You are on page 1of 19

RCA Minnitt

School of Mining Engineering


Private Bag 3, WITS. 2050.
Telephone: 011 717 7416
Mobile: 082 481 2357
Mining in Africa 2008 Conference
8-12 September 2008
NASREC, Johannesburg
Sampl e, sampl es, sampl ed,
sampl i ng
Sampl i ng Topi cs i n Repor t i ng
Codes
Sampl i ng er r or s
Accurate
& Precise
Accurate
& Imprecise
Inaccurate
& Precise
Accuracy
(Mean close
to true value, unbiased)
Reproducibility
(Variance, biased)
What you want:
Representativeness;
Unbiased, accurate
What you get: Real Life!
Biased, inaccurate, variable
+
Bias
Inaccurate
&
Imprecise
every particle in the lot has an
equal
chance of being in the sample
Component s of r epr esent at i veness
Representative Sampling : the defining, critical issue for all
materials description, study, characterisation and analysis as well
as in process monitoring and industrial process control.
Representativeness - impossible after the sampling event
Only understanding of the processes ensures representivity
Representivity - two components accuracy and precision
Representativeness (r) = Sample is accurate + Sample is
precise
The only scientific definition:
Representativeness (r) = r
2
(SE) = m
2
(SE) + s
2
(SE)
Wor ki ng mi ni mumf or Theor y of
Sampl i ng ( ToS)
Sampling errors reduced, sampling bias eliminated
Authoritative rules and guidelines provided by the
Theory of Sampling (ToS)
Seven Sampling Errors
Seven Sampling Unit Operations (SUO)
Seven sampl i ng er r or s
1. Fundamental Error (FE), related to CH, i.e. the material
properties of the lot
2. Grouping and Segregation Error (GSE) related to DH,
i.e. due to grouping and segregation in the lot
3. Increment Delineation Error (IDE) - geometry of outlined
increment is not completely recovered; can be completely
eliminated
4. Increment Extraction Error (IEE) - material extracted
does not coincide with the delineated increment; can be
completely eliminated
5. Increment Preparation Error (IPE), - all sources of non-
stochastic variation after extraction of the material; error
should always be nil
6. Increment Weighing Error (IWE), may be relevant or not,
depending on specific context
7. Continuous Selection Error (CSE), Random, Time, and
Cyclical fluctuations
M
a
t
e
r
i
a
l

c
h
a
r
a
c
t
e
r
i
s
a
t
i
o
n
E
q
u
i
p
m
e
n
t

a
n
d

H
a
n
d
l
i
n
g
P
r
o
c
e
s
s


a
n
d

P
r
o
c
e
d
u
r
e
Seven Sampl i ng Uni t Oper at i ons
Three general principles used once in optimizing sampling
procedures
1. Material characterisation Heterogeneity testing
2. Reduction of Lot dimensions
3. Characterization of process variability using replicate tests and
variography
Four practical procedures used several times during
practical sampling
1. Lot or sample homogenization by mixing or blending
2. Composite sampling, using the smallest possible increments
3. Particle size reduction (comminution)
4. Representative mass reduction
Gy s f or mul a f or Fundament al
Er r or
A nomogram is a summary of:
Fragment
size
Mass of
lot
Mass of
sample
Relative
variance

c
c

M
/ t
/ t
L L
l = (d
l
/d
N
)

f = 0.5 to 0.2
g = always 0.50
d
95
= always 95%

2
2
FE
FE
=
=
c
c

l
l

f
f

g
g

d
d
N
N
3
3
M
M
s
s
How t o sampl e l ar ge l ot s?
(Source: ACABS Research Group site
www.acabs.dk)
Appl i cat i on of var i ogr aphy
Mixing and blending reduces the Grouping and Segregation Error
(GSE)
Seven Sampl i ng Uni t
Oper at i ons
Grouping factor:
smaller
increments help
Segregation
factor: Mixing
helps
2 2
( ) ( ) s G S E = s F E
1 = 0 =
Mixing and blending - intuitively clear, almost trivial but very
important
F G F
G G
N - N N
=
N - 1 N

Vr edendal Wi ner i es
Sampl i ng by t he r ope met hod
Simulatio exemplum - credit: Hans-Henrik Friis Pedersen
- effects of reduction of avr. grain size, d by
comminution
6d 3d
1d
- which is (most)
representative?
Par t i cl e si ze r educt i on
- sample: 30 grammes
30 gram vs. 30.000 tonnes
30.000.000.000 gram
sampling rate: 1: 1.000.000.000 1: 10
9
sampling rate: 1: 10
2
1:10
9
QUESTI ONS i n need of answer s ! !
How good a sample is 1 kg of
rock with 1 cm top-size
fragments?
The lab crushes my gold ore
sample to 1.7 mm and takes a
500 g split. Is this always OK?
How would the lab know?
How much money should be
spent on sampling and sampling
equipment?
What is the impact of lousy
sampling and assaying on the
annual profit of a mine?
50t of rock
50g aliquot
How do we go from
50t of rock to a 50g
aliquot without losing
accuracy and
precision and still
preserve the integrity
of the sample?
Het er ogenei t y t est - used f or
nomogr am
Dat a: Lat ent i nf or mat i on
Entire chain of evidence vs. lot characteristics is
flawed without representative / correct sampling
Sampling errors reduced, sampling bias eliminated
by respecting authoritative rules, guidelines, and
structured framework provided by the Theory of
Sampling (ToS)
As a suggestion: Maybe we should be working
towards a more structured approach to sampling in
the Reporting Codes