 Error 4 Test 5 Unit 6 DrawCot 7 Stream8 9 UntFeed 0 RunSIM
F 33.3
95.6
183.
79.8
114.
41.3
183.
79.8
183.
54.5
150.
77.0
33.3
23.7
114.
41.3
80.2
80.0
80.2
81.4
33.3
19.0
P
CYCLONE
D20 CYCLONE
D20
MARCY BALL MILL
8´5”x12´
BALL MILL
9´x9´
TO LEACHING
PROCESS
FEED
Figure 5 – Simulated grinding circuit in serie at ‘El Pilon’ mine
The simulations in this flowsheet were conducted using the following relationship in ball size: (1) first mill
(8.5x12 ft), 70% at 76.2 mm and 30% at 50.8mm, and (2) second mill (9x9 ft), 100% at 50.8mm. The
simulated results are presented in Table 4. As observed in this table, this simulated design achieves the
objectives of ‘El Pilon’ mine. However, under the conditions of 19.0% of solids in the first cyclone
overflow, the recycle load represents a physical limitation for a stable operation in the first mill and
reduces the amount of solids in the regrinding circuit. Although, this simulation was conducted at 24% of
solid in the first cyclone overflow, is strongly suggested to increase the pulp density in the first cyclone
feed up to 68% of solids. Unfortunately, high pulp density in the cyclone feed may increase the
maintenance costs and this is a subject that needs to be evaluated.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
193
Table 4 – Simulation results of the grinding circuit in serie
700 tpd 800 tpd
Stream Solid Flow
Rate, tph
%Solids
% Passing
75µm
Solid Flow
Rate, tph
%Solids
% Passing
75µm
Ball Mill 8.5x12 Circuit
BM Discharge 182.0 79.7 13.0 216.0 80.0 11.4
Cyclone Overflow 29.2 19.0 56.6 33.3 19.0 53.9
Ball Mill 9x9 Circuit
BM Discharge 63.8 76.1 20.6 82.5 79.9 17.8
Cyclone Overflow 29.2 19.0 76.0 33.3 19.0 74.3
Selection of the New Grinding Design
The grinding circuit in serie showed the best performance as presented in Table 5. The proposed flowsheet
achieves the objectives of ‘El Pilon’ mine by increasing both grinding throughput and fineness. This
flowsheet increases the grinding throughput from 600 tpd to 700 tpd and increases the particle size in the
final product from 57% to 75% passing 75µm.
Table 5 – Evaluation of the simulated flowsheets at ‘El Pilon’ mine
Objectives of ‘El Pilon’ mine
Simulated Flowsheets
Grinding Throughput % Passing 75µm
Grinding in Serie
Grinding in SemiSerie
Grinding in Parallel Without change
The grinding circuit in serie shows the following advantages compared with the other two simulated
flowsheets:
• It is possible to increase the grinding throughput without affecting the particle size in the final product,
• The second mill can be operated with small ball size, which increase the fineness of the overall
grinding circuit at ‘El Pilon’ mine.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
194
GRINDING CIRCUIT IN SERIE – ITS VALIDATION AND OPERATION
Based on the analysis of the simulated flowsheets, the grinding circuit in serie was implemented in the
plant. New survey was conducted in the grinding circuit which was used to validate the simulation results.
The new sampling showed that this circuit is milling 800 tpd and the particle size in the final product is
76% passing 75µm.
Figure 6 shows the predicted particle size distribution using the JKSimMet computer package and the real
particle size distribution based on the survey data of the grinding circuit in operation. The new particle size
distribution in the circuit feed is similar with that distribution used in the simulations.
Particle Size, mm
C
u
m
u
l
a
t
i
v
e
%
P
a
s
s
i
n
g
40
50
60
70
80
90
100
0.01 0.1 1.0 10.0 100.0
Size Distribution of Final Grinding Product
Survey: 809 tpd, 20.1% Solids
Simulation: 800 tpd, 19.0% Solids
Figure 6 – Comparison of the particle size distribution of the final product in the grinding circuit in serie at
‘El Pilon’ mine
The real operation (survey data) and the simulated values are in 10% of confidentially. Therefore, the
JKSimMet computer simulator proves a valuable tool in the design of new grinding circuit.
Optimisation
Optimal parameters were optimised to maintain a more stable operation in the implemented grinding
circuit using the JKSimMet computer simulator. The effects investigated in the optimisation exercise were;
(1) ball size in the second ball mill, (2) maximum grinding throughput in the grinding circuit in serie, and
(3) optimal dimensions of vortex and apex in cyclones.
The simulation results shows that the following ball size should be used in mills; 100% at 76.2mm in the
first mill (8.5x12 ft) and 100% at 38.1mm in the second mill (9x9 ft). Additional simulations were
conducted by changing the ball size in the second mill as presented in Figure 7.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
195
New Feed Flowrate, tph
%
P
a
s
s
i
n
g
2
0
0
m
e
s
h
(
7
5
µ µ µ µ
m
)
C
y
c
l
o
n
e
1
U
n
d
e
r
f
l
o
w
,
t
p
h
80
100
120
140
160
180
200
60
65
70
75
80
85
800 850 900 950 1000
Ball Size Effect in Regrind Ball
Mill (9x9 ft)
2” (50.8 mm)
1 1/2” (38.1 mm)
33.3 35.4 37.5 39.6 41.7
New Feed Flowrate, tpd
Figure 7 – Effect of ball size in the second mill of the grinding circuit in serie
As you may see in Figure 7, the ball size in the second mill has control of the particle size in the final
grinding product. As expected, by increasing the grinding throughput a decrease in the particle size is
observed. However, the first cyclone underflow flowrate increases as increase the grinding throughput.
Again, this could be a limitation in the first ball mill performance. To avoid this effect, a large percent of
solid in the cyclone feed and vortex diameter would be used instead.
By decreasing the ball size in the second mill from 50.8 mm to 38.1mm an increase of 4.0% is observed.
The result highlights that the grinding circuit would be able to operate up to 1000 tpd keeping the particle
size in the final product as desired. However, the maximum percent of solids in the first cyclone and the
pump capacity should be investigated.
CONCLUSIONS
A grinding model developed at ‘El Pilon’ mine using JKSimMet (Version 4) was used to simulate three
different milling designs. Simulation shows that the grinding circuit in serie has better advantages in terms
of milling performance than the grinding circuit in parallel or the grinding circuit in semiserie. An
increase of grinding throughput from 600 tpd to 800 tpd was achieved by implementing a grinding circuit
in serie as well as an increase of the grinding fineness from 57% to 75% passing 75µm.
This case study shows how JKSimMet can be used to design grinding circuit accuracy and highlights that
this computer simulator is a powerful tool in the optimisation of comminution circuits.
Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine
JKMRC Conference 2001
196
ACKNOWLEDGEMENTS
The author would like to express his appreciation to the following parties for their help and support to
make this research work possible:
• Mr. Hector DavilaSantos, Mr. Armando IbarraAmaya and Mr. Sergio OlivaRangel for their
helpfulness in making ‘El Pilon’ visits possible
• Prof. Alban J Lynch and Prof. Alejandro LopezValdivieso for their supervision
• Ing. Armando Guerrero and laboratory staff for invaluable assistance in performing the testwork
• Consejo Nacional de Ciencia y Tecnologia (CONACyT) of Mexico for Juan L. ReyesBahena’s
scholarship Grant No. 66764/137256.
REFERENCES
Broussand, A., Connil, P., and Fourniguet, G., 1988, “USIM: An easy to use industrial simulator for mineral
processing plants”, Computer Applications in the Mineral Industry, ISBN 906191 7603, Balkema Rotterdam,
137145.
CANMET, 1989. SIMBAL Brochure, CANMET, 552 Booth Street, Ottawa, Ontario, Canada, K1A0G1.
Herbst, J.A., Schena, G.D. and Fu, L.S., 1989. “Incorporating state of the art models into a mineral processing plant
simulator”, Trans. Instn. Min. Metall., 98, C1C11.
JKTech., 1989. JKSimMet User Manual and Supplementary Information Manual (Version 4, Release 1), JKTech,
Indooroopilly, Qld., Australia.
Lynch A.J. and Morrell S. 1992. The understanding of comminution and classification and its practical application in
plant design and optimisation. Comminution: Theory and Practice, Ed. Kawatra, 405426 (AIME).
Lynch, A.J. and Rao, T.C., 1975. “Modelling and scaleup of hydrocyclone classifiers”. In Proceedings XI Int Min
Proc Congress, Cagliari, Paper 9.
Nageswararao K. 1978. Further developments in the modelling and scale up of industrial hydrociclones. PhD Thesis,
University of Queensland (JKMRC).
Whiten, W.J., 1976. “Ball mill simulation using small calculators”. In Proceedings AusIMM, 258, 4753.
The ore is crushed in a primary jaw crusher and transported by conveyor to the secondary and tertiary crushing circuit. Only one cyclone is operating at a time while the other is on standby. is sent to the leaching process. The classification circuit consists of two 0. Silver content ranges from 270 to 320 g/ton and gold from 0. 5. having a pulp density from 50 to 60% solids. JKMRC Conference 2001 . e. This material is conveyed via belt feeders to the 2.59m by 3. Lime addition increases the circuit pH to reduce volatilization of cyanide and to increase the settling of material in the thickener downstream. MODSIM (Herbst et al. The concentrator processes a sulphide orebody consisting of 1. Cyanide is used to leach silver/gold bearing particles throughout the circuit. The crushing circuit product contains material approximately 96% passing 9.20 m by 0.508m (20 inches) diameter cyclone. by means of a 0.20 to 0.31% Fe and 2. Ore is delivered to the mill by trucks and sent directly to the primary crusher.2 mm (3 inches) and 30% of 50. for classification. 1989). Thus. The ore is leached with cyanide to obtain the doré (i. AgAu alloy).182 Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine processing units has considerably increased in last decade.34 g/ton. 1988). Underflow from the cyclone is fed back to the ball mill while the overflow material (19% to 22% solid) with 58% passing 75 µm (200 mesh) is directed to a 19. The leached pulp and its silver/gold bearing solution are treated using the Merrill Crowe process to obtain the doré. is fitted with steel liners. JKSimMet (Version 4) computer package is used in this study to model and simulate the ‘El Pilon’ grinding and classification stages. It uses steel balls in a ratio of 70% of 76. 1989). Underflow of the thickener.0 ft in length) 300 HP Marcy ball mill. typical software packages are USIM PAC (Broussand et al. The grinding circuit uses a twostage cyclone process to classify mill discharge.64% Zn.10% Cu.67m (8 ft ½ inches in diameter by 12.53 mm (3/8 inches). and SIMBAL (CANMET. simulation results are implemented in the plant which were validated with a new sampling campaign. approximately 500 km east of the Guadalajara State in Northern Mexico. ‘EL PILON’ MINE ‘El Pilon’ mine mill is located in the town of San Martin de Bolaños. Current milling rates are approximately 600 ton/day (tpd). The pulp density in the ball mill is maintained at 76% solids.e. This simulator was used to assist in the design of a new grinding circuit with the addition of a new ball mill in order to increase both grinding throughput and fineness. JKSimMet (JKTech. The ball mill. and lime and cyanide are added to the mill feed. Discharge from the ball mill flows to a pump box where it is diluted to a density approximately 48% solids before being sent to the cyclone.81 m (65 foot) diameter thickener located outside the east wall of the mill complex.8 mm (2 inches) as grinding media.64% Pb. with 40% of mill filling and 70% of its critical speed.g. 0.15 m (DRL 8 inches by 6 inches) pump at 870 rpm. 1989).
Simulate the circuit.e. 1976) is quite similar to the general population balance model. as presented in this paper. p i = di s i (2) where. It allows engineers to: • • • • Design a circuit on the graphics monitor. di. The perfect mixed model start with the same population balance. and Graph and print the results. Feed in + Breakage in = Pr oduct out + Breakage out (1) However.Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine 183 GRINDING CIRCUIT SIMULATOR JKSimMet is a powerful tool for analysis and simulation of mineral processing plant data. JKSimMet performs steady state simulation of a range of comminution and classification operations. two factors control the breakage of particles inside the mill as follows: • Selection function which defines the breakage probability of particles. thus selection = ri s i (breakage out) (3) JKMRC Conference 2001 . This model assumes a perfectly mixed mill which removes most of the complexities in the general population balance model. Mathematical Models Process models of the two main units at ‘El Pilon’ grinding circuit are presented: • • Ball mill. where the purpose is to assess the suitability of different flowsheets to achieve a desired performance objective. Enter model and plant data. and Hydrocyclone Ball Mill The perfect mixed model (Whiten. The principal application of JKSimMet is to carry out process analysis and optimisation of existing circuits and also extremely useful for conducting conceptual design studies. is the discharge function in the mill. i. It gives engineers the ability to design and optimise any crushing or grinding circuit including stages of classification.
Lynch and L. the balance equation around each size fraction.20 (9) For cyclones of Krebs geometry treating identical feeds: JKMRC Conference 2001 . subject to a reasonable form of the appearance function (aij). incorporating terms suggested by dimensional reasoning with exponents estimated from data sets obtained by Nageswararao (1978) and Rao (Lynch and Rao. 1978).1 LC DC 0. 1975). P 0.15 (7) and − K D1 = K D0 D C0.93 P ρ p gD C −0. The predictive equations are: d 50c D = K D1 O Dc DC 0. Elber (Nageswararao.65 (8) where KD0 depends on feed solids characteristics only. Hydrocyclone The hydrocyclone model used in this study was that developed by Nageswararao under the supervision of A. This model comprises empirical equations for the main performance criteria.52 DU DC −0.20 θ 0.50 Qf = 2 K Q1D C ρp DO DC 0.68 Di DC 0.47 λ 0. is: fi + i j =1 a ij r j s j = p i + ri s i (5) Now substitute for the experimentally mill contents: i j =1 fi + a ij r j p j dj = p i + ri pi di (6) This means that the ratio ri di can be calculated for each size fraction from a set of actual feed (fi) and product (pi) measurements. which describes the mill behaviour.22 Di DC −0.184 Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine Breakage function which define how the selected particles are broken (appearance function) in the mill content: appearance = a ij r j s j (breakage in) (4) Then.50 LC DC 0.45 θ − 0. J.
% Rv is volumetric recovery of feed slurry to underflow. 9.19 R f = K W1 DU DC DU DC 2. degrees P is feed pressure at inlet. kPa ρp is feed slurry density.27 Di DC −0. m LC is length of cylindrical section.50 θ −0. t/m3 g is acceleration due to gravity.22 (11) −0.83 −0. v (1 − Cv )3 Cv is volumetric fraction of solid in feed slurry The d50c predicted by Equation 7 is inserted in the Whiten function to determine the corrected efficiency curve for the cyclone: E oa = C (1 + ββ x )(exp(α ) − 1) exp(αβ x ) + exp(α ) − 2 * * (13) where α is an efficiency parameter β is a fishhook factor C is recovery of water to overflow.25 −0.94 1. m DO is overflow (vortex finder) diameter. DO DC DO DC −1.24 LC DC 0.53 λ0.Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine 185 − K Q1 = K Q 0 D C0.10 (10) where KQ0 depends on feed solids characteristics only.31 R v = K V1 Di DC θ − 0. m DC is cyclone cylinder diameter. m3/h Rf is recovery of water to underflow.24 LC DC (12) where Di is inlet diameter (diameter of circle of some area as cyclone inlet).22 0. % JKMRC Conference 2001 . (C = 100Rf).81 m/s2 Qf is flowrate. % K is constant to be estimated from data C λ is hindered settling correction term.40 P ρ p gD C P ρ p gD C −0. m θ is cone full angle. m DU is underflow (apex) diameter.
two metallurgical sampling were conducted at ‘El Pilon’ grinding circuit (Figure 1) assuring steady state at least two hours prior surveys. dried and weighed again to calculate the percent of solids.% Solids P TO LEACHING PROCESS CYCLONE D20 F FEED MARCY BALL MILL 8´5”x12´ SAMPLING POINTS 1 Help 2 Operatn 3 Error 4 Test 5 Unit 6 DrawCot 7 Stream 8 9 UntFeed 0 RunSIM Figure 1 – Current grinding circuit at ‘El Pilon’ mine Samples were cut at each of these sampling point every 15 minutes for two hours and then combined into an overall sample for sieve analysis and solid content measurements. The surveys were conducted under normal operations. The percent of solid was then calculated by: %solid = Wt s ∗ 100 Wt (14) where Wts is the weight of solids. Cyclone underflow. the sampling points in each survey were: • • • • New circuit feed.186 Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine EXPERIMENTAL WORK Metallurgical Sampling In August 1998.TPH Solids Bottom . JKMRC Conference 2001 . The tonnage in the new feed stream was also recorded during each survey. CIRCUIT: CURRENT CIRCUIT WINDOWS: Top . and Cyclone overflow. Samples were weighed. As illustrated in Figure 1. and Wt is the weight of slurry. Ball mill discharge.
e. x is the adjusted measurement. RESULTS AND DISCUSSION The results are presented and discussed into four parts (a) mass balancing. (c) simulation and (d) validation and operation of the new grinding circuit at ‘El Pilon’ mine. Neither can it give the true values. 75µm. The balancing process produces bestfit estimates of flowrates and a set of adjusted size which is consistent with those flowrates. the new circuit feed is assumed to be accurate (i. a is the adjusted flow. X is a measurement. JKMRC Conference 2001 . and σij and σi are the weights or standard deviations for the measurements and flows respectively. A is the measured flow. The +45mm material was then dry screened at 12500µm. For fractions between 1% and 10%. and 45µm.1% is assumed. (b) model fitting. 1700µm. 2000µm. 180µm. a standard deviation of 0. Overall. 9520µm. Based on experimental results and personal judgement. 106µm. JKSimMet mass balancing subsystem is used to assess data efficiently. 150µm. a standard deviation of 1.0% is assumed. 250µm. To mass balance the grinding circuit data. JKSimMet mass balancing technique minimises the sum of weighed squared errors (JKTech 1989). L is the number of streams. Mass Balancing Mass balancing is the most useful method of data assessment. The accuracy of the experimental data depends entirely on the experimental design and handling.Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine 187 All the samples from the survey where then sized at ‘El Pilon’ Chemical Laboratory. Each sample was wet sieved at 45µm to remove any ultrafine particles. a standard deviation of 0. For fractions less than 1%. the standard deviations are assigned to the data as follows: • • • For fractions greater than 10%. contain zero error). that is: SSQ = N L X ij − x ij j =1 i =1 σ ij + L i=1 A i − ai σi (15) where N is the number of measurements.1% plus one tenth of the fraction is assumed. It should be noted that the mass balancing program tells how the experimental data (including both the stream data and sizing data) meet the mass conservation criterion. It does not tell which data are accurate or inaccurate.
7%. mm BM Discharge Cyc U/F 10.3% lower than the installed maximum (300 HP) while a reasonable safety margin for normal plant fluctuations may be 10%.19 106.63 1. The mass balancing results are presented in Figure 2 and the grinding performance is summarised in Table 1. the adjusted data were very similar to the experimental values.01 0. During the survey.04 24.00 Cyc O/F 100.5 75. 10 P80 10 F80 Power = Wi ⋅ − ⋅ Mill feed rate tph (16) 100 Cumulative % Passing 90 80 70 60 50 40 30 20 10 0 0.81 118.6 73. the milling operation was observed to have a very poor performance.99 149.0 [%] Water 0.14 [ton/m ] 3 V Flowrate 9.09 3.40 57.24 HP under the conditions of the survey.95 [ton/h] 96.27 [m /h] 3 % Passing 75µm 3.90 1.00 [%] JKMRC Conference 2001 . It should be noted that the power draw is 6.30 106.95 1.16 53. Under the conditions of the survey. Using the Bond power model (Eq.10 Feed marks ~ Experimental Data lines ~ Adjusted Data 1. the power draw for the mill is calculated to be 281. a large recycle load in the mill 697.42 115.95 173.7 19.91 56.4% is found and also the volume percent solids in the ball mill discharge too high 52.00 Particle Size.36 [ton/h] SG Pulp 2. These numbers confirm the observation why a large amount or coarse particles were rejected at the end of the mill. 16).00 Figure 2 – Mass balancing of the grinding circuit at ‘El Pilon’ mine Table 1 – Grinding circuit performance at ‘El Pilon’ mine (August 1998) Stream New Circuit Feed Ball Mill Discharge Cyclone Underflow Cyclone Overflow Units Solids 24.50 11.188 Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine Based on the criterion described above.
an increase in grinding throughput and fineness. Thus.3 tph) and large circulating loads keeping constant the percent solid (19. the machine parameters (e.5x12ft mill parameters can be used in the new ball mill because of the same ore treated. Overall. the behaviour of the circuit over a wide range of operational conditions and designs can be predicted. However. breakage rates) are calculated from survey data. etc. the ball mill and cyclone parameters are assumed to represent the grinding circuit within a 30% error. The new ball mill operating data are not available.g. Thus.2 tph) and 800 tpd (33. Because the main objective of this study is simulation and the ore characterisation tests was not available. %solids. 1989).g. the standard appearance function was chosen. The ore parameters should be determined from ore characterisation tests of representative samples. apex. Grinding Circuit in Parallel Figure 3 shows the simulated circuit in which the two mills are operating in parallel each one with a classification stage in closed circuit.0%) in the final grinding product as required by the metallurgical staff of ‘El Pilon’ mine.Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine 189 Model Fitting The mass balanced data were used to calibrate the mathematical models of ball mill and cyclone included in the JKSimMet (Version 4) simulator. JKMRC Conference 2001 . however. Simulation The calibrated models were used to simulate three different flowsheets by addition of a new ball mill 9x9 ft. the particle size in the final product (% passing 75µm) increases only 8% and it decreases 1% as increasing the grinding throughput from 700 tpd to 800 tpd. as described below. It is important to know the structure of the models and what parameters have to be fitted. each simulation was performed considering the volume percent solid not exceed 52% to avoid the slurry transport problems. These subjects are presented in details in the JKSimMet manual (JKTech. The simulation results are presented in Table 2 which highlight that this design in parallel increases the grinding throughput. the ball mill and cyclone design parameters (e. No significant differences were found in the calibration of ‘El Pilon’ models. Once the models have been calibrated to ‘El Pilon’ grinding circuit. Because ball mills are of the overflow discharge type. Therefore.) were adjusted to achieve the objectives of this study . it can be assumed that the 8. the simulations described below were performed at two new feed rates 700 tpd (29. vortex.
0 77.3 95.5 BALL MILL 8´5”x12´ 1 Help 2 Operatn 3 Error 4 Test 5 Unit 6 DrawCot 7 Stream 8 9 UntFeed 0 RunSIM Figure 3 – Simulated grinding circuit in parallel at ‘El Pilon’ mine Table 2 – Simulation results of the grinding circuit in parallel 700 tpd Stream Solid Flow Rate.2 14.6 19.6 %Solids % Passing 75µm 23.3 73.3 19.82 Solid Flow Rate. tph BM Discharge Cyclone Overflow BM Discharge Cyclone Overflow 58.2 16.5 33.3 16.0 Ball Mill 9x9 Circuit Grinding Circuit in SemiSerie This simulated flowsheet is named ‘semiserie’ because the first mill (8.TPH Solids Bottom .5x12 ft) is operating in open circuit following by a classification stage and the second mill in inverse closed circuit with classification.01 23.0 77.3 69.32 66.0 69.5x12 Circuit 74.2 CYCLONE D20 16.3 19.0 43.1 P TO LEACHING PROCESS 69.9 14.6 BALL MILL 9´x9´ 69.75 65.% Solids 52.6 F FEED 33.5 19. as illustrated in Figure 4.190 Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine CIRCUIT: GRINDING IN PARALLEL WINDOWS: Top .3 52. The ball mill discharge is fed into a first classification stage (cyclone D20) which the underflow is mixed with the second cyclone underflow and fed into the second mill (9x9 ft).2 19.7 73.7 95.0 74. the two cyclone overflows are sent to the leaching process.4 77. JKMRC Conference 2001 . Then.4 16.7 19.8 800 tpd %Solids % Passing 75µm 21. The 9x9mill discharge is sent into a secondary classification stage (cyclone D20).94 66.0 16.7 19.85 Ball Mill 8. tph 69.7 95.52 67.0 CYCLONE D20 69.4 77.1 69.6 59.4 43.9 69.0 77.0 16.0 77.05 21.
JKMRC Conference 2001 .0 79. Overall.3 33.0 70.3 800 tpd %Solids % Passing 75µm 36.7 19. tph 33.92 76.7 34.3 10.8 mm) in the second mill however 4. These results were achieved using a top ball size of 70% at 76. 70.1 BALL MILL 9´x9´ 142.55 Ball Mill 8.3 33.3 79.3 19.3 19. the solid flow rate in the first cyclone overflow is too low. tph BM Discharge Cyclone Overflow BM Discharge Cyclone Overflow 29.5x12 Circuit 79.0 Ball Mill 9x9 Circuit However. 70.0 63.0 142.% Solids TO LEACHING PROCESS 10.3 19. the particle size in the final grinding product is affected by the performance of the second cyclone.6% at 800 tpd. the particle size in the second cyclone overflow is coarse 66.0 33. the first cyclone is operating as a simple splitter increasing the final particle size in the cyclone overflow.25 66.3 51.7 142.7 19.3 FEED 142.94 20.7 tph (112.0 23.8 mm in both mills.5 %Solids % Passing 75µm 38. Subsequent simulations were conducted by decreasing the top ball size (100% at 50.8 tpd) of 2mm particles (> 2000µm) remains without being breakage.2mm and 30% at 50. 50.63 76. 70.7 109.7 33.0 P 23.3 BALL MILL 8´5”x12´ 1 Help 2 Operatn 3 Error 4 Test 5 Unit 6 DrawCot 7 Stream 8 9 UntFeed 0 RunSIM Figure 4 – Simulated grinding circuit in semiserie at ‘El Pilon’ mine Simulation results of this design are presented in Table 3. 75.0.85 56.3 79.3 95.9 142. Under these conditions.57 kPa at 800 tpd.64 Solid Flow Rate. As may be seen.0 CYCLONES D20 23.0 20.3 19.2 8.6 F 33.34 kPa at 700 tpd and 6.3 118. 19.41 13.6% passing 75µm at 700 tpd and 56.TPH Solids Bottom .Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine 191 CIRCUIT: GRINDING IN SEMISERIE WINDOWS: Top . Table 3 – Simulation results of the grinding circuit in semiserie 700 tpd Stream Solid Flow Rate. The simulator predicts a very low operating pressure in this cyclone 5.
54. JKMRC Conference 2001 . the first cyclone overflow is combined with the second mill discharge and fed into a second classification stage to avoid overgrinding. this simulation was conducted at 24% of solid in the first cyclone overflow. under the conditions of 19.6 BALL MILL 9´x9´ 80.5x12 ft). As showed in Figure 5.5x12 ft) is operating in closed circuit with a classification stage and the second mill (9x9 ft) is in inverse closed circuit with classification (Figure 5).8 MARCY BALL MILL 8´5”x12´ 80.2 80. 41.2 mm and 30% at 50.0 183. 79.0 114. 77. The simulated results are presented in Table 4. the recycle load represents a physical limitation for a stable operation in the first mill and reduces the amount of solids in the regrinding circuit. As observed in this table. 41.% Solids 33.3 95. 100% at 50.2 81. 79. Unfortunately. Then.0 CYCLONE D20 P TO LEACHING PROCESS CYCLONE D20 FEED 150. the second cyclone overflow is sent to the leaching process. is strongly suggested to increase the pulp density in the first cyclone feed up to 68% of solids.4 F 33.192 Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine Grinding Circuit in Serie In this design the first mill (8.0% of solids in the first cyclone overflow. and (2) second mill (9x9 ft).TPH Solids Bottom .3 183. this simulated design achieves the objectives of ‘El Pilon’ mine. Although.8mm. high pulp density in the cyclone feed may increase the maintenance costs and this is a subject that needs to be evaluated.3 19.8 114.8mm.7 33.3 23.3 1 Help 2 Operatn 3 Error 4 Test 5 Unit 6 DrawCot 7 Stream 8 9 UntFeed 0 RunSIM Figure 5 – Simulated grinding circuit in serie at ‘El Pilon’ mine The simulations in this flowsheet were conducted using the following relationship in ball size: (1) first mill (8.5 183. However. 70% at 76. CIRCUIT: GRINDING IN SERIE WINDOWS: Top .
0 Ball Mill 9x9 Circuit Selection of the New Grinding Design The grinding circuit in serie showed the best performance as presented in Table 5. The proposed flowsheet achieves the objectives of ‘El Pilon’ mine by increasing both grinding throughput and fineness. The second mill can be operated with small ball size.5x12 Circuit 79.0 29.0 80.5 33.9 19.3 82.0 56. JKMRC Conference 2001 . tph 216.Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine 193 Table 4 – Simulation results of the grinding circuit in serie 700 tpd Stream Solid Flow Rate.2 63.7 19.8 74.3 800 tpd %Solids % Passing 75µm 11.1 19.0 Solid Flow Rate.9 17. Table 5 – Evaluation of the simulated flowsheets at ‘El Pilon’ mine Simulated Flowsheets Grinding in Serie Grinding in SemiSerie Grinding in Parallel Without change Objectives of ‘El Pilon’ mine Grinding Throughput % Passing 75µm The grinding circuit in serie shows the following advantages compared with the other two simulated flowsheets: • • It is possible to increase the grinding throughput without affecting the particle size in the final product.2 %Solids % Passing 75µm 13.4 53.3 Ball Mill 8.6 76.6 20. tph BM Discharge Cyclone Overflow BM Discharge Cyclone Overflow 182.0 19. which increase the fineness of the overall grinding circuit at ‘El Pilon’ mine.0 33.0 76. This flowsheet increases the grinding throughput from 600 tpd to 700 tpd and increases the particle size in the final product from 57% to 75% passing 75µm.0 79.8 29.
and (3) optimal dimensions of vortex and apex in cyclones.2mm in the first mill (8.01 Size Distribution of Final Grinding Product Survey: 809 tpd. Optimisation Optimal parameters were optimised to maintain a more stable operation in the implemented grinding circuit using the JKSimMet computer simulator. mm 10.0 Figure 6 – Comparison of the particle size distribution of the final product in the grinding circuit in serie at ‘El Pilon’ mine The real operation (survey data) and the simulated values are in 10% of confidentially.1mm in the second mill (9x9 ft).0 Particle Size. The new sampling showed that this circuit is milling 800 tpd and the particle size in the final product is 76% passing 75µm. Additional simulations were conducted by changing the ball size in the second mill as presented in Figure 7.0% Solids 0.194 Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine GRINDING CIRCUIT IN SERIE – ITS VALIDATION AND OPERATION Based on the analysis of the simulated flowsheets. 100% at 76. 20. New survey was conducted in the grinding circuit which was used to validate the simulation results. the JKSimMet computer simulator proves a valuable tool in the design of new grinding circuit. The simulation results shows that the following ball size should be used in mills. the grinding circuit in serie was implemented in the plant.0 100. Therefore. (2) maximum grinding throughput in the grinding circuit in serie. JKMRC Conference 2001 . 100 90 Cumulative % Passing 80 70 60 50 40 0.1 1. The effects investigated in the optimisation exercise were.5x12 ft) and 100% at 38. 19.1% Solids Simulation: 800 tpd. Figure 6 shows the predicted particle size distribution using the JKSimMet computer package and the real particle size distribution based on the survey data of the grinding circuit in operation. The new particle size distribution in the circuit feed is similar with that distribution used in the simulations. (1) ball size in the second ball mill.
tph Figure 7 – Effect of ball size in the second mill of the grinding circuit in serie As you may see in Figure 7. As expected.6 1000 33. the first cyclone underflow flowrate increases as increase the grinding throughput. a large percent of solid in the cyclone feed and vortex diameter would be used instead. This case study shows how JKSimMet can be used to design grinding circuit accuracy and highlights that this computer simulator is a powerful tool in the optimisation of comminution circuits. Again. the ball size in the second mill has control of the particle size in the final grinding product.8 mm to 38. JKMRC Conference 2001 Cyclone 1 Underflow. tph 80 .8 mm) 1 1/2” (38. Simulation shows that the grinding circuit in serie has better advantages in terms of milling performance than the grinding circuit in parallel or the grinding circuit in semiserie.0% is observed. by increasing the grinding throughput a decrease in the particle size is observed.4 41.7 New Feed Flowrate. tpd 37. An increase of grinding throughput from 600 tpd to 800 tpd was achieved by implementing a grinding circuit in serie as well as an increase of the grinding fineness from 57% to 75% passing 75µm. To avoid this effect.Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine 195 85 200 180 160 140 % Passing 200 mesh (75 µ m) 75 70 Ball Size Effect in Regrind Ball Mill (9x9 ft) 2” (50. The result highlights that the grinding circuit would be able to operate up to 1000 tpd keeping the particle size in the final product as desired.5 39. CONCLUSIONS A grinding model developed at ‘El Pilon’ mine using JKSimMet (Version 4) was used to simulate three different milling designs.1 mm) 120 100 80 65 60 800 850 900 950 New Feed Flowrate. this could be a limitation in the first ball mill performance. However. the maximum percent of solids in the first cyclone and the pump capacity should be investigated. By decreasing the ball size in the second mill from 50.3 35. However.1mm an increase of 4.
The understanding of comminution and classification and its practical application in plant design and optimisation. CANMET.C. JKTech. 1975. and Fu. P. Ed. 1989. 1989. 1978.. Alejandro LopezValdivieso for their supervision Ing. Mr. Kawatra. • • • REFERENCES Broussand. 552 Booth Street. Sergio OlivaRangel for their helpfulness in making ‘El Pilon’ visits possible Prof.D. ReyesBahena’s scholarship Grant No. 1988. “Ball mill simulation using small calculators”. G. Trans. Ontario. Alban J Lynch and Prof.J. A. 1976.A. Nageswararao K. Comminution: Theory and Practice. 405426 (AIME). SIMBAL Brochure. CANMET. 1992. Lynch A. “Modelling and scaleup of hydrocyclone classifiers”. JKMRC Conference 2001 . Hector DavilaSantos.S. “Incorporating state of the art models into a mineral processing plant simulator”.J. Whiten. PhD Thesis. “USIM: An easy to use industrial simulator for mineral processing plants”. University of Queensland (JKMRC).J. 1989. and Rao. Australia. Ottawa. W. and Fourniguet. C1C11. and Morrell S. 137145..196 Modelling and Simulation of the Grinding Circuit at ‘El Pilon’ Mine ACKNOWLEDGEMENTS The author would like to express his appreciation to the following parties for their help and support to make this research work possible: • Mr. J... Herbst. ISBN 906191 7603. Balkema Rotterdam. Armando Guerrero and laboratory staff for invaluable assistance in performing the testwork Consejo Nacional de Ciencia y Tecnologia (CONACyT) of Mexico for Juan L. Armando IbarraAmaya and Mr.. Indooroopilly. Qld. Schena. Computer Applications in the Mineral Industry. Paper 9. Canada. Cagliari. In Proceedings XI Int Min Proc Congress. JKSimMet User Manual and Supplementary Information Manual (Version 4. 66764/137256. Lynch. Release 1). Metall. A.. 98. In Proceedings AusIMM. G. 258.. Further developments in the modelling and scale up of industrial hydrociclones. Instn. Min.. 4753. T. Connil... JKTech. K1A0G1. L.