AH

HISTORY
Y OF EUR
RUS MIN
NERAL
CONS
SULTAN
NTS AND
D
DEV
VELOPM
MENT OF
F KinCalcc and
S
SUPASIM
M
The South Africa
an Mining Industry
I
in
n the 1990ss
Eurus M
Mineral Con
nsultants de
eveloped ass a result off a specific need
n
in thee minerals in
ndustry.
Prior to
o 1996, the leading
l
Min
ning Housess in South Africa
A
histo
orically held
d a large con
ntingent
of engin
neers and in
n-house con
nsultants w
which was a repository
y for the ind
dustrys sub
bstantial
and valluable technical know
wledge that had been built-up
b
ov
ver many deecades. Fo
ollowing
global ttrends and rising costts, these deepartments were axed over a nu
umber of ye
ears and
from ab
bout 2000, the
t industry
y consultan
nt became an
a integral part of the Mining lan
ndscape,
often ass a SME (sm
mall to mediium enterprrise).

The Deevelopmentt of KinCa

alc and SUP
PASIM in the 1980s
The usee of flotatio
on kinetics and the deevelopmentt of KinCallc and SU
UPASIM began
b
in
1985 in Gold Fields Laboratorries [GFL] (n
now SGS, Lakefield
L
La
aboratories)) with the arrival of
the firstt commerciaally availab
ble desk top
p PC. The MS-DOS
M
sysstem alloweed program
mming in
Turbo P
Pascal on a low resoluttion green aand plasma
a yellow screen. At thaat time, Golld Fields
of South
h Africa [G
GFSA] (unbu
undled in 19998 to form
m Gold Field
ds Limited, GFL) had acquired
a
the Norrtham PGM
M property down-dip
d
frrom JCIs (n
now Anglo Platinum) A
Amandelbu
ult mine.
It later transpired that this was
w quite a coup and the result of an intrigguing and stealthy
dawn rraid on lapssed explora
ation and m
mining rightts on the Lo
ondon Stocck Exchange
e. Until
the new
w millenniu
um the plattinum indu
ustry mainta
ained an offten fanaticcal level of secrecy.
Bringin
ng the depo
osit into pro
oduction po
osed some major issues which w
were not resolvable
with cu
urrent know
wledge or tecchniques ass they existeed at that tim
me.
The deeposit began
n at about 1,200m, a total of only 60kg of
o borecore was available for
designing the plan
nt and no plant operatiing data, eq
quipment sizes or test data was available
a
dance and benchmarki
b
ing. Togeth
her with Miining house employeess, mill, flota
ation cell
for guid
and pum
mp supplieers were sub
bject to legaal action if they divulg
ged any infoormation to
o a third
party. The probleem was how
w to size a plant and predict final concentrrate mass, recovery
r
o for whicch there waas no plant scale data from otherr operationss and no
and graade for an ore
Bond w
work index tests could
d be done aas this wou
uld consum
me at least a third of the
t core
sample.
m sizing was a deca
ade away frrom being developed.. Sizing
The JK test on borrecore for mill
float plant from 1k
kg laborato
ory rate testts was based on variou
us thumb-su
uck method
dologies
g plant and laboratory
y final conce
entrate recoovery derived from
using scale-up facttors linking
udies on orres of particcular characcteristic. Due
D to the lack
l
of opp
portunity to
o widely
case stu
benchm
mark, scale--up factorss from theese limited
d case stud
dies were considered
d to be
universsally correctt, whereas they actually
y varied wiidely and were
w
mostly conflicting.

Necesssity and Inv

vention
Mill and float circu
uit sizing an
nd predictio
on of flotation perform
mance was oone of eithe
er take a
a
attemp
pt to dress it up as a scientificc proceduree or develop new
complete guess and
dologies th
he latter beiing a classicc situation of
o necessity
y is the motther of invention.
method
Using ssamples fro
om Gold Fie
elds gold an
nd base meetal operatio
ons, method
eveloped
ds were de
using 11kg samplees to determ
mine operaating work index usin
ng GFLs laaboratory mill
m and
flotation
n kinetics from
f
a standard rate teest in a 2.5l Denver D12
D float celll. Outlier samples
with paarticularly low/high Bond work
k index an
nd ones possessing hiigh mass pull
p
and
low/hig
gh metal reecovery during previou
us pilot plan
nt campaign
ns were obttained from
m Mintek
on a no
o-name basiis. These provided
p
exxtreme poin
nts and faciilitated mod
delling from
m which
variouss scale-up faactors were
e determineed from lab
boratory to plant. 8 caase studies of gold,
pyrite, ccopper, pollymetallic base metal aand tin operrations valid
dated the m
methodologies.
Having
g found Kelssalls two-co
omponent eequation sp
pecifying fasst and slow
w fractions and rates
in a 19661 paper an
nd a similarr equation in
n a paper by
b Jowett in
n 1960, a prrogram (late
er called
KinCalcc) was writtten in Turb
bo Pascal by
y John Foxcrroft to inpu
ut rate test d
data and de
etermine
flotation
n kinetics. A mass balancing pro
ogram was then writte
en for a flotaation circuiit with a
degree of flexibilitty. Inputs were flotattion kineticcs, feed ton
nnage and p
pulp densitty, head
grade aand numberr and size of
o flotation
n cells. Thiss was eventtually calleed SUPASIM
M. John
was a g
graduate cheemical engiineer whosee programm
ming skills were
w
taughtt to him by his wife
who haappened to be one of
the earrly and few
w women
computter program
mmers at
the timee. They em
migrated to
Edmon
nton, Canad
da in the
late 19880s - one off the many
loses off high calib
bre skill to
other co
ountries thaat remains
a dilem
mma for South Africa.
SUPASIIM was validated
using ttest and plant
p
data
from tthe pyritee, copper,
polymeetallic base metal and
tin flotaation circuitts detailed
in the case studiees. Some
false sttarts highliighted the
need for a reelationship
n float celll air rate
between
and a w
water rate constant.
c
In
n wide-ran
nging plant trials on UG2
U
and Meerensky ore
es in the
mid-19990s this was later ex
xpanded to
o include fllotation kin
netics to m
maintain prrediction
accuraccy of recovery-grade-m
mass profil es at very low and very
v
high aair rates. The
T
first
conclussions that em
merged werre;

a) Scale up factors for the various fractions and rates of mineral and gangue were
consistent and not ore or circuit dependant and
b) Fast floating fraction of mineral (or metal) was the primary decider of circuit
recovery and represented the only 1:1 relationship between laboratory and plant
performance.
These proved very convenient and made prediction of plant performance a much simpler
task than originally anticipated. Point a) was later seen as obvious, because scale up factors
are a function of the efficiency of particle-bubble contact and subsequent particle recovery
and not the type of mineral and gangue being treated.

20 Months of Aspirin
A brief thought experiment over coffee concluded that, logically, there should be a
relationship between each kinetic parameter and some physical aspect of a plants mass
balance of recovery, grade, mass pull, stream pulp density and flow rate. There indeed was,
but it took the best part of 20 months to fully define these correlations. For each ore type
different circuit configurations, each kinetic parameter was incrementally changed and a
mass balance generated. Each time the mass balance had to be printed out, the result noted
and the next change run. With 4 kinetic parameters for each of the economic minerals and 4
for gangue; 8 ore types each with an average of 2 economic metals, 2 input streams for
kinetics (roughers and cleaners) and 3 circuit configurations per ore, making 7 changes per
kinetic parameter the number of scenarios to be analysed totalled 5,376. At about 14 mass
balances a day, it took much paper, perseverance, two Christmas breaks and many cups of
coffee to define these relationships.
Today, with the Windows operating system, solver, goal seeker and what-if functions in
Excel or using Monte Carlo simulation software this could probably be completed in a few
days about 125x quicker!

First Applications Maranda and Northam

In 1988 bench and pilot scale data were used to size and simulate the sequential Cu-Zn
flotation circuit for the Maranda deposit owned by Metorex. Flotation performance was as
per simulation.
In 1988/9 the circuit to treat Merensky ore for Northam Platinum was sized and simulated.
From start-up in 1993, flotation performance was also as per simulation. However, sizing
the mills was open to significant error because, unlike flotation, it was not possible to
directly measure key parameters that could be used for design. As far as possible 1kg tests
in the GFL laboratory mill on known ores were correlated with Bond work index values.
Once an operating work index for the Northam run-of-mine ore had been determined and
optimum grind established, benchmarking found that the milling circuit at the Chino copper
operation in New Mexico, USA was almost an exact fit in throughput, grind and overall
operating work index to that envisaged for Northam.

Design
ning from Laboratory
L
y Data bee Prepared
d to get it Wrong
W
Chino u
used a welll-establishe
ed pancakee SAG-Ball North American miilling circuit. This
was un
nknown in South Africca and view
wed with much
m
suspicion. The local indusstry was
comforttable with convention
nal circuits u
using threee stage crusshing follow
wed by rod
d-ball or
ball-balll mills and
d the autoge
enous ROM
M sausage
mill developed for ggold plants by Alec
Mokken
n of Gencorr in early 19
970s and op
ptimised by
y Peter Baile
ey into the 1980s. At the
t back
of every
yones min
nd was the failure
f
of th
he countrys only 9.75
5m pancakke mill at Palabora
P
which ssuffered fro
om a cracked
d shell.
Furtherr benchmark
king and fa
act finding w
world-widee of operatio
ons treatingg the same ore
o with
differen
nt milling
g circuits
revealed
d that the difference
in opeerating wo
ork index
between
n ROM prim
mary mills
and miills treating
g 10-20mm
feed
was
reaassuringly
consisteent. Large savings in
capital and operaating cost
way and a
eventuaally held sw
8.53m
x 2.95m ROM
R
SAG
and 5.550m x 6.70m
6
ball
mill ciircuit with
h variable
speed d
drives was installed.
The prim
w designe
ed to treat 2225 t/hr to 30%
3
-75m at
a 35.8 kWh
h/t-75m.
mary will was
Feedbacck from thee companys Mining E
Engineers in
ndicated it most likelyy that ore hardness
h
increasees with dep
pth, so the primary
p
milll was sized with a libe
eral dose of capacity an
nd could
operatee with a maaximum of 18% steel. At full prroduction th
he primary mill unexp
pectedly
settled into autogeenous mode
e producing
g 22% -75
m at 25.3 kWh/t-75m
k
m, drawing
g 900kW
n was 2,400 kW). In terms
t
of efffective millling capacitty of LD2.5 (effective grinding
g
(design
length x diameterr inside she
ell^2.5) thee primary mill
m was so
ome 2.5x oover-sized! It later
hat area of th
he Bushveld
d ore hardn
ness decreasses with dep
pth.
emergeed that in th
It transspired abou
ut 14 years later (Angllo American had since
e acquired JCI to form
m Anglo
Platinum
m) that theey obtained
d an intereest in North
ham. Duriing a periood of strike
e action,
Amand
delbult truck
ked some of their stock
kpiled Mereensky ore to
o Northam to be treate
ed, since
there w
was significaant milling capacity av
vailable. Affter 6 hours of operatioon the prim
mary mill
filled-up and had to be shut down and emptied. Unrelated
U
and
a a few m
months late
er Anglo
m commisssioned EMC to audit all milling
g circuits in
n the grou
up. Amand
delbults
Platinum
6.10m
x 4.88m primary
p
RO
OM Merenssky mill grrinds to 36%-75m an
nd operatess at 39.8
kWh/t--75m. Thee Northam primary m
mill was corrrectly sized
d for an up
p-dip Aman
ndelbult
ore wh
hich laboratory milling
g tests on w
waste samp
ples had ind
dicated, but
ut evidently
y not for
softer d
down-dip Northam
N
ore
e. And all because sufficient sam
mple could n
not be sparred for a
proper Bond work
k index test on reef!

Benchmarking at any cost?

With the amount of risk involved in sizing production-scale plant from laboratory data
based on a newly developed methodology, benchmarking was a critical component of the
evaluation and design process and to do this properly means obtaining similar ore to test.
During the construction of Northam in the early 1990s, the Mine Manager requested
samples from Amandelbults waste rock dump so that GFSA could perform Bond work
index tests to check Northam mill sizing. The pervasive level of secrecy in the Platinum
industry in those days evoked some unusual responses. Within a few days razor wire was
placed on top of the chain-link fence dividing the two properties and both waste rock dump
and tailings dams were patrolled by guards with dogs. In any event, if waste rock had been
obtained, no one would have been any the wiser as it would have confirmed mill size
according to harder Amandelbult ore.
Obtaining a Merensky sample from another operation for flotation testing was more
successful and involved a cloak and dagger approach which has not been revealed outside
of the now defunct Gold Fields of South Africa until now. I had described the development
of SUPASIM and the desire to obtain any amount of Merensky ore to a friend (Ian) who
was sales manager at one of the major pump suppliers. He suggested an obvious approach
of just walking into one of the plants and grabbing material. Being a junior level
Metallurgist, and particularly in a back-water of a research laboratory, no one in Head Office
was particularly interested in my comings and goings. With this advantage of non-descript
employee and under the guise of a trainee, we went off to Rustenburg on one of Ians
routine customer-related plant visits. I was armed with an empty briefcase fitted with thick
plastic sample bags and a sample scoop. In the 1980s security may not have been nonexistent but it was certainly very laid-back. After Ians brief conversation with the
mechanical supervisor, and being introduced as a trainee learning the ropes, we headed into
the plant to do a service check on the pumps. The plant was of 1960s origin with 5 milling
lines consisting of 9ft primary and 8ft secondary ball mills. Typical of design at that time,
the mill feed conveyors began in the bowls of the earth with the tail pulley in a dimly lit
chamber below surface level ideal for sample theft!
With much fear and as rapidly as possible, material was frantically scooped from the mill
feed belt into my plastic lined briefcase, levelling after each scoop. Luckily no one saw
although down there it was highly unlikely and I went to find Ian, swinging my briefcase
in as nonchalant a fashion as possible with the idea that any observer would immediately
note that my briefcase was obviously light and held nothing more than odd papers. The
trouble with doing something you know is wrong is that you think everyone you see
automatically focuses on the item of ones crime.
I had managed to obtain just short of 4.5kg. Back at Gold Fields Laboratories this was
enough to do a rougher and cleaner rate test, produce some batch final concentrate and test
mill 1kg to estimate operating work index. Bristling with success and when enthusiastically
informed, GFSA management was horrified and forbad any mention of the escapade.

Despitee the fact th

hat today th
his level of ssecrecy and
d paranoia is considereed unnecesssary and
even craazy, the reaality at that time was th
hat there wo
ould have been
b
seriouss consequen
nces had
I been ccaught. Tw
wo other hassty grab sam
mples had also
a
been ob
btained from
m other ope
erations;
but und
der the circcumstances it was bestt not to mention this. The engraained system
m of old
style m
managementt expected conformity
y rather th
han initiativ
ve from a M
Metallurgisst at D1
Paterso
on grade.

The firsst sortie had

h
been in
nto the Fraank concenttrator in Ru
ustenburg aand the oth
hers into
Mortim
mer south off Amandelb
bult. Flotatiion circuit size
s
and configuration
n, feed tonn
nage and
recoverries were ap
pproximately known sso, importa
antly for pro
ogram deveelopment a further
three vaalidations could be claiimed.
Whilst auditing Anglo Platinum milling
g and flotatiion plants in
i 2004/5, I stood in th
he same
y that tail pulley
p
in Frrank concen
ntrator and mused oveer the contribution
spot as in 1988 by
that sam
mple had made
m
to the initial deveelopment off SUPASIM
M. Togetheer with oth
her small
plants iin the area with high operating
o
c ost, the Fra
ank concenttrator was d
decommissiioned in
2009.

The Esstablishmeent of EMC

C and Dev
velopment of KinCallc and SU
UPASIM
in the
1990s a
and 2000s
SUSPAS
SIM conttinued to be
b develop
ped, but more
m
as a side-line,
s
in
n the 1990ss whilst
Consultting Metalllurgist at Lonmin.
Averagee PGM re
ecovery foor the com
mpanys
concenttrators increeased at the
e rate of 0.775% per annum. The companys manageme
ent style
and stro
ong emphaasis on high
h operating standards were
w
the main drivingg factors. KinCalc
K
and SU
UPASIM played a parrt by enablin
ng the flota
ation processs to be bettter understo
ood and
being able to simu
ulate and mo
odify propo
osed changees. Lonmin
ns slag treaatment conccentrator
nd commisssioned in 1995 at th
he Western
n Platinum
m smelter w
was design
ned and
built an
perform
mance simulated from laboratory
l
sscale tests using
u
SUPA
ASIM.
EMC w
was started in February
y 2002 in reesponse to changing conditions
c
iin the locall mining
industry as describ
bed at the beginning
b
o
of this history. The com
mpany wass fortuitous enough
d 90% of the
e work for that calend
dar year;
from daay 1 to starrt with five clients who
o provided
Anglo P
Platinum, Cluff
C
Platinu
um (becamee Ridge Min
ning who were
w
bough
ht later by Aquarius
A
Platinum
m), Harmon
ny Gold, OOkiep Cop
pper and Veenmyn Rand
d (later renaamed Venm
myn).
Anotheer fortuitou
us event was meeting
g Dave Wiseman, the
e developerr of LIMN
N, at the
Flotatio
on 03 confeerence in Helsinki.
H
Latter in the year
y
the old
d Turbo Paascal KinCalc and

SUPASIIM prograams were trranslated in

nto more usser friendly
y Excel spreead sheet fo
ormat by
Dave. In 2004, EM
MC boughtt LIMN wh
hich Dave customised to accomm
modate SUPASIM.
misation con
ntinued ove
er the next two years with
w
the ad
ddition of oother functions and
Custom
calculattion routinees. EMC refers to the ccustomised version of LIMN as SU
SUPASIM-in
n-LIMN,
or SIL.
The M
MS Excel veersion of
KinCalcc was a vast
improv
vement on the old
Turbo P
Pascal prog
gram but
could only accom
mmodate
8-10 setts of test daata before
it
slo
owed-up.
The
translattion into MS
M Excel
format had beeen done
ve on a
rapidly by Dav
rough-aand-ready
basis
which made it ussable but
m
new
required a great many
ns. As proccessing of
function
client d
data grew, flotation
rate tesst data wou
uld arrive
in batcches of 200-60 and
several hundred seets of test
data were being processed
p
num to deteermine flotation kinetiics. These data then had
h to be m
manually ta
abulated
per ann
with asssociated kinetic ratios calculated
d. Apart frrom this be
ecoming an
n onerous task
t
the
possibillity of data input errorr was alway
ys a factor. A new verssion of KinC
Calc was required
r
which w
would be laargely autom
mated.
A third
d fortuitouss event wa
as coming into conta
act with Ad
drian Jardiin who, lik
ke Dave
Wisemaan, was a Metallurg
gist by deegree but had augm
mented thiis with high-level
program
mming skillls. Adrian sat
s in EMC offices for the
t best parrt of 11 mon
nths whilst KinCalc
was up
pgraded into Visual Ba
asic. The p
project was supposed to take 3 m
months but as each
new veersion camee out, testing on the lattest set of client
c
data highlighted
h
d the need for
f extra
function
ns. A few of
o these werre,

Sorting and
d ordering of
o data in sp
pecific wayss,
Highlightin
ng individu
ual data setss to be transsferred to an
a average p
page and th
he result
put back in
nto the inpu
ut page for rre-estimatio
on of kineticcs,
A facility to
t change g/t
g into % to accomm
modate clients who reeported basse metal
assays in pp
pm instead of %,
The facility
y to re-run a data set w ith certain analytes
a
exccluded via a tick-box.

The first time saver was to develop an import wizard to automatically transfer data in any
excel format into KinCalc following which the program would calculate, tabulate and graph
the subsequent results and kinetics. Probably counter-intuitively, this proved to be amongst
the biggest headaches. Character and word recognition functions had to be programed in to
accommodate clients use of acronyms, abbreviations, aliases, spelling mistakes and foreign
language alphabet.
The fourth fortuitous event, was meeting Andy Holloway at the annual Canadian Mineral
Processors conference in Ottawa, 2004 (another quality South African Metallurgist who had
decided to move to Canada). Andy introduced me to the scroll bar facility of Excel. This
became one of the largest and most useful components of KinCalc as the ScrollCalc
function. With two exponential functions, Kelsalls equation fits a specific shape of
recovery-time curve. Due to mineral composition and mineral-gangue associations not all
ores adhere to this profile, especially nickel ores where the metal can be distributed between
fast, medium and slow floating nickel-bearing minerals. ScrollCalc is a manual means of
adjusting the visual fit between calculated and actual profiles.

EMC, KinCalc and SUPASIM owes much to others

Serendipity and the input of others has enabled a company to be formed that, based on the
continuing work received, provides a useful service.
Isaac Newton famously remarked in a letter to his rival Robert Hooke dated February 5,
1676 that:
"What Descartes did was a good step. You have added much several ways, and especially in
taking the colours of thin plates into philosophical consideration. If I have seen a little further
it is by standing on the shoulders of Giants."
Source: Wikipedia