Alex G.

Doll
C O N S U L T I N G L T D

TECHNICAL MEMORANDUM
December 12, 2011

TO: UBC MINE331 CLASS

FROM: ALEX DOLL, ALEX G DOLL CONSULTING LTD.
SUBJECT: SAG MILL + BALL MILL CIRCUIT SIZING

Summary:
Most large tonnage mineral processing facilities consist of a primary SAG milling and secondary ball milling
grinding circuit. There are three principal methodologies for sizing SAG mills that are widely used:
• Bond work index based methods;
• Australian “drop weight test” based methods;
• The SPI/SAGdesign methods.
Each methodology involves a series of grindability laboratory tests and a suite of equations used for mill power
and throughput predictions. This presentation deals only with the Bond work index suite.
The desired outcome of all of these methods is an estimate of the grinding specific energy required to reduce a
rock from a “F80” size to a “P80” size. This energy is expressed as kW·h/tonne, where kW is the grinding energy re-
quired “at the mill shell” multiplied by the inverse of throughput (h/tonne). This energy requirement (for a par -
ticular F80 and P80) is a property of the rock and any difference in specific energy of different comminution
devices is (largely) reflective of the efficiency of the grinding device.
Once the specific energy is determined, the amount of grinding power needed (at the mill shell) is found by mul-
tiplying the specific energy by the desired throughput. Then a motor size and mill geometry is chosen that will
satisfy the power demand at the mill shell.

The Bond Work Index

Before the 1950's the grinding industry had no ability to scale-up grinding results from laboratory to industrial
mill sizes, or to reliably estimate the specific energy when either a feed or product size changed. The industry had
been measuring specific energy consumption for decades, but didn't really know how to make reliable predictions
involving them1.
Fred Bond worked at the Allis Chalmers mineral processing equipment company in the middle part of the 20 th
century. He had access to one of the best equipped grinding laboratories in the world, and access to data from
hundreds of operating plants all over North America. His employer, Allis Chalmers, was one of the largest sup -
pliers of grinding equipment. Allis Chalmers had a huge stake in selling “correctly” sized grinding equipment, so
they were motivated to do research find better ways to scale up laboratory grindability results into full-sized
mills.
What Bond developed was a simple formula that expressed specific energy consumption with three parameters:
the feed size, product size, and a grindability measurement of the ore. This measurement became known as a
“work index”2. Bond's formula is usually presented in this form:

E =10×Wi
(√ 1
−
1
P 80 √ F 80 ) (1)

1 For the purists, there were two models in existence at the time: the Von Rittinger and Kick models. Neither of
these gave results that matched the observations of specific energy consumption from tumbling mills.
2 F.C. Bond; The Third Theory Of Comminution, Transactions AIME 1952. Available from www.onemine.org

MINE 331 -1- 2011-12-12

where:
• E is the specific energy consumption, kWh/tonne;
• Wi is the work index, unitless, but applicable only to metric calculations;
• F80 is the 80% passing size of the feed, µm;
• P80 is the 80% passing size of the product, µm.
The work index laboratory measurement can be made with a variety of equipment. The most common is the
Bond ball mill work index:

Figure 1: Ball mill for determining Bond ball mill work index
Many people refer to “the” Bond work index when they really mean the Bond ball mill work index. There are two
other Bond work index measurements commonly used: the rod mill work index and the crushing (low-energy im-
pact) work index.

Figure 2: Rod mill for determining Bond rod mill work index

MINE 331 -2- 2011-12-12

So why are there three machines measuring three different versions of the same index? Because the nature of
rock breakage changes with size. Coarse rock breakage is dominated by fractures, whereas medium-size rocks
break through the rock matrix, and fine-size rocks break along and through mineral grains. To capture these
three mechanisms, the Bond method makes use of three tests, each suited to a particular size range and type of
breakage.
Table 1: Typical size ranges applicable to the Bond work index tests

Feed Size, F80 Product Size, P80

Crushing work index, WiC 100 mm 10 mm
Rod mill work index, WiRM 10 mm 1500 µm
Ball mill work index, WiBM 1500 µm 100 µm

Determining Overall Circuit Specific Energy Consumption

To predict the cumulative breakage from a coarse size to a fine size requires doing three calculations using Equa-
tion 1, one for each of the three work index measurements and handing-off intermediate sizes per Table 1. But
before doing that calculation, there are some correction factors that need to be applied.
Chet Rowland worked with Bond's original equations and found that some adjustments were needed in certain
applications where machines were being operated outside of their range of highest efficiency. These “EF” factors
are tabulated in many texts1 and not all are used in particular calculations. Be aware that many old texts refer to
work index on a short ton basis, and these must be converted to a metric tonne basis to perform modern calcula-
tions.
Of interest to this style of calculation are the Rowland EF4 and EF5 factors for rod milling and ball milling. The
following equations are for metric work index values and particle sizes in µm:

[ ( )]
(0.907×Wi RM −7) F 80
EF4 RM = 1+ −1

( ) ( )
0.5
F 80 14.33 (2)
16000
P 80 Wi RM

[ ( )]
(0.907×Wi BM −7) F 80
EF4 BM = 1+ −1

( ) ( )
0.5
F 80 14.33 (3)
4000
P 80 Wi BM

[
EF5 BM = P 80 +
10.3
1.145×P 80 ] (4)

Values of EF4 can only be greater than 1. If a value evaluates to less than 1, skip that factor and substitute 1 in its
place. The EF5 factor only applies below 75 µm to ball milling, and a value of 1 can be substituted above that P 80.
Combining the Rowland efficiency factors with the three-stage Bond equation gives this overall equation for
grinding specific energy consumption from a primary crusher product size, F 80, to a ball mill cyclone overflow
product size, P80.

E=10×Wi C
(√ 1
−
1
)
10000 √ F 80
+10×Wi RM
( √ 11500 − √101000 )×EF4 RM +10×Wi BM
(√ 1
−
1
P 80 √ 1 500 )
×EF4 BM×EF5 BM

(5)

1 C. Rowland. Chapter 23: The Design and Installation of Comminution Circuits. AIME 1982

MINE 331 -3- 2011-12-12

This energy is the “maximum efficiency” comminution energy that would be associated with highly efficient com-
minution machines: typically multi-stage crushing, rod milling and ball milling or a Noranda-style multi-stage
crushing and single-stage ball milling circuit. Barratt, 1989, writes that a power-efficient circuit of SAG milling
followed by ball milling should be 10% less efficient that the “maximum efficiency” grinding circuit (crushing, rod
mill, ball mill). Thus, the Etotal of a SAG and ball mill circuit is:
E total =1.10×E (6)

Partitioning the Total Energy Between SAG and Ball Mills

The Bond-based method applied by Barratt involves deducting a SAG mill specific energy consumption, E SAG,
from the overall circuit specific energy consumption, E total, with the difference being the ball mill specific energy
consumption, Ebm. One of the advantages of this approach is it accounts for the tendency for SAG mills to make
more fines than a crusher or rod mill would in the same duty. These fines are passed to the ball mill sump and
into the cyclone where they exit the circuit without actually appearing in the ball mill. This results in an apparent
reduction in ball mill specific energy consumption versus what would be predicted using the straight Bond for-
mula and the ball mill work index. This phenomenon is often called the phantom cyclone because of how some
of the other modelling approached account for the extra fines.
The specific energy consumption for the SAG mill can be estimated using Barratt's 1979 formula (which was re -
published in the 1989 SAG conference1):

[
ESAG = 10×Wi C
(√ 1
−
1
)
PC √ F 80
+10×Wi RM
(√ 1
−
1
PR √PC )
×EF4 RM +10×Wi BM
(√ 1
−
1
110 √ P R) ]
× EF4 BM ×EF5 ×1.25
(7)

( )
1 1
−10×Wi BM − × EF4BM × EF 5
√ 110 √ T 80
The form of this equation is similar to Equation 5, except that it introduces a few new terms and it includes a
factor of 1.25 that accounts for the difference in grinding efficiency of a SAG mill versus a ball mill. The original
Barratt papers do not specify values to use for P C and PR, but we can use the same values as Table 1 of 10 000 µm
and 1500 µm, respectively. The value T80 is the “synthetic” transfer size between the SAG mill and ball mill – a
fines-corrected 80% passing size 2 of the SAG mill product (or the combined SAG mill product and crushed
pebbles if the circuit is sending crushed pebbles to the ball mill circuit, a.k.a. SABC-B operation).
The transfer size is normally set to a value in the range of 1 to 3 mm for a “SABC-A” circuit where the pebbles dis-
charging the SAG mill are screened and crushed in a pebble crusher, then the crushed pebbles are returned to the
SAG mill feed. The recommended design value is 2 mm for T80 in most calculations.
Pebble crushers are considered to be part of the SAG mill portion of the circuit in these calculations. Normally,
the energy that pebble crushers input to the ore breakage is insignificant (on the order of 2% of the overall
SAG+pebble crusher power) and can be neglected.
The SAG mill specific energy consumption is deducted from the total circuit specific energy consumption to get
the ball mill specific energy consumption:
E bm =E total −E SAG (8)

Determine Mill and Motor Size

Knowing the SAG and ball mill specific energy consumption allows mill sizes to be calculated if a desired
throughput rate is known. Note the units of specific energy: kW·h/tonne. This is power (kW) times the reciproc-
al of throughput (t/h).

1 D.J. Barratt. An Update On Testing, Scale-up and Sizing Equipment for Autogenous and Semi-Autogenous
Grinding Circuits. 1989 SAG Conference.
2 The actual transfer size you measure in a plant survey will be coarser than this synthetic value. To determine
the fines-corrected T80 from a survey value, perform a “phantom cyclone” or “reduced-recovery” calculation
(see Barratt, 1989).

MINE 331 -4- 2011-12-12

The power in Bond calculations is the power as seen by the ore at the shell of the mill. This is often called the
“power at the pinion” because smaller mills use a pinion and gear arrangement to transmit motor power to the
mill. A gear and pinion have mechanical losses, meaning the actual motor output power will be greater than the
grinding calculations which operate in “power at the pinion”1.
The SAG mill motor will normally operate at 90% of its rated output power. Combining this design criteria with
the pinion/gear efficiency (use 0.985 for a synchronous motor, pinion and gear; use 1.000 for gearless) provides
the following equation to estimate the motor size for a gear-driven SAG mill:
Total SAG Motor Power = (ESAG × tonnes/hour) ÷ 0.985 ÷ 0.90 (9)
The ball mill motor can be sized using a similar equation, but ball mills normally operate at a higher proportion
of their rated power output, use 0.94. The ball mill gear/pinion efficiency is the same as for a SAG mill. For a
gear-driven ball mill, use the following:
Total Ball Mill Motor Power = (Ebm × tonnes/hour) ÷ 0.985 ÷ 0.94 ( 10 )
The largest motor sizes currently on the market with pinion and gear drives are in the 10 MW range (these would
be a twin-motor arrangement with one motor on either side of the mill). Gearless drives are available for over
20 MW motor output (23.5 MW is running at Esperanza in Chile and 28 MW drives are under construction for
mines in Peru and Australia). If your motor power exceeds these sizes, split the duty into multiple mills of the
same type.

Determining Mill Dimensions

Nominal mill diameters are specified inside the mill shell, before the liners are installed. Two types of mill length
are commonly used in the industry.
• “Effective Grinding Length,” or EGL, is the length of the mill cylinder where grinding can occur. This is
generally the length from the inside of the mill feed end liner to the mill discharge grate. This is the pre -
ferred designation of process engineers.
• “Flange to Flange Length,” or F/F, is the length of the cylinder of the mill as seen from the outside flange
positions where the feed and discharge cones are bolted to the cylinder. This is the preferred designation
for mechanical engineers.
The detailed calculation methods for determination of mill dimensions to yield a particular power draw (at a mill
shell) is beyond the scope of this Memorandum. Instead, two ways to size mills are offered:
1. Look in literature for a mill that has the same motor as you have calculated. Plagiarize Copy that design.
Don't use any mills from Africa – they use a completely different design (low-aspect-ratio and high
speed). Also, don't use an iron mine as a basis for a copper or gold project because the density difference
in the ore affects the mill designs. Be aware that both the major vendors quote mill lengths as flange-to-
flange, so you must convert your SAG mill effective grinding length to F/F by adding 2 feet to the EGL
and your ball mill by adding half a foot to the effective grinding length.
2. For SAG mill only, use the diagram in Figure 3 and choose a point on or to the right of the diagonal blue
line. The nominal mill diameter is given by the colour of the points, and the effective grinding length can
be determined using the X-axis formula.

Pebble Crushers
For initial design purposes, always assume a pebble crusher is required. Determine the size of the pebble crusher
by choosing a crusher that can treat the volumetric flow rate equivalent to 20% of the SAG mill feed. Download
vendor catalogues from their websites to see the volumetric capacities of crushers set at a 13 mm closed side set-
ting.

1 See http://www.sagmilling.com/articles/TechMemorandum-MotorPower-Rev4.pdf for more discussion of

power measurements at different parts of a grinding mill electrical & control circuitry.

MINE 331 -5- 2011-12-12

 Figure 3: SAG Mill Motor Sizes Versus Function of Diameter (ft) and Effective Grinding Length (ft)1

Example: Tenke Fungurume

Composite 1 has the following grindability results (J. Starkey et al, CMP 2007):
• WiBM=8.3 (metric)
• WiRM=12.7 (metric)
• WiC=9.0 (metric)
Estimate the specific power consumption to grind primary crusher feed of 150 mm to a flotation feed size of
150 µm.

E=10×Wi C
( √ 10000
1
−
√ 150000 )
1
+10×Wi
( √11500− √101000 )×EF4
RM RM +10×Wi BM
(√ 150
1
−
√ 1 500 )
1
×EF4 BM ×EF5 BM

( 11 )
The EF4 for rod milling evaluates to 0.72, so it doesn't apply. The EF5 for ball milling doesn't apply because the
product size is coarser than 75 µm, and the EF4 for ball milling evaluates to 0.61, so it also doesn't apply. Use a
value of 1 for all the EF factors.
Equation 11 simplifies to: E = 0.67 + 2.01 + 4.63 = 7.31 kWh/tonne
Total SAG and ball mill circuit specific energy consumption: Etotal =8.04 kWh/tonne
Determine the SAG mill specific energy consumption using Equation 7:

[
ESAG = 10×Wi C
(√ 1
−
1
)
10 000 √ F 80 ( √ 1500
+10×Wi RM
1
−
√ 10 000 )
1
×1+10×Wi
( √110
BM
1
−
√ 1500 )
1
]
×1 ×1 ×1.25

−10×Wi
( √ 110
BM
1
−
√
1
2000 )×1×1
( 12 )
Equation 12 simplifies to: ESAG = (0.67 + 2.01 + 5.77) × 1.25 – 6.06 = 4.50 kWh/tonne

1 Doll & Barratt, Choosing the Right Motors for your Mills, Procemin 2010

MINE 331 -6- 2011-12-12

Determine the ball mill specific energy consumption using Equation 8:
Ebm = Etotal - ESAG = 8.04 - 4.50 = 3.54 kWh/tonne ( 13 )
Determine the SAG and ball mill motor sizes required for 2000 tonnes/hour. Using Equations 9 and 10:
Total SAG Motor Power = (4.50 × 2000) ÷ 0.985 ÷ 0.90 = 10 156 kW ( 14 )
Total Ball Mill Motor Power = (3.54 × 2000) ÷ 0.985 ÷ 0.94 = 7 647 kW ( 15 )
Rounding off, set the SAG mill motor to 10 MW. On Figure 3, a 10 MW SAG mill motor equates to about 100 on
the X-axis. This size motor is also in the regime of 34 foot diameter SAG mills. Use this value and the X-axis for -
mula to determine the mill length:
2.5
100= 34 ×EGL ( 16 )
1000
This simplifies to EGL = 100 000 ÷ 6741 = 14.8
Round this off to a SAG mill effective grinding length of 15 feet. To compare to a vendor table, add 2 feet to con -
vert the EGL to a flange-to-flange length: 17 feet.
Set the ball mill motor to 7.7 MW, and look for a motor of a similar size in the vendor literature. The vendors typ-
ically quote motor sizes in US horsepower, so look for a motor of around 10 350 hp. FLSmidth installed a
10 500 hp ball mill at Minera Los Pelambres – that is close enough. The Pelambres mill was 21 feet diameter and
33.5 feet flange-to-flange length. Convert this length to EGL by deducting half a foot, and you have a 21 feet dia-
meter by 33 feet EGL ball mill.
End result:
• SAG mill: 34 ft diameter by 15 ft EGL with 10 MW of motor output power
• Ball mill: 21 ft diameter by 33 ft EGL with 8 MW of motor output power
• Add a pebble crusher. Metso MP800 is suitable for this duty.

Exercise #1
Tenke Fungurume Composite 2 has the following grindability results (J. Starkey et al, CMP 2007):
• WiBM=10.4 (metric)
• WiRM=13.5 (metric)
• WiC=11.10 (metric)
Determine a SAG + ball mill circuit to grind primary crusher product of 150 mm to a flotation feed size of 200 µm
at 2500 tonnes/hour.

Exercise #2
Using the milling circuit you determined for Exercise #1, what would be the throughput of Composite #1 if you
feed 150 mm F80 rock to the SAG mill and make a ball mill product P 80 of 200 µm? Assume the SAG mill draws
90% of available power and the ball mill 94% of available power.

MINE 331 -7- 2011-12-12