Muestreo Metalurgico

Sampling in the South African minerals
industry
by R.C.A. Minnitt*
cuttings from reverse circulation or blast-hole

drilling, to the collection of dust and fine
particulate products of drilling and rock
Synopsis
cutting. The focus in dealing with sampling
Although not fully accepted in South Africa, the Theory of Sampling issues is to use the principles from the Theory
originally proposed by Pierre Gy is fast becoming the cornerstone of
of Sampling as a template against which to
sampling practice throughout the world. The growing acceptance of
compare different sampling practices. The
Gys Theory of Sampling in South Africa can be attributed to a
number of factors, chief amongst them being the development of a paper is not aimed at identifying either best or
tradable mineral asset market, the promulgation of the Mineral and worst practice, but tries to present an impartial
Petroleum Resources Development Act (MPRDA), the growing view of what sampling practices are prevalent
number of commercial and academic courses that are offered on in the gold, platinum, non-ferrous metals,
sampling, and the regulation of the industry through internationally ferrous metals, and coal industries. The focus
acceptable guidelines and rules for reporting and trading in mineral is on sampling practice in South Africa, and
assets. The size of the South African minerals industry and the where possible, to indicate the levels of
dependence of our economy on mineral production have also meant compliance with the globally accepted
that correct sampling is of key importance to mineral trade. ISO
principles of the Theory of Sampling as laid
standards have been the principal guides for producers of mineral
out by Gy (1979, 1982, 2003). It is clear that
bulk commodities who produce to customers specifications, whereas
Gys insights have been most readily accepted by precious and base
many practitioners, particularly in the bulk
metals producers whose product is sold into metal markets. commodities trade (coal, iron ore, manganese,
Understanding of small-scale variability is essential in the precious chrome, and vanadium), apply the principles
and base-metal industries, but detailed studies of the effects of of sound sampling practice through application
heterogeneity have not been as productive in the bulk commodities. of the ISO standards. These standards uphold
Sampling practices at different stages of mineral development from all the principles of sampling as espoused in
exploration, face sampling and grade control, ore processing and the Theory of Sampling, although they do not
handling, metallurgical sub-sampling, point of sale sampling, and apply the terminology, such as the
sampling in the laboratory are considered in the gold, platinum, Delimitation Error, Extraction Error,
ferrous metal, and coal industries. A summary of the impact of poor
Preparation Error and so forth, as prescribed in
sampling in these industries is presented. Generally it appears that
the Theory of Sampling. As the Theory of
poor sampling practice is most likely to erode mineral asset value at
the early stages of mineral development. The benefits of good Sampling gains acceptance it is more than
sampling are considered, especially with regard to the financial likely that the associated terminology will also
implications of bias and error on large and consistent consignments be incorporated in the international mineral
of bulk commodities. resources and reserves reporting codes and
codes for metallurgical balance reporting.
Keywords
A number of events in the preceding three
theory of sampling, sampling errors, ISO standards.
decades of South African political, economic,
and regulatory history have driven interest,
awareness, and recognition of the importance
of sampling in the minerals industry to the
Introduction
This paper provides a brief review of sampling
issues covered in a more extensive consid-
eration of the state of sampling practice in the * School of Mining Engineering, University of the
South African minerals industry (Minnitt, Witwatersrand, Johannesburg, South Africa.
2010). The term sampling here is taken to The Southern African Institute of Mining and
include all types of sampling in the minerals Metallurgy, 2014. ISSN 2225-6253. This paper
was first presented at the, Sampling and analysis:
industry, from face or chip sampling of solid
Best-practice in African mining Conference,
ore in mining operations, through cross-belt, 46 June 2013, Misty Hills Country Hotel and
cross-stream, or cross-ram sampling of Conference Centre, Cradle of Humankind,
streams of broken ore or slurry, sampling of Muldersdrift, South Africa.

The Journal of The Southern African Institute of Mining and Metallurgy VOLUME 114 JANUARY 2014 63
Sampling in the South African minerals industry
forefront of concerns. Firstly, the unbundling of mineral location, unique numbering of each sample, and measures
assets and the commercial activity in mineral asset markets taken to ensure representivity must be reported. For all
in the early 1990s introduced a heightened awareness of sample types, the nature, quality and appropriateness of the
mineral development valuations all of which are reliant on sample preparation techniques must be indicated, including
sample values. Secondly, growing foreign control of South quality control on all subsampling stages and measures taken
Africas mineral assets also stimulated trade in mineral to ensure that the sampling is representative of the in situ
assets, and with that a necessary focus of attention on material collected. Requirements in regard to subsampling
sampling and related issues. Thirdly, the promulgation of the and non-core samples and their treatment are to be disclosed.
Mineral and Petroleum Resources Development Act (Act 28 of Furthermore:
2002) (MPRDA) provided opportunities for historically A view must be expressed as to whether the sample sizes
disadvantaged entrepreneurs to invest in mining operations. are appropriate for the grain size of the material being
A total of 25 867 applications for prospecting and exploration sampled. The results of selected intersections must be verified
rights were received by the Department of Mineral Resources by both independent and alternative personnel, and the use
(DMR) since 2004 (Department of Mineral Resources, 2009). of twinned holes, deflections or duplicate samples must be
Due diligence studies and re-evaluation of many older reported. With respect to assaying, the laboratory used and
operations meant that sampling took on a new importance. In its accreditation with SANAS must be stated. Laboratories
the mid-1990s interest in sampling was stimulated as a result not yet accredited, should be able to demonstrate that their
of a number of short courses presented by Geoff Lyman, application for accreditation is not older than one year from
Dominique Francois-Bongaron, Francis Pitard, and more the date of publication of the report. Accreditation,
recently by Kim Esbensen. Most of these courses were application for accreditation, and/or rejection of accreditation
presented through the Geostatistical Society of Southern by international standards, such as ISO 9000, must be
Africa in the School of Mining Engineering, at the University disclosed. The nature, quality and appropriateness of the
of the Witwatersrand, although these consultants have also assaying and laboratory procedures, and whether the
presented many in-house courses to private mining technique is considered partial or total, should be explained.
companies. In a number of cases this led to the identification The nature of quality control procedures adopted must be
of individuals who then took on the role of the sampling highlighted (e.g. standards, blanks, duplicates, external
champion for their companies, much to the benefit of laboratory checks) including whether acceptable levels of
company and individual. Finally it is probable that the single accuracy (ie lack of bias) and precision have been
most important event to stimulate the strengthening of established. (Minnitt, 2010, p. 2).
mineral industry-related institutions and with it an interest in Reporting codes have enhanced the importance of
sampling, particularly in countries with vibrant minerals sampling in the mineral industries even though the Theory of
industries, was the Bre-X affair in the late 1990s. This Sampling and Gys formula for the Fundamental Sampling
episode exposed weaknesses in these institutions ability to Error are not mentioned per se. While sampling practice and
rightly guide and control the reporting of mineral resources laboratory performance are consistently emphasised in both
and reserves. This brought the processes of sample taking, the SAMREC code and the JSE Listing Rules, scant attention
quality assurance and quality control (QA/QC) practices, the is paid to the very important aspects of material characteri-
creation of guidelines for fledgling codes, due diligence, and zation, large-scale variability in the plant, and the specifi-
the critical role of the sample audit trail into sharp focus. The cations and performance of sampling equipment, which are
event also galvanized the already growing interest in the main generators of sampling bias. This deficiency is in
sampling and Gys Theory of Sampling in southern Africa need of attention.
and fomented an interest in the sampling-parallel topic of
quality assurance and quality control (QA/QC) (Minnitt,
Economics of South Africas minerals industry
2010).
The checklist and guidelines in the South African Code for The Department of Mineral Resources (DMR) reports that
Reporting of Exploration Results, Mineral Resources and research by Citigroup indicates South Africas non-energy
Mineral Reserves (SAMREC, 2007: Table I) for reporting mineral reserves are worth $2.5 trillion, making her the
relevant information and assessment criteria in public reports worlds richest country (Department of Mineral Resources,
require that Mineral Resource and Mineral Reserve estimates 2009, p. 7). The four most economically important products
be supported by disclosure of sampling methods, sample are gold, platinum, ferrous metals, and coal, and these are
preparation, analyses, quality assurance and quality control, shown with their primary sales values for 2009 in Figure 1.
and the integrity of the database. The range of issues about South Africa accounts for 88 per cent of global platinum
the way sampling is undertaken and reported is wide. They group metals (PGMs) reserves, 80 per cent of manganese, 72
include sampling method, sample collection, validation, per cent of chrome, 32 per cent of vanadium, and 13 per cent
capture and storage, sample preparation, and sample analysis of gold. South Africas exploration budget of R220 million in
(Table I, Sections T3.1 to T3.4), data capture and storage, 2009 placed her 10th in line behind the big spenders,
validation and integrity, QA/QC practices, and policies amounting to about 3 per cent of the global exploration
affecting sample retention (Minnitt, 2010). Downstream, this budget (Department of Mineral Resources, 2009).
affects the way in which listing companies disclose sampling In 2009 South Africas mining industry employed about
techniques along lines set out in Section 12 of the 500 000 people or 2.9 per cent of the countrys economically
Johannesburg Securities Exchange Listing Rules. A active population (Figure 2), and about 1120 jobs were
description of the nature and type of sampling, the precise provided through 112 new small, micro, and medium
64 JANUARY 2014 VOLUME 114 The Journal of The Southern African Institute of Mining and Metallurgy
(Esbensen, 2003). Since then world conferences have been
held every second year, with WCSB6 being planned for 2013
in Lima, Peru. Generally there seems to a separation between
the applications of ISO standards and ToS in the minerals
industry. ISO standards have been applied mainly in regard
to the international trading of bulk commodities, while the
ToS finds its most important applications in the precious and
base metal sectors. The rise and popularity of ToS as a means
Figure 1Value of South African mineral outputs in ZAR billion; note of identifying, categorizing, and minimizing sampling errors
the dominance of gold, platinum, coal, and iron ore has not overshadowed the importance of the ISO-linked
methods and equipment that are embedded in the minerals
industry. The ISO is in a strong position to incorporate and
accommodate the benefits available through the ToS.
The sampling process aims at ensuring that a sample
bears all the representative characteristics of the lot from
which it was taken. Diversions from this goal arise during
increment collection due to the characteristics of the material,
the equipment used for increment extraction, the handling of
increments after collection, and finally the analytical process
itself. The combination of errors from these sources identified
by Gy (1992, 2003) and Pitard, 1993) is termed the Total
Sampling Error (TSE), shown in disaggregated form in
Table I.
Figure 2Total and local mineral sales compared to employment and The sampling protocol along the lot-to-aliquot route is
wages from 1988 to 2009 for the South African industry
generally understood by process operators, but it is rare to
see the total lot-to-aliquot process documented and displayed
in such a way that it can be followed irrespective of the
enterprises (SMME). Apart from the readjustment in 2008,
person involved. Each step in the protocol involves the
Figure 2 also shows a steady increase in both local and
application of tools or technology to particulate materials. The
foreign mineral sales, the latter increasing faster than the
design and application of such sampling equipment must be
former.
implemented such that the principle of sampling correctness
South Africas mineral diversity and abundance is
all fragments must have an equal probability (p) of being in
demonstrated by the fact that 53 different minerals were
the sample is upheld for each volume of observation at
supplied from 1 548 mines and quarries to 80 countries in
each step of the protocol. The incursion of bias due to the
2009. Gold was produced from 53 mines, platinum-group
four errors DE, EE, WE, and PE can be avoided only if p is
metals from 45 mines, non-ferrous metals from numerous
maintained when the lot is presented to the sampling
mines, ferrous mineral ores were produced at 32 mines, coal
equipment. Sampling equipment should not be selective
from 108 mines, and diamonds from 395 mines, all as
either in terms of fragment size, shape, or density, and
primary commodities and as value-added ferroalloys at 23
increment sizes must be proportional to the mass of the lot
metallurgical works during 2009. Since 2003 there has been
presented for sampling. Contamination, loss of mass, or
a steady increase in revenue from mining (Figure 2), but
chemical alteration of material increments during crushing,
revenues from processed minerals in 2009 (R46.8 billion)
milling, screening, drying, filtering, and transportation
were down almost 50 per cent on 2008.
between sample selection events, must be avoided (Pitard,
These economic highlights simply emphasise the fact that
2009).
South Africas economy is deeply reliant on the mining
The defining framework within which the protocol is
industry. Even though it is seventh in line after manufac-
constructed is referred to as the sampling nomogram. This is
turing, electricity, financial services, tourism, communi-
a staggered series of comminution and mass reduction
cations, and construction as a contributor to national GDP,
processes applied to the broken ore and designed so that
mining is still the economys largest employer (Figure 2).
changes in the variance of the FSE with changes in fragment
Thus, despite a general lack of appreciation of its relevance
size and increment mass can be followed. The nomogram is
and importance, sampling in the mining industry plays a
constructed so that the FSE should not exceed 10 per cent
major role in establishing mineral resource value and we do
precision at any point along it. The two constants K and
well to pay attention to it. Since sample values form the basis
alpha () are determined through an experimental process of
for many major financial decisions, the manifestation of
calibration for each specific ore type (Francois-Bongaron
sampling errors and bias can have considerable implications.
and Gy, 2002a, 2002b). These values are then substituted
into a modified version of Gys formula and form the basis
Theory of Sampling and the ISO standards for compiling a nomogram.
Pierre Gys Theory of Sampling (ToS) (Gy, 1979, 1982, 1983, The importance of the International Organization for
2003) was the central subject at the First World Conference of Standardization (ISO) standards in providing stable
Sampling and Blending (WCSB1) held in Denmark in 2003, guidelines for sampling prior to the widespread dissemination
entitled 50 years of Pierre Gys Theory of Sampling of the ToS is referred to by Minnitt (2010). This organization

Table I
Taxonomy of sampling errors in regard to material characteristics, plant processes, sampling equipment, and
analytical procedures
Sources of error Nature of error
Total Sampling Material characterization In situ Nugget Effect (INE), A function of the occurrence and distribution of
Error (sampling uncertainties) nuggets in the host rock to be sampled
Fundamental Sampling Error (FSE). Related to compositional heterogeneity
due to the material properties of the lot
Grouping and Segregation Error (GSE). Related to distributional heterogeneity
due to grouping and segregation in the lot
Sampling equipment and Increment Delimitation Error (IDE or DE), Geometry of outlined increment is not
materials handling completely recovered; can be completely eliminated
(sampling errors) Increment Extraction Error (IEE or EE), Material extracted does not coincide
with the delineated increment; can be completely eliminated
Increment Preparation Error (IPE or PE), All sources of non-stochastic variation
after extraction of the material; error should always be nil
Increment Weighing Error (IWE or WE), May be relevant or not, depending on specific context
Plant process and procedures Short-range Process Integration Error (PIE1)
(combined sampling Long-range Process Integration Error (PIE2)
uncertainties and errors) Periodic Process Integration Error (PIE3). Random, time, and cyclical fluctuations,
usually sub-divided into random and non-random effects
Analytical processes Analytical Error (AE), All sources of error associated with materials handling and
(mainly sampling errors) processes in the laboratory
Source: Gy (1992, 2003), Pitard (1993, 2009)
was established in 1947, at about the time that Pierre Gy at the port of destination. The ISO provides a valuable
began to develop the underlying concepts and ideas for the standard and basis against which international trading
ToS. Although not restricted to the minerals industry, the ISO partners can measure product specifications.
standards represent minimum requirements for product
quality and provide definitions, terminology, and concepts for Sampling at South Africas mineral deposits
effective quality management in the trading of mineral
The economic significance of gold, platinum, ferrous metals,
commodities and products. According to Cook (2010) the
and coal (Figure 1) to South Africas export earnings forms
main aim for the ISO is to facilitate trade by institutionalizing
the basis on which the sampling practices in the industry
attitudes towards quality that are based on right policy,
were investigated. For each of the four principal mineral
procedures, records, technologies, resources, and structures.
products the sampling procedures and processes during
The benefits to ISO-registered suppliers are product differen-
exploration, face sampling and grade control, ore processing
tiation through superior quality products and a measure of
and handling, and metallurgical subsampling of solids and
customer loyalty because their expectations in regard to
slurries are considered in tabular form in the following
quality are being considered. The ISO provides carefully
sections. The influence and effect of the two main sampling
researched off-the-shelf standards for sampling-related
errors, Fundamental Sampling Error (FSE) and the Grouping
issues in mining and quarrying (73.020), coals (73.040),
and Segregation Error (GSE), as well as the biases that arise
metalliferous minerals and their concentrates (73.060), non-
due to the Delimitation, Extraction, Preparation, Weighting,
metalliferous minerals (73.080), and equipment for
and Analytical errors, are noted. The possible magnitude of
processing of minerals (73.120) (Minnitt, 2010).
each error or bias for the mineral product at each stage of
The application of ISO standards also appears to rest on
sampling is estimated on a scale of one to ten and the
an important distinction between precious/base metals and
sampling practices are graded in this way. While the impact
bulk commodities. Generally, precious and base metals are
of laboratory practices is shown in the tables, the issues of
sold into metal markets where prices are fixed through a
sampling in the laboratories for these minerals are dealt with
number of different methods, but all are related to the
elsewhere in this paper.
balance between supply and demand. In addition, because
precious and base metals can be produced to a five-nines
standard, there is very little in the way of product differen- Sampling gold deposits
tiation in the market place. For the bulk commodities, For more than 100 years the gold mining industry has
however, once a supplier-customer relationship has been applied itself to a variety of aspects of sampling, particularly
established, linkages between producer and consumer are face sampling, which are controlled by strict, documented
likely to be much stronger. The producers of ferrous metals, protocols for sampling at all stages of gold production. The
for example, produce to meet fairly inflexible customer methods and reasons for good sampling, including health and
specifications that are established before the product arrives safety measures are provided in detailed standard operating
procedures (SOPs), or codes of practice, for example Internal and Sichel 1960; Whitten, 1966; Rowland, 1966; Storrar,
underground chip sample standards (Flitton, Bernard, and 1981; Lerm, 1994) and are documented by Cawood (2004).
Spangenberg, 2009). Monitoring of sampling sites, Although these insights were not ordered in a structured
underground and on surface, during regular planned task taxonomy like that in the ToS, Cawoods (2004) study of the
observations (PTOs) by company sampling champions historical development of sampling practice on South African
ensures that protocols, standards, and SOPs are adhered to. mines suggests that a complete spectrum of sampling errors
Most large mines have information management systems that equivalent to those that Gy identified were appreciated and
verify data through checks and approvals (Minnitt, 2010). accounted for in the principles underlying sampling practice.
Details of sampling in gold mining operations are The combination of error and bias is generally not as
summarized in Table II. severe for the wider more homogenous reefs such as the
Ventersdorp Contact Reef (VCR) and Main Reef (Brokken,
Face sampling 2010) as it is for the narrow composite reefs (Carbon Leader,
The run-of-mine gold content and day-to-day economic Basal Reef, Vaal Reef). The latter reef types consist of a small
viability of South African gold mining operations is pebble conglomerate 10-20 cm wide, with a carboniferous-
maintained through a grade control process that relies on rich layer, between 0.1 cm and 2 cm wide, at the base of the
chip sampling of mineralized rock faces. Training for the reef. The carbon-rich layer may contain fly-speck or
practice and procedures for channel sampling of the stope columnar carbon that makes this narrow unit much softer
faces for samplers and surveyors is provided as part of the than the overlying quartz pebble conglomerates. The softer
Mine Surveyors Certificate of Competency (Department of carbon-rich layer may contain up to 90 per cent of the gold
Minerals and Energy), through technical colleges, the mineralization and easily produces a 20 per cent
Chamber of Mines, and mining companies. oversampling, something that Sichel (1947) and later
The sample mass, sample spacing, sample size, and the Harrison (1952) warned of. Even under ideal circumstances
ratio of sample mass to the gross mass as the critical factors channel samples are defined only in two dimensions and
determining the representivity of samples were described by owing to the hardness of the rock the most carefully
Storrar (1981). Various aspects of face sampling of the delimited samples can never be fully extracted. Sample
Witwatersrand-type gold reefs has been investigated by extraction using hammer and chisel is not acceptable, and
many (Watermeyer and Hoffenberg, 1932; Hoover, 1933; generally the sampling process is dictated by what is possible
Beringer, 1938; Jackson, 1946; Truscott, 1947; Sichel, 1947a, in practice rather than the requirements of a prescribed
1947b; Harrison, 1952; Krige, 1952, 1964, 1966; Rowland sample mass. The common use of accumulation values
Table II
Sampling in the gold mining industry
Sampling site and mechanism, sampling problems arising from site and mechanism Error/ Impact
Exploration
Very detailed protocols for dealing with borehole cores. Use of diamond drill holes for exploration. DE, EE/ High (7)
Random systematic or stratified sampling methods.
Face sampling and in-mine grade control
Authoritative sampling mode. Very well developed face sampling protocols. In situ sampling has the greatest potential to erode FSE, DE, EE/
value. Life-of-mine plans are based on in situ sampling. Any bias at this stage can seriously affects long-term forecasting and life Extremely high
of mine plan. Large pebble conglomerate reefs (VCR) are easier to sample low mine call facror (MCF). VCR generally has a depending on
higher MCF that thin carbon base reefs. Some reefs are thin, composite conglomerate-carbon reefs (Basal, Vaal, Carbon Leader). mining method (9)
Each component has very different sampling characteristics. Calculation of tonnages could be affected
Ore processing and handling
Larger operations have clearly defined sampling protocols for broken ore sampling, (BOS). Sample mass and number of FSE, GSE/ Med
increments per unit mass clearly defined at go-belt sampler. Large composite samples are collected. Assay aliquots the same for high (6)
BOS and plant samples. BOS samples are reduced in size in preparation plant to produce assay aliquots. Optimized protocol for
stages of BOS sub-samples. Increment size and rate of collection carefully controlled by protocol.
Handling procedure of broken ores causes softer materials to degrade and break up causing liberation and segregation of fine
gold. Composite sample collected from cross-belt sampler (go-belt sampler). Material is very coarse; up to 30cm. Segregation
and loss of fine gold during transport leads to low MCF. Coarse material extracted, fines left on belt.
Metallurgical sub sampling solids and slurries
Vezin-type cross-stream samplers and in-line, automatic cross-stream samplers: EE/ Low (2)
Process sub-sampling solids and slurries
Detailed protocols for collecting, handling and analysing process control samples PE/ Med (4)
Sub-sampling in the Laboratory
Face samples are pulverized and assayed. Data used for grade control and evaluation for short- and long-term forecasting. FSE/ med-high (5)
Different protocols for handling BOS and face samples in laboratory. Face samples use 30g aliquots, BOS samples use larger
aliquots.Treatment and sub-sampling of broken ore samples (go-belt) more detailed and controlled by detailed protocols. BOS
samples are larger and assay procedures more carefully controlled than for normal face sampling protocol. Data used for month-
to- month mining reconciliation, metallurgical balances and metal accounting. Detailed optimized sampling protocols are
designed and implemented.
Total 31/60 =50%
(Source: Minnitt 2010, used with permission of AusIMM, 2013; See Table 1 for an explanation of FSE, GSE, DE, EE, PE, WE, and AE)

(centimetre-grams per ton, cmg/t) means that different representative composite sample should be determined from a
densities (e.g. Ventersdorp Contact Reef uses 2.71 g/cm3 heterogeneity study. Information from go-belt sampling is
whereas Carbon Leader and C Reef use 2,78 g/cm3) are valuable if the results are used in a fit-for-purpose manner.
applied to a reef width, rather than sample mass (Cawood, Unfortunately, the inability to reconcile the geostatistically
2004). evaluated call on gold from underground face sampling with
Overstatement of sampling values is the principal what is found on the go-belt has meant that go-belt samplers
contributing factor to apparent loss and low mine call factors have been decommissioned at several mining operations.
on gold mines where narrow carbon-type reefs are mined. However, the experience at Harmony Gold Mining Company
Attempts to use single- or multiple-blade diamond saws have is that RoM sampling can be used for monthly metal
made little difference to the overstatement of sampling accounting to allocate production to different shafts, despite
values, and nondestructive radiation analysis in the stope the high variance of go-belt sampling.
face has not gained much acceptance (Lerm, 1994). Despite Moisture may affect the sample mass leading to a bias in
some small advances over the years, it is now nearly sixty the sampling results, particularly if the proportion of fines is
years on and there is still no mechanized tool that can take high (Pitard, 2009, personal communication), but moisture is
the perfect sample at an even depth efficiently and econom- usually measured on samples taken specifically for the task
ically (Cawood, 2004). Irrespective of how well samplers are rather than using material that is collected for assay
trained, face sampling differs from one sampler to the next; purposes.
even under the strictest supervision, uniform sample
extraction is almost impossible (Lerm, 1994, Flitton, 2010, Waste stream sampling
Minnitt, 2010). Material sent to the waste dump is analysed on a regular
basis using a stop-belt sampler consisting of a static steel
Go-belt sampling frame that exactly matches the profile of the stopped
Although no studies have been reported, it is likely that a conveyor belt. The frame is placed on a stopped belt, allowing
good proportion of the softer, carbon-rich gold ores are the total increment to be easily removed from the belt
broken during the seven-fold handling from underground together with all the fine material that accumulates at the
workings to the plant. Much of the contained gold will be bottom of the belt; the ideal situation is shown in Figure 4.
liberated and either segregated or lost to the system. On The nominal top size of the waste rock is 150 mm and the
surface the ore is fed from the tip to the go-belt, sampled width of the cut is 50 cm, at least three times greater than the
using a cross-belt hammer sampler (Figure 4), and then nominal top size. The opening through which the sample is
stored in a silo from where it is fed to the plant. The ore skips delivered and collection chute are designed to minimize any
are equipped with weightometers that provide a record of spillage or loss of sample (Spangenberg, 2009).
tons hoisted. In terms of sizing, the first constraint occurs as
the ore is passed through the 30 cm x 30 cm grizzly, at the tip
in the underground workings or on surface, before reaching
the go-belt sampler.
Typically the go-belt sampler is the first point at which
both grade and tons are estimated in the reduction of gold
ores. Some gold mining companies characterize their ores by
means of heterogeneity tests, detailed deportment studies,
and calibration of the sampling constants (K and ). Such
studies provide the basic information for compiling an
optimal sampling protocol that describes the changes in
sampling variance as the mass of the samples is reduced as
recorded in the sampling nomogram. Studies indicate that a
composite sample collected at the go-belt should be 200 t to Figure 3Go-belt sampler installation on conveyor belts from shaft
head (source: Letsie, 2009; Minnitt, 2013)
500 t (Letsie, 2009). Variographic analysis of ore streams is
an important step in minimizing large-scale cyclical
variability due to process variations in the plant. Such
insights from the daily go-belt samples provide guidance for
the optimal size and timing of increment extraction, e.g. one
70 kg sample per 500 t of ore.
Go-belt samples provide an estimate of the total gold
produced from each reef every month. The go-belt sampler is
screened off to prevent either injury or interference with the
collection of the ore (see Figure 3). The two main factors
affecting representative sampling at the go-belt are sampling
frequency and sample mass (Spangenberg, 2007). At some
operations go-belt samplers are activated on a mass basis by
a weightometer, such that a single 200 t increment is Figures 4(a) Design and shape of the stop-belt sampler is correct, (b)
collected over a 24 hour period from 10 increments of about profile of the stop-belt sampler fits the shape of the belt exactly
20 kg each (Kelly, 2007). Generally, the increment mass for a (source: Kelly, 2007)
Plant sampling acid wash phase pH and Ca are determined before and after
the wash and during loaded carbon lifting. Loaded, barren,
Sampling in the plant is used to ensure that in-plant
and regenerated carbons are sampled by in-line carbon
processes as well as flow rates, pulp densities, and pH
samplers during the transfer from the acid wash to the
measurements are maintained within specifications. In
elution column, and during transfer from the elution column
contrast, sampling for recoveries and efficiencies of
to the kiln feed bin. Solutions from the electrowinning feed
metallurgical balances ensures that gold entering and leaving
and the barren electrolyte are sampled manually by an
the plant is accounted for.
operator at the electrolyte drain valve. Smelthouse bullion
In gold plants, pulp head feed to the thickeners (thickener
and slag are manually sampled by drilling and increment
overflow) is sampled using robust Vezin-type cross-stream
collection by an operator. Waste products, including carbon,
samplers to produce a daily composite made up of sample
wood chips, and solids, are also sampled manually as
increments collected every 15 minutes. Radial cutter blades
required.
on Vezin samplers (Figure 5a) are radial along the full length
of the collector, but may accumulate a residue of pulp on the
blades immediately adjacent to the portion that cuts the Sampling platinum group metals
slurry stream. In some cases adequate and effective cleaning The 2 060 millionyear-old Bushveld Complex contains 68
mechanisms have been installed as shown in Figure 5b. per cent of the world's economically viable chromite reserves
Turbulent flow of slurry along the edges of the launder at and 86% of all known platinum group metals. The Critical
the thickener underflow (Figure 6a) may generate a bias as Zone hosts the Upper Group (UG) chromitite layers, among
the cutter blades cross the stream at this point. Cross-stream them the UG2 Chromitite Layer, which averages 1 m in
slurry samplers may be installed where the flow rate in the thickness, carries economically important platinum group
launder is too high so that unbiased, representative samples element (PGE) deposits. The uppermost unit of the Critical
cannot be collected. While the cutter blades for sampling Zone carries the PGE-bearing Merensky Reef, a feldspathic
device illustrated are not worn, the length of the blade is pyroxenitic assemblage with associated thin chromitite layers
inadequate for the flow rate (Figure 6b, Kelly 2007). that rarely exceeds 1 m in thickness. Apart from faults and
Residue pulp samples collected in the plant (final tail dykes the Merensky Reef and UG2 Chromitite Layer are
sampler) are sent to the filter press and submitted for characterized by potholes slightly sunken, circular features
analysis. Training in the simple aspects of sample correctness where lithological layers of economic significance are
and good sampling practice is essential if staff across the completely or partially absent, causing geological losses of up
plant are to be sensitized to the importance of sampling. In to 25 per cent. Chromitites of the UG2 Chromitite Layer
any responsible position, untrained staff could potentially typically consist of fine-grained (1-2 mm) chromite (>80 per
generate biases during sample preparation because of a poor cent) and interstitial feldspar (<20 per cent ), and accessory
understanding of the sampling processes involved. interstitial pyrrhotite and chalcopyrite. The platinum group
Several samples are taken at different points in the minerals (PGEs) cooperite, braggite, sperrylite, and laurite are
elution stream, most of which are extracted manually. In the typically very fine-grained, and hosted within the base metal
sulphides (Lomberg, 2009, Minnitt, 2010). A summary of the
sampling issues in the platinum mining sector identified in
this study is provided in Table III.
Exploration
Exploration borehole cores are treated with particular care in
that they are the primary indicators to prospective investors
of the value of the potential mineral development (Donlon,
2010). Borehole cores are split using a diamond saw and
half-core samples 1520 cm long are taken within the UG2
Figure 5(a) Vezin sampler splits the primary residue sample, and Chromitite Layer, with slighter longer samples (up to 30 cm)
(b) cleaning nozzle pointing at head grade sampling by cross-stream in the immediate hangingwall and footwall. Samples include
cutter (source: Kelly, 2007) approximately 2 cm of hangingwall or footwall, the same as
in the gold mining industry. Samples are placed in a
numbered bag, documented, and dispatched to the laboratory
for analysis. The remaining split core is marked showing the
relevant sample numbers is stored for future reference
(Lomberg, 2009). Core samples, between 500 g and 600 g,
are crushed to 70% -2 mm and milled to 85% -75 m, and
30 g and 50 g aliquots assayed by fire assay using lead
collection. PGE fire assays are undertaken locally, but Cu and
Ni assays (XRF) are undertaken abroad. QA/QC controls
ensure that PGE and base metal assay techniques are
Figure 6(a) Turbulence created at the lip (LHS of launder) will appropriate for mineral resource estimates, and that the assay
introduce bias at this point, (b) cutter blades are not long enough to techniques for recent drilling are compatible with the
match the flow rate (source: Kelly, 2007) techniques used for the historic data.

Table III
Sampling in the platinum mining industry
Sampling site and mechanism/ Sampling problems arising from site and mechanism Error/impact
Exploration
Very detailed protocol for siting, drilling, and recovering core samples. Random systematic or stratified sampling methods. DE, EE/ high (6)
Only mineralised portions of core are split, bagged and tagged. Samples are then weighed, crushed and LM2 milled to 80%
-75 m. Samples are analysed for density, base metals and PGEs.
Authoritative sampling mode. Sample cut from block in face. Data used for evaluation and long-term planning and DE, EE/ very high (8)
comparisons with plant input measurements. Merensky and UG2 reefs significantly different. Distribution of grades varies
through the reef profile for both Merensky and UG2 reefs. Higher grades may occur near footwall contact, distribution is
inconsistent. Most serious impact on long-term planning and life-of-mine.
Ore from various shafts are not batch processed, but mixed at random when delivered at the concentrator. Ore is GSE/ medium (5)
transported from shafts to the plant via conveyor belts, trucks or trains. Handling causes degradation with segregation and
loss of fine sulphides. Mixing ore and waste underground increases the material heterogeneity. Composite samples are
collected with cross-belt sampler (go-belt/hammer sampler). Bias due to incorrect sample extraction. Varying belt profiles
have a detrimental effect. Due to sample size it is very difficult to determine representative moisture, where dry mass may
be required. For run-of-mine material (top size approx. 300 mm) where material is very heterogeneous or mixed with waste up to
50 t of sample is required to minimise sampling constant. Minimum of 30 cuts per sampling campaign is required. Moisture
content of sample may be incorrect. Grade content of sample used only to calculate ounces split ratios, to allocate
concentrator product content back to shafts.
Metallurgical sub sampling solids and slurries
Batch transfers at smelters and refineries can be controlled via a semi-automated sampling system. The system software DE, EE, PE/ medium (5)
can communicate with the PLC/weighbridge software etc. to do system checks and facilitate the use of barcode scanners.
Transfers and the associated samples can thus be traced via barcodes and electronic transaction identities creating a proper
audit trail and simplifying processes.
Moisture sample to determine dry mass.
Sampling for grade: at the concentrators a Cross-stream/Vezin or Vezin/Vezin sampler at the mill discharge/rougher feed
slurry line is used. At the smelters for concentrate wet cake material Auger sampling in tipper trucks/or Spiral samplers. At
the refineries interleaved Vezin samplers can be used.
Process sub-sampling solids and slurries
Instrumentation for online analysis can be used. Only measuring part of the stream. Data may be biased and can only be GSE, DE, EE/ med-high
used for trending. Mass and density meters. Requires a full upward flow. Difficult to calibrate representatively. Error in (6)
measurement decrease as density decrease. Recovery decreases as density increases
Sub-sampling and analysis in the laboratory
Semi-automated sample preparation; samples can be prepared and analysed at robotic laboratory. Samples are labelled with GSE, DE, EE/ med-low (4)
bar codes. Robotic laboratory then crush and mill samples to 80% -75 m before analysis. The system software can
communicate with the PLC/weighbridge software etc. to do system checks and facilitate the use of barcode scanners. The
samples can be traced via barcodes and electronic transaction identities, creating a proper audit trail and simplifying
processes. Sample preparation protocols are designed to minimize preparation uncertainties. Milled or finely
crushed products are normally sampled, but where possible the sampling constant is determined, to further define the
nominal top size before sample size reduction. When weighted composite samples are required, Variable split sample
dividers (VSSD) are now replacing matte rolling and dip sampling.
Generic sample preparation protocol includes: drying in conventional oven at 110C, lump breaking with lump breaker or GSE, DE, EE/ low (2)
screen, rotary split to reduce sample size and collect representative sample for gradings (if required), use of VSSD to make up
daily weighted composite with an ideal mass, LM2 milling of individual samples or weighted composite to 95% passing 75 m,
rotary split composite sample to provide sample for laboratory (normally between 200 g and 500 g). If sample has not yet
been milled to 95% passing 75 m then Sieb mill sample. Selection of the final aliquot can be problematic especially e.g. if only
example 0.5g is needed for base metal analysis. Current practises involve matte rolling and dip sampling but the use of 3
dimensional tubular mixers and optimized micro-splitters are being investigated.
Total 36/70=51%
(Source: Minnitt 2010, used with permission of AusIMM, 2013)
A recent comparison of chip sampling against reef sampling intervals of 26 m. Sample positions are carefully
intersections in boreholes indicate that the chip samples marked off, using a chalk marker, as a channel perpendicular
return higher grades than the borehole intersections by 35 to the dip of the reef and across the full width of the reef
per cent on UG2 and 3040 per cent on Merensky Reef (Figure 7a). The channel width of the sample is 20 cm, and
(Mosholi, 2009). the full width of the reef, hangingwall to footwall, is
subdivided into 20 cm intervals. Sample borders are cut into
the face along two vertical guide lines using a circular rock
Face sampling saw to a depth of approximately 4 cm; samples are then cut
The similarity in dip, thickness, and stratiform nature of the along horizontal lines (Figure 7b). A hammer and chisel are
reefs means that face sampling protocols for the platinum then used to chip, from the bottom upwards, about 500 g of
mines and Wits-type gold mines are similar. Underground rock from the face with the rock fragments being collected in
chip sampling is undertaken for evaluation purposes on all a metal tray and transferred to a numbered sample bag
stope faces and reef development ends (winzes and raises) at (Figure 7c, Hol, 2009).
performance of UG2 ores. UG2 flotation concentrates still
contain entrained chromite; this in turn affects the
metallurgical behaviour of UG2 concentrates in the
downstream smelting process (Donlon, 2010).
Sampling departments at the larger mining operations
have well-designed and rigorously enforced codes of
Figure 7(a) Marked off sample section, (b) cutting sample from stope sampling practice. Larger mining operations treat Merensky
face, and (c) chipping samples from stope face (source: Hol, 2009) and UG2 reefs separately using conveyors to deliver
Merensky and UG2 ore to their respective mills; smaller
operations simply process a combined stream of ore. Milled
Sampling standards are maintained by careful training products are sampled using a cross-belt hammer sampler to
and regular auditing of samplers, because the likelihood of provide management with shaft-specific indicators of
contamination and loss of material is high (Hol, 2009). metallurgical balances and the effectiveness of grade control.
Although the sampling is made easier using diamond saws to Sampling of the UG2 and Merensky reefs are considered
cut along the correctly delimited sample boundaries on the separately because of the large amount of by-product
exposed rock face, traditional sampling using a hammer and chromite associated with the UG2 reef. Types of samplers
chisel has been a preferred method since the opening of encountered in the plant at UG2 operations include
South Africas precious metal mines in the late 1880s. automated cross-belt samplers (Figure 8a), as well as manual
sampling of slurries, falling streams (chromite concentrate),
Go-belt sampling and tailings. Typically the plant is designed to separate PGEs
The first sampling of ore on surface is by means of an and chromite in UG2 ores by crushing and cleaning with
automated cross-belt (hammer) sampler that is activated by a heavy media separation. Particle sizing of the crushed ore is
weightometer on a mass basis (Figure 8a and 8b). Although checked every 4 hours before it is milled; ore is acceptable if
standard cross-belt cutters may meet all the necessary 75 per cent of a 3 kg sample passes -19 mm. Head grade and
engineering specifications, one must accept that non-biased moisture are measured on an eight-hourly, 1215 kg
sample extraction is not possible with this device because the composite of increments taken by an automated cross-belt
cross-belt cutter does not sweep close enough to the belt to sampler.
ensure that all the fines are removed (Figure 8c). At some mines spirals are used to separate the chromite
The fines invariably carry higher grades than the larger and to produce a saleable metallurgical product. After primary
lumps of ore and the extraction of larger fragments in crushing, ores being conveyed to the silos are sampled on an
preference to smaller ones introduces a bias in the grade of hourly basis using stop-belt methods. Increments are
the samples. Poor adjustment, sub-standard installation, or accumulated over an eight-hour shift into a 1015 kg
simple neglect may mean that the already-biased samples composite sample that is assayed for head grade and
incur additional bias (Mosimane, 2007). Time-based moisture. Chromitite ores pass through a primary mill
increment extraction has the disadvantage that large followed by primary floatation that removes PGE-bearing
amounts or very little ore may be on the belt at the instant sulphides. Chromite is removed and the material flow is
the cutter is activated, and the efficiency and effectiveness of subject to a secondary crushing followed by secondary
the device is questionable. The mass of some increments may floatation for PGE-bearing sulphides.
be too small, giving rise to biased and inaccurate samples. Milled Merensky ore is sent to a primary flotation plant
where the ore is slurried with water and reagents. Once the
Plant sampling waste material is removed, the floatation concentrate and the
Detailed mineralogical studies together with material charac- sinks (tailings) are sampled using a time-based cross-stream
terization of the ores and precious metal assays provide the cutter and a continuous Vezin sampler. Installation of the
database for the compilation of sampling nomograms that equipment is done by the mine according to the manuals
allow one to minimize the FSE. Many such studies are provided by the manufacturers, who will provide equipment
conducted, as recovery from both the Merensky and UG2 guarantees only if they perform the inspection phase of the
reefs depend critically on comminution and liberation of the installation; training and manuals for the use and
host sulphide minerals. Comminution protocols and sample maintenance of the equipment are provided by manufac-
preparation are very sensitive to grain size because the turers.
sulphides containing the PGEs are interstitial to the grains of
the chromitite host. The UG2 Chromitite Layer breaks along
grain boundaries and liberates the sulphides, which means
that fines on a UG2 mine typically have twice or three times
the in situ grade (Lomberg, 2007). The liberation of the
precious metals is a function of the liberation of host
sulphides, which in turn is a function of their grain size and
the chromitite-sulphide grain structure. Samples with lower
grades are more sensitive to the comminution protocol; this is
particularly true of lower grade gold mineralization. Figure 8(a) Go-belt sampler for Merensky reef, (b) for UG2 reef, and
Recoveries from the UG2 are lower than for the Merensky (c) swath cut by the sampler is evident, but fine ore remains on the belt
Reef because of the poorer metallurgical flotation (source: Mosimane, 2007)

In summary, possible biases have been reduced by current density measurement process (i.e. consistency of
considering the feed size, the feed rate, mill capacity, SAG sample mass, possible dilution and contamination) can be
mill loading size, and the product size in the design phase of avoided.
comminution circuits. In the design of flotation circuits the Over the last few years the largest of South Africas
potential for sampling biases in slurries has been reduced by platinum producers has improved its evaluation processes
considering feed rates and grade, froth flows and grade, and through implementing strict corporate governance and Best
the percentage solids. Spot checks for obvious defects, Sampling Practice. The Home and Away Sampling strategy
contamination, broken parts, and sample masses are allowed samples to be analysed at the Sender and Receiver
performed at random on a daily basis by senior staff sites, and was convenient due to the geographic distribution
members who record their findings on a checklist. of plants in the eastern Bushveld Complex. However as the
operations expanded, maintaining sampling quality and
Summary for platinum deposits sample throughput was threatened, so that single-point
Consistent training and exposure to courses given by leading sampling and analysis in an ISO17025 accredited laboratory
international experts over the past 15 years is bearing fruit as was introduced. This facility was then upgraded knowing
practitioners and manufacturers incorporate concepts from that the best defence against sampling bias is correctly
the ToS into plant designs and custom-built sampling designed mechanical sampling.
equipment used in both Merensky and UG2 plants. To Sampling protocols that define stages of comminution
consolidate the advances in equipment manufacture, further and mass reduction, minimising and eliminating dust losses,
process optimization in platinum plants requires sampling and monitoring of quality control parameters, for different
campaigns, mineralogical analysis, fractional analysis, data material streams have been introduced. Improved QA/QC for
analysis, and feedback. However a lack of management buy- sample preparation procedures, improved audit trails, and
in, inappropriate flotation cell sizes, unstable feed character- electronic transfer of data became essential as the number of
istics, poor plant control, and skills shortage act as samples to be processed has increased. Electronic
constraints on such developments (Naidoo, 2010). management of sampling protocols, sample bar coding, audit
The cross-belt, rather than cross stream, samplers are trails, process variation, and statistical process control are
generally used on the run-of-mine ores before or after the now standard procedures. New mechanical equipment (the
primary crush. Such samplers are easy to retrofit, cheap to Variable Split Sample Divider) has been successfully designed
manufacture, and need no transfer point as for cross-stream and developed to replace manual methods for compositing
sampling (Naidoo, 2010). Cross-belt samplers are adjusted to weighted samples. Optimisation of the mass determination as
minimize the chance of damage to the conveyor belt while well as correct sampling of concentrate slurries has reduced
taking the best possible sample. Segregation of coarse and operational cost significantly
fine material occurs on conveyor belts during transportation The likelihood of serious introduction of bias in the
of material over a large number of idlers before sampling analytical stages of sampling has been minimised by the
(Robinson and Cleary, 2009). Consequently there tends to be installation of completely robotic systems that eliminate
an over-representation of coarser material from the top of the human handling and intervention from the time the sample is
belt and the side of the belt nearest the sample chute, with a submitted to the time the grade appears as a figure on the
substantial amount of finer material left on the belt. Such report document. (Minnitt, 2010).
samplers are not recommended (Lyman et al., 2009). Rotary
and Vezin-type samplers generally have good extraction Sampling ferrous metals
ratios and negligible bias under a range of operating
conditions (Robinson and Cleary 2009) so their use at the
Manganese deposits
plant is recommended.
The strongly biased nature of grab samples, as well as the The Kalahari Manganese Field in the Northern Cape Province
ease with which a cross-stream sampler can be fitted at a hosts the stratiform manganese beds that occur in the banded
transfer point on the conveyor from underground to stockpile, ironstone sediments of the Hotazel Formation in the
should preclude the use of grab sampling, especially to lowermost Postmasburg Group at the base of the Transvaal
determine the head grade of feed to the plant (Lyman et al., Supergroup. Of the three sub-horizontal manganese beds,
2009). These investigators also suggest that the stop-belt only the lowermost (Mn1) is currently mined at Nchwaning
sampling creates undue strain on the belt, causing Mine. However, numerous small mining operations have
overloading of the belt, spillages, and overflow into the entered the market and are extracting ores from other units.
sample collected (Naidoo, 2010). In the same way the use of Experience has shown that of the 6 m thick beds, only a more
spigots and pipes to manually sample slurry streams in the or less central 3.5 m mining cut in the lowermost unit
chrome and tailings re-treatment plant should be replaced by provides the optimal grade and recovery. The ISO 9000
a sampling mechanism that will introduce turbulent flow. standards for manganese sampling are strictly applied during
This will break up the segregation that develops during the stacking and reclamation of numbered, sized, and
laminar flow and allow a more representative sample to be sampled stockpiles that are about 300 t (Figure 9).
taken (Robinson and Cleary, 2009). According to Naidoo The ores being stockpiled are sampled using automatic
(2010), test work has indicated that in the flotation cells, cross-stream samplers to collect mass-based increments.
recovery decreases with higher density in the primary Inspection of the cross-stream samplers indicated that they
thickener underflow (Valenta, 2009). It is therefore important obey the principles of correct sampling as per CSIRO (2000)
to keep a good check on densities, so the pitfalls in the requirements. Each sample of a stockpile consists of 10
increments of between 150 kg and 170 kg to give a every 30x30 cm block being about 200300 g. This means
composite sample of between 1.5 t and 1.7 t per stockpile that every mining cut (about 1300 t of ore) is covered by
(stack); the sampling rate here is 0.64 per cent. This allows 1215 samples with an average mass of 5 kg.
extremely accurate grade control for all shipments from the
mine. The average size distribution for the manganese ores is
shown in Table IV. Products are typically split into lumpy (9 Table IV
75 mm) and fines (<9 mm) (Assmang, 2010).
Approximate size distribution of classified
Typical analyses of the fines and lumpy manganese ores
are listed in Table V. manganese ores
A summary of the main issues in regard to sampling in
Size fraction %
the ferrous metals industries, taken here to be represented by
iron and manganese, is presented in Table VI. -75 to -60mm 12%
-60 to -32mm 27%
All exploration holes, 36mm BQ core, are drilled vertically -32 to -6mm 32%
to intersect the three sub-horizontal manganese layers. Core Fines: > -6mm 19%
from all three seams is split, logged, and sampled at 50 cm Fines: > -0.5mm 3%
Slime: 7%
intervals so that the best 3.5 m composite section can be Total 100%
identified. In-mine grade control is done by profiling of the
orebody for manganese and iron by marking off samples in a
30 cm wide vertical channel across the face. The channel is
divided into 30 x 30 cm blocks, the sample mass collected for Table V
Typical analyses of manganese ore grades
Element, % 47% fines 48% lumpy
Mn 48 49
Fe 11.4 10.5
SiO2 6 4.5/6.0
P 0.05 0.05
S 0.19 0.17
AI2O3 0.39 0.35
CaO 5 4.9
MgO 0.7 0.7
K2O <0.1 <0.1
Na2O <0.1 <0.1
Figure 9Numbered and sampled manganese stockpiles ready for

blending Source: Assmang 2010.
Table VI
Sampling in the ferrous metals industry
Sampling site and mechanism. Sampling problems arising from site and mechanism Error/impact
Exploration and blast hole sampling

Very detailed protocol covering drilling method preference, material recovery, core drilling, percussion drilling, reverse DE, EE/ High
circulation (RC) drilling, transport of drilling samples, sampling, sample preparation, analyses, chain of custody,
administration, and metallurgical test work.
Detailed protocol for in-pit mine-grade control. Random systematic or stratified sampling methods used for blast-hole and DE, EE/ Very High
reverse circulation drilling. Sample defined in 2D, Sample mass depends on depth drilled, every one or two metres. Sampling
of chips from blastholes using pie-slice catcher. Labour intensive, operator dependent. Data used for evaluation, planning
product specifications, and life-of-mine planning. Bias due to differences in sample support, incorrect delimitation and substandard
extraction of sample chips. Bias can influence long term planning or downgrade resource estimation and life of mine.
Mass based, automated, cross-belt sampling of different sized streams after crushing and sizing.Bias introduced due to DE, EE/ Med
excessive sampling of coarser material that typically has a slightly higher grade than the average of the stream. Samples split
17%:83% for sizing and grade, respectively. Screening may be inefficient if screen blinding occurs when fragment size is close
to screen size. Sample mass and number of increments per unit mass at go-belt sampler are clearly defined. Time-based
increment extraction.
Metallurgical sub-sampling solids and slurries
Cross-belt samplers are commonly used in the plant for monitoring size and grade. These can introduce significant bias in regard DE, EE, PE/medlow (4)
to both sizing and grade. Cross-stream and belt-end samplers are used on three types of final products and the waste stream.
Sub-sampling in the laboratory
Very carefully controlled protocols for XRF and wet-chemical analysis. PE low (2)
Total 29/50=58%

In the metallurgical plant the ore is crushed, washed, and Iron ore is stockpiled in designated beds as it arrives at the
screened to various sizes and then is stacked according to port and is sampled using a radial arm sampler in a sampling
size and grade. Processes in the plant are strictly controlled tower en route to the vessel for loading (Minnitt, 2010).
using the ISO9000 standards for manganese sampling. The
run-of-mine ores are crushed to -150 mm, and are then Exploration and grade control sampling
further crushed in a secondary crusher to -75 mm. A detailed Extensions to the orebodies are explored with diamond
flow sheet of the process indicates that there are four main drilling at 50 m intervals and reverse circulation infill drilling
products, namely -75 to 60 mm, -60 to 32 mm, -32 to 6mm, at 25 m centres. Detailed geological mapping of benches and
and the fine-grained product, i.e., -6 mm fines. These sized, mining faces as well as drillhole information provides the
sampled products are stockpiled in numbered 280320 t basis for outlining the orebody. Information for geological
stacks (Figure 9), built by compositing ten loads of 25 t each orebody domaining, modelling, mine planning, and
delivered by truck to the pile. production is provided by in-pit grade control drilling on a
25 m x 25 m grid to a depth of 40 m, and percussion drilling
Sampling tower to 12 m. Analysis of data allows a grade distribution block
The samples taken during creation of the stockpiles and model providing a geometallurgically cost-efficient pit layout
loading of the railcars are collected in sampling pots, each and quality control on ores delivered to the plant. Blast-hole
holding up to 2 t of material (Figure 10a). Composite samples sampling provides the basis for classification of the ore types
from the stockpiles (1.51.7 t) and from the railcars (1.2 t) as pure ore (5867.5 per cent Fe), ore (5058 per cent Fe),
are taken to the sampling tower, where they are reduced in and low-density material (2650 per cent Fe) (Nel, 2010).
size and mass to a manageable size of about 200 g that is The grade of the blended iron ore products is initially
sent to the analytical laboratory. controlled during the open pit mining operation by
Samples arriving at the sampling tower are split 17 per demarcating suitable run-of-mine ores for the primary
cent and 83 per cent by mass in a vertical splitter and the crusher. Both the beneficiation properties of the ores and the
smaller fraction is subjected to a size analysis using a series content of iron and silica, together with the deleterious
of screens. The larger fraction (83 per cent) is then hoisted to element (K, P, S, and Al) content (Minnie, 2010, Minnitt,
the top of the tower again and fed through the sampling 2010) which are Kumba Iron Ore specific names (Nel, 2010).
tower to reduce both size and mass, and about 200 g of Producing and blending an acceptable product of 64 per
pulverized sample is sent to the laboratory for XRF and wet cent Fe at 25 mm or 64.5 per cent Fe at 70 mm takes place in
chemical analysis. It is difficult to follow exactly what the the pit at the time of mining. The iron ore quality from
size-mass reduction processes in a typical sampling tower different mining faces (sites) is combined to indicate the
are. The sampling towers are usually aged and there is no required proportion of ore from each face so that an
certainty that equipment on the floors between the top and
bottom of the towers is functioning as it should.
Loading station
The ores are loaded to open railcars via a loading station
(Figure 10b). Each railcar hold approximately 63 t of ore,
with two samples being collected every five railcars loaded.
This means that there are 10 increments of approximately
120 kg each, collected to give a composite sample of 1.2 t
representing 315 t of ore shipped. The sampling rate is 0.38
per cent; the ores are primarily shipped through Port
Elizabeth (Minnitt, 2010).
Figure 10(a) Pots holding up to 1.7 t of ore in which sample material is
collected in the metallurgical plant, and (b) railcar loading facility fed by
Iron ore deposits trucks moving material from blended stacks
South Africas iron-rich sediments of the Transvaal
Supergroup occur in the Northern Cape Province, adjacent to
the massive Kalahari Manganese Field, and in Thabazimbi in
Limpopo Province. South Africas main producers deliver
about 38 Mt/a from these mining operations (Mbendi, 2010).
The ores consist of laminated haematite ore and a
conglomerate ore. These originated through a combination of
primary and secondary haematization producing orebodies
that range is grade from 50 per cent to 68 per cent Fe.
Approximately 37.1 Mt per annum is railed to Saldanha Bay,
from where about 82 per cent is exported and the remainder
used locally. Export volumes have increased annually and
reached nearly 53 Mt of iron ore in 2012, with projections Figure 11(a) Tippler in motion as 100 t of ore is discharged from a
that the total port exports will be 60Mt in 2012. Discharge at wagon, and (b) cross-belt sampler on iron ore conveyor belt (source:
the port is done in a dual line using a tippler (Figure 11b). Kumba Iron Ore, 2010)
acceptable product can be blended. The initial production is splitting stages to obtain a particle size smaller than 0.5 mm.
fed to the so-called Taster Plant to test the beneficiation Compressed powder pellets are prepared from 150 g of
characteristics of the RoM product, and the information is pulverized ore for a combination of wet chemical and XRF
supplied to the DMS plant where necessary changes are made analysis (Fe, Al2O3, SiO2, K2O, P, and S). There is very little
to enable planning of the blending beds. A 45 000 t bed is in the way of QA/QC, and although duplicate samples are
stacked in two hours, so rapid turnaround time on sample taken it is rare that duplicate analyses are made. There is
information is essential (Minnitt, 2010). nothing in the way of application of certified reference
materials, duplicate analyses, or blanks in the analytical
Sampling in the plant process.
Early determination of quality and composition of stockpiled
ore is compiled to produce an inventory of ores prior to their Port loading facilities
dispatch to the plant for sizing and beneficiation. Lower Manganese ore is exported through Port Elizabeth ore
grade ores are washed by dense medium separation or terminal where it arrives by rail after a 30 h 1100 km
jigging to remove diluting contaminants. The final products journey from Hotazel. In the marshalling yard railcar loads of
are sampled on a time basis by cross-belt samplers (Figure manganese ore are split on the basis of grade and conveyed
10b) installed on all final product conveyors. The cross-belt either directly to the ship or to the stacker-reclaimers where
hammer samplers and the cross-stream cutter samplers are of stockpiles up to 460 000 t can be built. When required, the
the highest quality and performance. Unfortunately, the reclaimers are positioned alongside the required grades and
sampling is done on a time rather than a mass basis, which extract the ore from the stockpiles and the ships are loaded.
may introduce significant bias to the results. It has been For iron ores a calculated loading grade for the 200 000 t
suggested that the sampling revert to mass-based collection vessel, against which the invoice is issued, is derived from
of increments (Minnitt, 2010). Samples are split 18:82 for sampling that takes place as the vessel is loaded (Figure
sizing and chemical analysis. Size distribution of samples is 12a), and this value is reconciled back to the original
determined using a range of screens with apertures from 70 blending bed value at the mine. Four stacker-reclaimers
mm to 4 mm (Minnitt, 2010). manage the stockpiles (Figure 12b) with a capacity of 4.8 Mt;
The process for sample handling is a robust, well- the facility is capable of stacking seven different product
designed system. Approximately 45 samples per day, made types.
up of sample increments collected in adequately sized, Stockpiles are reclaimed at 8 000 t/h and the product is
electronically identifiable steel pots, are collected in the delivered to the quay on two conveyors, where dual ship-
metallurgical plant. Geoscan on-line analysers run 24/7 to loaders can load up to 16 000 t/h. The new sampling facility
provide a very good result for iron ores that have been provides continuous sampling of iron ore streams being
washed and screened (Minnitt, 2010). loaded to vessels.
Sampling of blending beds Sampling coal deposits

Blending beds of 100 000 t to 200 000 t are built on a The most extensive and currently productive coalfields occur
chevron arrangement and are sampled exactly according to in the Witbank-Secunda-Carolina area, but the importance of
ISO chemical and sizing specifications every 2 hours as the these fields will be eclipsed by the production anticipated
ore is stacked. Thus an accurate total composition of the bed from the Waterberg coal deposits in the next five to ten years.
is reported to the plant quality control division. The fact that In terms of its global position South Africa is ranked fifth for
iron ore is produced to strict customer specifications means exports (60.5 Mt), sixth for reserves (30 408 Mt), and
the ability to build, reclaim, and control stockpile seventh for production (250.6 Mt), in 2009 (Department of
compositions is essential. Cumulative qualities using Mineral Resources, 2011). The country hosts 7.4 per cent of
weighted averages (analyses weighted with tonnages the world's total coal reserves. Of the total coal production
produced) for each blending bed are used to meet customer South Africa consumes 75 per cent locally and exports about
specifications. The consistency of bed blending and product 25 per cent. The 18 operating coal-fired power stations
quality is determined during reclamation of the beds. owned by Eskom absorb about 65 per cent of the domestic
Sampling of the ore during loading of all railcars (34 200 t) consumption to produce 240 300 GWh annually (Department
allows each car to be matched with the bed specifications. of Mineral Resources, 2010).
The ore is then transported and stockpiled at the Saldanha
port loading facility (Minnitt, 2010).
Sampling tower
Samples collected in specially designed steel pots at sampling
stations throughout the plant report to a dedicated sampling
tower where both sizing and XRF analysis are undertaken.
The sampling tower is a recent and robust addition to the
mines capacity to monitor and control the iron ore products.
The mechanisms are largely automated and allow minimal Figure 12(a) Dual shiploaders at the Saldanha port loading facility,
human intervention in the processes. Preparation for and (b) stacker-reclaimers in the stockpile area at the Saldanha port
chemical analysis consists of four consecutive crushing and loading facility (source: Kumba Iron Ore, 2010)

Total run-of-mine production increased to 317.3 Mt in to be at least three times the nominal top size of the coarsest
2009, but saleable coal production fell to 250.6 Mt under fragment, that cutter capacity should be sufficient to hold the
depressed European markets. The depressed demand from entire increment, and that the cutter velocity must be less
Europe means that the 50 per cent of South African coal than 0.6 m/s (Minnitt, 2010).
exports that Europe accepted until 2009 is currently being
sold off to eastern markets, with current prices (February, Sampling at the open pit mine and port
2013) being around R85 per ton. Total sales value for coal The primary source of information for a new coal
declined 12.6 per cent, with 75.3 per cent being sold locally, development or extension is the exploration drilling. Very
while the remaining 24.7 per cent was exported. Coals are specific sampling protocols govern the way in which such
beneficiated to export quality, low-ash metallurgical, and samples are collected and the core is treated before, during,
steam coal and exported through Richards Bay (Figure 13a and after sampling. This information is the basis for a
and b; Department of Mineral Resources, 2011). feasibility study that includes a layout of the minable orebody
Sampling procedures for coal vary considerably and there is generally no further in situ sampling of the coal.
depending on what the analyses are to be used for, which Once production commences the coal is reduced to -300 mm
may include technical evaluation, process control, and quality in the primary crusher and delivered via conveyors to the
control for international trade. Sampling of coal forms an silos or stockpiles. A secondary and tertiary crush reduces the
integral part of the coal production route, yet it is often a coal to -120 mm and to -32 mm or less, respectively. The coal
misunderstood and underrated activity on collieries. It is a is sampled on a daily basis during the building of the on-site
vital aspect of ascertaining that the coal provided to the stockpiles to monitor the quality of production. There may be
customer is of the correct quality. Since large amounts of two or three sites in the open cut from which coal is
money change hands based on results of samples, it is vitally produced. The coal is reclaimed from the stockpile and the
important that the activity of sampling be made -32 mm to -0.5 mm coal is deslimed by removing the -0.5
unambiguous and regulated especially in the case of mm material. The 0.5-32 mm product is then beneficiated in
international trade (De Korte, 2005). heavy medium cyclones. After washing (removal of high-
Intersections of coal seams in diamond drill core during density rock and inertite) the upgraded product is again
exploration for new coalfields are analysed for calorific value sampled as the stockpiles are built, mainly with respect to
and washability. Bulk samples of coal are extracted as sizing, but percentage ash and calorific value are also taken
rectangular blocks from the full seam thickness. At Richards into account. Coal from these stockpiles is then loaded on
Bay commercial consignments of coal are checked for quality trains destined for Richards Bay Coal Terminal (RBCT). At
as they move along conveyor belts towards the stockpiles. Richards Bay the South African Bureau of Standards (SABS)
Results from stopped belt sampling are compared with samples the coal on an hourly basis-using automatic
automated mechanical sampling devices during bias testing. samplers at two transfer points, before the coal is loaded and
The choice of mechanical or manual sampling from moving when the coal is being loaded into the ship (Johns, 2010).
conveyor belts, stockpiles, or stationary wagons depends on The procedure for coal sampling requires firstly that the
the accessibility of the lot. ISO 13909 (2001) and 18383 coal quality parameters to be analysed be defined. The range
(2005) standards provide the protocols for representative of variables includes calorific value (CV), percentage ash,
sampling by different methods, and are extensively applied in moisture, volatiles, sulphur, phosphorus, ash fusion
South Africa. Sellers and purchasers may jointly modify the temperature, Hardgrove index, etc. The lot and its dimensions
sampling methods depending on costs and technical should be defined, together with the required overall
constraints (Minnitt, 2010). A summary of the sampling precision on the analytical results. This should include a
issues faced by coal producers and sellers is provided in determination or an assumption about the variability of the
Table VII. coal and the variance associated with preparation and testing.
A choice as to the use of continuous or intermittent sampling
Application of international standards methods and equipment, as well as the method of combining
Most coal quality parameters are determined using ISO or increments into composite samples, should be made. The
ASTM standards. These are deeply embedded in the coal number of sub-lots, the number of increments per sub-lot
industry, and while the concepts are virtually identical the required to attain the desired precision, and the minimum
terminology and description of errors, such as the average increment mass must be determined. A decision as to
Delimitation Error, Extraction Error, and Weighting Error as the sampling intervals and whether time- or mass-based
used in the ToS are not used. However, a statement of the increments will be extracted must be stated. The nominal top
basic principles of coal sampling by Johns (2010) suggests size of the coal and the resulting minimum mass of sample
that there is considerable agreement between ISO and the must be ascertained.
ToS. General sampling, coal preparation, and bias testing, Continuous sampling extracts a prescribed number of
mechanical sampling, and sample preparation are covered in increments from every sub-lot. In the case of manual
ISO 1988-1975 and ISO 13909-2001. Manual sampling, sampling a relatively large number of increments must be
sampling of stationary and moving streams, and sample taken, spread evenly over the whole lot. The average value of
preparation are covered in ISO 18283-2005. According to the a cargo can be established with reasonable accuracy if
ISO standards the number of increments and minimum mass sufficient samples are taken. Each sub-lot should be approxi-
of increments per composite depends on the coal size. Johns mately the same size; however, for practical reasons
(2010) provided some additional recommendations in regard sometimes sub-lots of different sizes are used. The estimated
to coal sampling that require cutters on sampling equipment mean should have the correct precision. Intermittent sampling
Table VII
Sampling in the coal mining industry
Sampling site, mechanisms and procedures sampling problems arising from site and mechanism Error/ impact
Exploration
Detailed protocols documented in regard to drilling operations, borehole core logging, coal seam sampling, sample submissions, data DE, EE, PE/
entry, and geological modeling. Reports include coded log sheets, core recovery, and sample mass prediction. Protocol deals with every high-very high
facet of the drilling, logging, and sampling processes. Handling core, splitting, weighing, submission for analysis, proximate analyses. (8)
Relative density estimated by categorizing coals. Coal quality is a critical component of decision making about resource use thermal
coal or export quality. Sampling at the exploration stage can be critical because value can easily be destroyed or overlooked. Sample
Advice sheets record sampling intervals based on geology, quality, or lithology. Geological orebody models compiled from borehole
data are audited, updated, and validated annually. Geological model incorporating orebody structure and coal qualities is overlaid with
the mine plan Layout and the individual mining blocks.
Face sampling, blast-hole sampling, and in-mine grade control
Coal sampling procedure:All coal sampling instructions to operators must be based on the ISO standards, simple, easily understood, DE,EE/high
and capable of only one interpretation. The sample must contain all the particle sizes and all the relative densities in the same (7)
proportions as in the lot to be representative. Face sampling used on rare occasions, to confirm the coal characteristics for a specific
market, or to compare face sampling and borehole intersections in a specific pillar; generally the correlation is good. Typical variables
include moisture, sizing, volatile matter, calorific value, ash, sulphur, phosphorus and ash fusion temperature. Face sampling is
difficult and laborious to do correctly; a very stringent protocol must be used. Sample defined in 2D only. Delimitation and
Extraction errors. Blast hole chips and cuttings are laid out sequentially for every metre of drilling as heaps for logging and sampling.
Samples provide an early indication of coal quality - predicts the product specifications.
Ore Processing and handling
Run of mine: Coal is delivered via conveyors to the silos or stockpiles. Coal from mine is crushed to -300mm in primary crush, and DE, EE PE/med
sampled on its way to the beneficiation plant using a cross-belt hammer sampler, activated on a mass basis. Secondary and tertiary (6)
crush reduces the coal to -120 mm and to -32 mm, respectively. May be two or three sites in the open-cut from which coal is produced.
Sampling rate depends on the required number of increments. Sampled for proximal analyses and efficiency testing.
Blending beds: After beneficiation (removal of high density rock and inertite), coal is again sampled (2-hourly composite) as the stockpiles
are built tomonitor the quality of production. Ensures sizing, % ash and CV are in-spec done on a daily basis. Cross- stream samples
are taken from the lip of the screen for efficiency testing. Both tonnage and grade are used to cumulate an analysis for the stockpile that
is built to ensure it is within specification. Segregation as fines settle to the bottom of conveyor belt and perimeters of stockpiles. In old
stockpiles graded coals can cover the fines, or larger coal may be buried in excess fines, giving rise to pockets of varying grade.
Loading trains: Cross-belt hammer samples of final product are collected as the coal is loaded into the train. Screening panels for
sizing are operated vigorously enough to prevent blinding of the screens. Same procedure as coal is loaded to train bound for RBCT.
Segregation occurs during loading as fines settle to the bottom of railcars or trucks. Segregation is aggravated by vibration during transport.
Manual sampling: The reference method - stopped belt sampling-requires insertion of profile plates into coal ribbon and removal of all
coal between plates. Truck sampling takes increments from 0.3 m below surface. Only applicable to freshly loaded trucks of smalls or
duff if where no layer-loading has occurred. Other manual sampling involves use of probes, mechanical augers, quantum sampling,
ortrenching of compacted stockpiles.
Metallurgical sub-sampling of coal
Aspects to be considered include the inherent variability of the coal, the nominal top size, the minimum sample mass, delimitation, DE, EE, PE/
extraction, and preparation errors, the number of sublots, and the number of increments to be taken per sub-lot. The dimensions med-low
of the lot, the required overall precision, and the sampling intervals (time- or mass-based), must be considered. PGNAA analytical (5)
results are more representative of the primary sample. Size of errors associated with mechanical sampling, provides strong motivation
for use of online scanners. Since 2000 Coalscan has been used for tariff purposes for moisture, CV, volatiles, and ash (Johns, 2010).
Port loading facility at Richards Bay Coal Terminal (RBCT)
Coal stockpiles are reclaimed and loaded on trains destined for Richards Bay Coal Terminal (RBCT). Coal discharged at train tippler. DE, EE, PE/low
In-going sampling by cross-belt samplers. At RBCT the South African Bureau of Standards (SABS) samples the coal on an hourly basis (2)
using automatic cross-belt samplers at two transfer points, when the coal arrives at the terminal (in-go), and when it is loaded into
the vessel (out-going).
Sub-sampling in the laboratory
Conducted under strictly controlled ISO conditions. DE, EE, PE/low
(2)
Total 30/60=50%
be collected from some of the sub-lots, but not from others,

but the same minimum number of increments should be
taken from every sub-lot that is sampled. Sub-lots for
sampling should be chosen at random, except if systematic
choice of sub-lots is demonstrated to be bias-free. If the
variation between sub-lots is too large, it may be necessary
to introduce continuous sampling to achieve the desired
precision (Johns, 2010).
Figure 13(a) Coal tipplers transferring coal at Richards Bay before (b)
being stacked on blending stockpiles
Design of the sampling scheme
Written sampling instructions strictly based on the ASTM or
is used by agreement between contracting parties when ISO standards are issued to sampling operators by
frequently sampled coal has demonstrated a consistency in supervisors once the data requirements and the design of a
type and quality. In such cases intermittent increments may sampling scheme have been reviewed. The material to be

sampled, the overall precision of results, the division of lots, monetary terms are managers able to identify with them
the basis of sampling, the bias of results, and the (Minnitt, 2010).
requirements for test samples must be documented in the The possibility of eroding value from a ferrous metal
instruction. Coal stockpiles at the mine site are reclaimed and operation occurs in the early stages of exploration, extraction,
loaded on trains destined for RBCT. On arrival the coal is and grade control. At current metal prices a grade control
discharged at the train tippler and is sampled at the in-going blast-hole evaluates a block of ground that is potentially
sampling site before it is stockpiled in 68 Mt lots. The worth R140 000. If samples from such blast-holes are biased
stockpile is reclaimed and again sampled at the out-going site or unrepresentative and if 50 such holes are drilled per day
on its way to the vessel. The South African Bureau of the overall revenue at risk could be substantial. During the
Standards (SABS) is the sampling authority at RBCT, evaluation stage the likelihood that a high-grade block (say
sampling the coal on an hourly basis using automatic 64.5 per cent Fe) could be classified as a low-grade block
hammer-type cross-belt samplers that are activated on a (say 62.5 per cent Fe) could mean that the opportunity to
mass basis (Johns, 2010). use the high-grade block as a sweetener for a number of
lower grade blocks is lost. In some cases the difference in
On-line coal analysis grade on which a decision as to how a particular block should
Prompt gamma neutron activation analysis (PGNAA) be utilized is based could be as low as 0.5 per cent. Where
technology has undergone extensive tests over a long period mining projects are in the process of expansion, newly
of time, and has faced strong opposition from those with installed capacity has to be filled by increased production,
interests in maintaining the existing quality control systems. and the larger tonnages are likely to come at slightly lower
In 1999 Eskom appointed the CSIR to evaluate the results of grades and increased overall cost. Economies of scale and
an initial test. The results indicated a better and more new technologies should keep the mine profitable, but the
representative analysis, the primary sample reduction being lower grades mean that accuracy and precision, especially in
about 1:120 with the scanner and 1:2 x 1010 in the the vicinity of cut-off grades, must be carefully guarded.
laboratory. The insight that at least 80 per cent of the For gold and base metal mining operations, better metal
sampling error arises during the sampling event (even if prices mean that ores with lower grades can be profitably
sampling is done properly) provides strong motivation for mined, but this also means that the limits on accuracy and
installation of the on-line scanners, and since 2000 Coalscan precision near the much lower cut-off grades is crucial,
has been used for tariff purposes for moisture, CV, volatiles, especially as the workable limits of detection are approached.
and ash (Johns, 2010). The advantages of the in-line scanner Where selective mining is used the effects of sampling error
are that it has no vested interest in the result, it requires no and bias can be especially severe (Minnitt, 2010).
training, it has no learning curve, it provides real-time data Bulk commodities chrome, iron ore, and manganese
for process control, it warns immediately of out-of-specifi- are produced to customer specifications that are generally
cation coal supply to power stations, and it has performed quite rigidly controlled, with payments for contained metal
well in regard to tariff purposes. The main disadvantage is based on the current price and penalties for out-of-specifi-
that there is no contingency plan for when the analyser is out cation sizing. Pricing for chromite varies between R580 per
of action (Moodley and Minnitt, 2009). ton for 3437 per cent Cr2O3, and R720 per ton for 37-42 per
cent Cr2O3 so the difference in price of R150 per ton is based
on a 1 per cent difference in grade. At the margin a 3000 t
Benefits of good sampling stockpile could be at risk for R450 000 if the analysis is
The benefits of good sampling are difficult to quantify, but incorrect. Penalties of about 30 per cent are payable for out-
there are numerous examples of the costs and losses, in some of -spec lots below 70 mm, so a lot selling for R500 000
cases amounting to millions of dollars, incurred by companies could incur penalties of R150 000 (Minnitt, 2010).
when good sampling practice is compromised (Carrasco et al., Suppliers of manganese and iron ore usually settle within
2004; Carrasco, 2008; Johns 2010; Pitard, 2009; Tice, 2013). 0.5 per cent around the specified grade, usually 46 per cent
Poor sampling practice will influence outcomes, evaluations, for Mn and 64 per cent for Fe. However, at the margins the
estimations, and bottom line cash flows from the earliest 0.5 per cent difference can amount to R150 000 for a 40 000
stages of mineral exploitation and extraction. For example, t manganese consignment and R160 500 for a 150 000 t iron
Carrasco (2008) records losses of US$156 million over a ten- ore consignment. The gain or loss within the 1 per cent range
year period at a copper mine in Chile when the grade of of variability around the specification limits could be to the
tailings was not properly ascertained. The introduction of supplier as well as the customer. On 50 such transactions in a
errors and bias as a result of poor sampling practice is year the revenue at risk would be around R12 million for
probably the main reason for failure to reconcile material manganese and R240 million for iron ore. As mentioned by
flows and balances when trying to account for the metal that Minnitt (2010), the question is: how easy is it to make such
is finally produced. This especially true for precious- and losses?
base-metal deposits where the metal produced is the outcome The back-of-the-envelope calculation by CSIRO (2000)
of what should be a closed circuit in which all the inputs a provides some insight. If during loading of the vessel a 2 t
known tonnage of ore with a known grade, and the outputs composite sample is taken from the stream, a single 70 mm
a known mass of tailings at a certain grade, are matched. particle weighing about 140 g is 0.00007 as a fraction of
Persuading managers to spend the money to rectify the the composite. So the larger particles are underrepresented by
problems, even when they are clearly identified, can be less than one part in 10 000. If the average grade of a large
challenging. Only when errors and bias are translated into particle is 0.5 per cent Mn higher than the average for the
entire stream of ore, a single particle failing to enter the Generally, the risks to project value due to poor sampling
sample cutter would cause a bias of this grade difference tend to be lower in the laboratory. The introduction of QA/QC
(0.5 per cent Mn), times the proportion by which large measures using blanks, duplicates, and standard reference
particles were under represented or 0.5 per cent Mn materials means that problems in the laboratory are identified
0.00007 = 0.000035 per cent Mn. In order to make this up reasonably early, can be monitored, and are usually
to 0.1 per cent you only need 0.1/0.00004 = 2500 particles addressed swiftly. As a general point there is a possibility
or a loss of about 350 kg of material. Since lost material that the value of ferrous metals industry projects could be
simply re-joins the stream of material one would never know more severely affected by errors and bias in all stages of
about it. (Minnitt, 2010, p. 16). mineral developments compared to other mineral
The value of maintaining good mechanical and sampling commodities.
integrity for coal mining operations was described by Long The importance of sampling in the minerals industries is
(2000). A modified example by Minnitt (2010) demonstrated being acknowledged more and more, and the human and
that an increase in moisture of 0.25 per cent or eight cents financial resources needed to ensure good sampling takes
per ton does not sound like a lot. However, this translates place at mining operations are being applied by the industry
into R8000 on a monthly sale of 100 000 t. The example at large. Corporate culture is changing, and in most mining
illustrates that poor sampling could equally favour the operations those who have been ready to take on the role of
supplier or the buyer, and that there is potentially great value champion for the cause of sampling within companies have
in maintaining good sampling integrity. emerged.
It is also evident that the ToS has supplied a framework
Conclusions within which the different aspects of sampling across the
The events of the past three decades have promoted minerals industries can be captured. In the precious and base
awareness of the importance of correct sampling techniques metal industries an understanding of the small-scale
and equipment. Managers are taking note of the potential variability, the Fundamental Sampling Error, and the
and hidden costs associated with poor sampling and rather Grouping and Segregation Error is crucial in compiling a
than retrofitting, which has had to take place in some cases, meaningful nomogram and establishing a usable protocol. As
new mineral processing plants are being designed with one moves towards the bulk commodities the ferrous
stateof-the-art sampling equipment. metals and coal the need for such detailed studies becomes
A summary of the main sampling errors at each of the less important. Instead the use of ISO standards has proved
main stages of mineral development, from exploration to to be sufficient, although the principles of ToS are not to be
point of sale, for the four main commodities examined here is violated in any part of the minerals industry. The fact that
shown in Table VIII. The table is necessarily subjective, but grades in precious and base metal deposits are usually
nevertheless shows some broad and interesting results. The lognormally distributed, while the variables of interest in the
table suggests that the most vulnerable stage in the bulk commodities are normal or negatively skewed, also
development of mineral projects is early, during initial tends to affect the impact and influence of sampling errors.
feasibility and pre-feasibility studies. This is also true for Other aspects that emerged in the 2010 study (Minnitt, 2010)
operating mines where the effects of unrepresentative include the fact that precious metal deposits tend to be
sampling could potentially put huge revenues at risk. For the overvalued (Krige, 1994), while bulk commodities tend to be
bulk commodities the revenue at risk is large for out-of- undervalued at the grade control stages. In the metallurgical
specification product, but generally the standard deviations plants, however, it is the precious metals that are
around the specified grades are less than 0.25 per cent. undervalued and the bulk commodities that are overvalued.
Table VIII
Sampling errors at different stages of mineral development
Stage of mineral development Gold Platinum Ferrous metals Coal

and processing
Exploration DE, EE/ High (7) DE, EE/ High (6) DE, EE/ High (7) DE, EE, PE/ High to very high (8)
Face sampling and in-mine grade control DE, EE/ Extremely high (9) DE, EE/ Very high (8) DE, EE/ Very High (9) DE, EE/ High (7)
Ore processing and handling GSE/ Med-high (6) GSE/ Medium (5) DE, EE/ Med (7) DE, EE, PE/ Med (6)
Metallurgical sub-sampling solids and slurries EE/ Low (2) DE, EE, PE/ Medium (5) DE, EE, PE/ Med low (4) DE, EE, PE/ Medium low (5)
Process sub-sampling solids and slurries/ PE/ Med (4) GSE, DE, EE/ Med-high (6) DE, EE, PE/ Low (3) DE, EE, PE/ Low (2)
point of sale for commodities
Sub-sampling in the laboratory FSE/ Med-high (3) GSE, DE, EE/ Low (2) PE Low (2) Low (2)
Impact score 48% 51% 58% 50%
(Source: Modified after Minnitt 2010, used with permission of AusIMM, 2013)

In general, the finer-grained portions of precious and DE KORTE, G.J. 2005. Sampling in the coal industry, environmental, mining,
base metal material tend to contain the higher grades, metallurgical engineering. Operating and Developing Coal Mines in the
whereas the coarser fragments in the bulk commodities tend Republic of South Africa. Department of Mineral Resources, Pretoria.
to have the higher grades. Bulk commodities require Directory D2/2009.
relatively little in the way of beneficiation to bring them to DEPARTMENT OF MINERAL RESOURCES. 2009. South Africas Mineral Industry
saleable products according to customer specifications, 2007/2008. Department of Mineral Resources, Directorate: Mineral
whereas precious and base metals require many and costly Economics, Pretoria. 207 pp.
steps in order to concentrate and recover the metals.
DEPARTMENT OF MINERAL RESOURCES. 2011. South Africas Mineral Industry
2009/2010. Department of Mineral Resources, Directorate: Mineral
Acknowledgements Economics, Pretoria.
The original research for this paper was carried out in 2009,
ESBENSEN, K.H. 2003. 50 years of Pierre Gys Theory of SamplingWCSB1: a
and the contributors have all been thanked in an earlier
tribute. Aalborg University Esbjerg, Denmark.
publication. However it would be wrong not to thank the
people who helped me again in the compilation of this paper. FLITTON, T.M. 2009. Structural problems at sample stations from the face to the
Numerous individuals assisted in arranging mine and port ingot. Assignment in partial fulfilment of the GDE course in Sampling
visits and provided input from different sources. Special Theory and Practice, School of Mining Engineering, University of the
thanks are due to Marie Nel for organizing the visits to Witwatersrand. 16 pp.
Sishen Iron Ore Mine and the port loading facility at Saldanha
FLITTON, T.M. BARNARD, A.C., AND SPANGENBERG, I.C. 2009. Proposal for an
Bay, and Panna Loubscher and Lionel for their help at the
investigation into the current sampling methodology and the testing of a
Saldanha loading facility. Jonathan Wolfe (African Rainbow new sampling method at Kopanang mine. Internal report REP2009 1020.
Minerals) arranged visits to the Black Rock manganese and AnglogoldAshanti.
Khumani iron ore mines. Sakkie van der Merwe, Pierre Du
Plessis, and Andre Fulhard all provided input at the Khumani FRANOIS-BONGARON, D. and GY, P. 2002a. The most common error in applying
iron ore mine. The assistance of Magda Klindt, Gerhard Gys Formula in the theory of mineral sampling, and the history of the
Botha, Ben Ruziwe, and Danny Maluleka, at Black Rock Mine liberation factor. Journal of the South African Institute of Mining and
operations is acknowledged. Patricia Tennant (Assore) and Metallurgy, November/December 2002. pp. 475480.
Chris Minnie (Kumba) provided valuable insights into the FRANOIS-BONGARON, D. and GY, P. 2002b. Critical aspects of sampling in mills
marketing of ores. Conne Spangenberg and Tarryn Flitton and plants: a guide to understanding sampling audits. Journal of the
(Anglogold Ashanti) assisted in gathering information for the South African Institute of Mining and Metallurgy, November/December
section on gold sampling. Bob Lavery provided guidance and 2002. pp 481484.
input in regard to the sampling in Wits-type gold deposits.
GY, P.M. 1964. Variography. International Symposium of Applied Statistics,
Special thanks are due to Coert Kruger and Erika van Tonder
Golden, Colorado, USA, April 1964. Colorado School of Mines Quarterly,
(Anglo Platinum) who provided exceptional input and contri-
vol. 59. pp. 693711.
butions for the platinum section. Patrick Donlon (Wardrop
Consulting) is thanked for reading and editing portions of the GY, P.M. 1967. LEchantillonage des Minerais en Vrac (Sampling of Particulate
manuscript. Alan Johns (Witlab), Cecilia Barbosa Materials). vol. 1. Revue de lIndustrie Minerale, St. Etienne, France.
(Inspectorate), Gerhard Esterhuizen (Coalserve), Fanie Botha, Numero special (special issue), 15 January 1967.
and Ricky Pinhero provided assistance with the section on
GY, P.M. 1979 and 1983. Sampling of Particulate Materials, Theory and
coal. Dr Geoff Visser made a significant contribution in regard
Practice. Developments in Geomathematics 4. Elsevier Scientific,
to ISO standards applications. The danger with thanking
Amsterdam.
people individually is that I am bound to have inadvertently
omitted someone; in which case please forgive the oversight. GY, P. 1982. Sampling of Particulate Materials: Theory and Practice. 2nd edn.
Elsevier, Amsterdam.
References GY, P.M. 1988. Heterogeneite Echantillonnage Homogeneisation: Ensemble

BERINGER, B. 1938. Underground Practice in Mining. 2nd edn. Mining coherent de theories. Masson Editeur, Paris.
Publications Ltd, London. 277 pp.
GY, P. 2003. Sampling of discrete materials a new introduction to the theory
CAWOOD, F.T. 2003. Underground face sampling on narrow gold reefs: sampling of sampling. I. Qualitative approach. Chemometrics and Intelligent
practice and its impact on the mine call factor on Witwatersrand gold Laboratory Systems, vol. 74. pp. 724.
mines lessons learnt. Journal of the Institute of Mine Surveyors of South
HARRISON, R.N. 1952. Comment on the mine call factor. Journal of the Institute
Africa, vol. XXXI, no. 7. pp. 202224.
of Mine Surveyors of South Africa. pp. 102104.
CARRASSCO, P. 2008. Some key facts and concepts in the evolution of sampling
and assaying practices at Codelco. Proceedings of Sampling 2008. HOL, K. 2009. Sampling protocol at Northam Platinum Limited. Assignment in
Australasian Institute of Mining and Metallurgy, Melbourne. pp. 38. partial fulfilment of the Mineral Resource Management Project on
Sampling Theory and Methods, School of Mining Engineering, University
CARRASCO, P., CARRASCO, P., and JARA, E. 2004. The economic impact of correct of the Witwatersrand. 10 pp.
sampling and analysis practices in the copper mining industry.
Proceedings of the First World Conference on Sampling and Blending. HOOVER, T.J. 1933. The Economics of Mining. Stanford University Press,
Chemometrics and Intelligent Laboratory Systems, vol. 74. pp. 209213. California. 547 pp.
Cook, C.D. 2010. What is ISO 9001 and why is it important to my organisation? JACKSON, J.O. 1946. Mine Valuation: Rand Practice. Louis Gordon Booksellers.
http://www. entrepreneur.com/tradejournals/article/109569568.html 155 pp.
JOOSTE, R, 2007. Audit of sample preparation at a laboratory. An assignment Century: An International Forum in Honour of Michel Davids
completed in partial fulfilment for the postgraduate course on Sampling Contribution to Geostatistics. Dimitrakopoulos, R. and David, M. (eds).
Theory and Methods (MINN7061). School of Mining Engineering, Kluwer Academic Publishers, Boston.
University of the Witwatersrand. 13 pp.
PITARD, F. 2004. New sampling technologies for ore grade control, Metallurgical
KELLY, S.M. 2008. Audit Plant Sampling. Assignment for the course MINN accounting, and laboratory preparation. Metallurgical Plant Design and
7061, Modern Sampling Theory and Methods. Assignment completed in Operations Strategies 2004, Perth. Australasian Institute of Mining and
partial fulfilment of the Masters in Engineering, School of Mining Metallurgy.
Engineering, University of the Witwatersrand, Johannesburg. 9 pp.
PITARD, F.F. 2009. Theoretical, practical and economic difficulties in sampling
KRIGE, D.G. 1952. A statistical approach to some basic mine valuation problems for trace constituents. Proceedings of the Fourth World conference on
in the Witwatersrand. Journal of the Chemical, Metallurgical and Mining Sampling and Blending. Southern African Institute of Mining and
Society of South Africa. pp. 201-215. Metallurgy, Johannesburg. pp. 312.
KRIGE, D.G. 1964. A review of the impact of statistics on mine valuation in the ROBINSON, G.K. 2005. Economic justification for capital expenditure on sampling
gold mining industry. Journal of the South African Institute of Mining and and blending. Proceedings of the Second World Conference on Sampling
Metallurgy, vol. 2, no. 2. pp. 373380. and Blending. Australasian Institute of Mining and Metallurgy,
Melbourne. pp. 111120.
KRIGE, D.G. 1966. Two-dimensional weighted moving average trend surfaces
for ore valuation. Symposium on Mathematical Statistics and Computer ROBINSON, G.K. and CLEARY, P.W. 2009. Should we aim for zero bias or merely
Applications in Ore Valuation. Southern African Institute of Mining and for acceptable bias? Proceedings of the Fourth World conference on
Metallurgy, Johannesburg. pp. 373380. Sampling and Blending. Southern African Institute of Mining and
Metallurgy, Johannesburg. pp. 195202.
KRIGE, D.G. 1994. A statistical approach to some basic mine valuation problems
on the Witwatersrand. Journal of the South African Institute of Mining ROWLAND, R. ST J. 1966. Broken ore sampling. Symposium on Mathematical
and Metallurgy, vol. 1, no. 1. pp. 95111. Statistics and Computer Applications in Ore Valuation. South African
Institute of Mining and Metallurgy, Johannesburg. pp. 324326.
LERM, C. 1994. Evaluation of chip sampling vs the gold analyser. Journal of the
Institute of Mine Surveyors of South Africa, vol. XXVII, no. 5. pp. 7787. ROWLAND, R. ST J. and SICHEL, H.S. 1960. Statistical quality control of routine
underground sampling. Journal of the South African Institute of Mining
LOMBERG, K.G. 2007. Sampling theory and methods; Construction and use of a
and Metallurgy, vol. 60. pp. 251284.
nomogram. Assignment completed in partial fulfilment of the
postgraduate course on Sampling Theory and Methods (MINN7061). SICHEL, H.S. 1947a. Mine valuation and maximum likelihood. MSc dissertation,
School of Mining Engineering, University of the Witwatersrand. 20 pp. University of the Witwatersrand. 21 pp,
LONG, A.M. 2000. Maintaining coal sampling systems managing objectives SICHEL, H.S. 1947b. An experimental and theoretical investigation of bias error
with limited resources. National Weighing and Sampling Association. in mine sampling with special reference to narrow gold reefs. Transactions
http://www.nwsassn.org/files/pg3_amm.pdf of the Institute of Mining and Metallurgy, vol. 56. p. 403.
LYMAN, G., NEL, M., STEINHAUS, R., and BARTLETT, H.E. 2009. Bias testing of SPANGENBERG, I.C. 2007. A standard for broken ore, plant head, and residue
cross belt samplers. Proceedings of the Fourth World Conference on sampling in a typical gold mine. Proceedings of the Third World
Sampling and Blending. Southern African Institute of Mining and Conference on Sampling and Blending, Porto Alegre, Brazil, 2325 October
Metallurgy, Johannesburg. pp. 183192. 2007. pp 282327.
MBENDI. 2010. Iron ore mining in South Africa overview. SPANGENBERG, I.C. 2009. Audit on plant sampling: Kopanang gold plant and east
http://www.mbendi.com/indy/ming/iron/af/ sa/p0005.htm#5 [Accessed gold acid and float plant, broken ore sampling: Kopanang mine and assay:
22 January 2010]. Vaal River Chemical Laboratory. Internal report, Anglogold Ashanti.
MINNITT, R.C.A. 2010. The state of sampling practice in the South African
STORRAR, C.D. 1981. South African Mine Valuation. Chamber of Mines of South
minerals industry. Sampling 2010, Perth, 1112 May 2010. Australasian
Africa, Johannesburg. 470 pp.
Institute of Mining and Metallurgy, WA. 20 pp.
TICE, B. 2013. Screening for a sticky situation, aggregates and road building,
MOSHOLI, T. 2009. Sampling at the Crocodile River mine. Assignment completed
http://www.rocktoroad.com/content/view/1339/67/ [Accessed 22
in partial fulfilment of the postgraduate course on Sampling Theory and
February 2013].
Methods (MINN7061), School of Mining Engineering, University of the
Witwatersrand. 21 pp. TRUSCOTT, S.J. 1947a. Mine Economics. 2nd edn. Mining Publications Ltd,
London. 366 pp.
NAIDOO, P. 2009. Sampling theory identification of structural and circum-
stantial problems in the process sampling of a platinum group metal VISSER, G. 2010. Personal communication. International Organization for
concentrator plant. Report in completion of the MINN7061 course of the Standardization (ISO), South Africa.
Graduate Diploma in Engineering, School of Mining Engineering,
WHITTEN, T.E.H. 1966. The general linear equation in prediction of gold content
University of the Witwatersrand. 25 pp.
in Witwatersrand rocks. Symposium on Mathematical Statistics and
PINHEIRO, H.J. 2010. Coal quality requirements for the metallurgical industries, Computer Applications in Ore Valuation. Southern African Institute of
principles of coal selection, acquisition and substitution, and coal classifi- Mining and Metallurgy, Johannesburg. pp. 124141.
cation for marketing purposes. Presented to the postgraduate course in
WILLS, B.A. 1979. Mineral Processing Technology.1st edn. Pergamon Press.
Coal Utilization, University of the Witwatersrand, 19 February 2010.
WILKE, A.V., CARRASCO, P.C., and CORTES, M.G. 2009. Evolution of the estimation
PITARD, F. 1993. Pierre Gys Sampling Theory and Sampling Practice.2nd edn.
of the metallurgical balance. Fourth World conference on Sampling and
CRC Press, Boca Raton, Florida.
Blending. Southern African Institute of Mining and Metallurgy,
PITARD, F.F. 1993. Exploration of the nugget effect. Geostatistics for the Next Johannesburg. pp. 5362.


Muestreo Metalurgico

Uploaded by

Copyright:

Available Formats

Muestreo Metalurgico

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Muestreo Metalurgico

Uploaded by

Copyright:

Available Formats

Sampling in the South African minerals

cuttings from reverse circulation or blast-hole

Sources of error Nature of error

Source: Gy (1992, 2003), Pitard (1993, 2009)

Sampling in the gold mining industry

Sampling in the platinum mining industry

(Source: Minnitt 2010, used with permission of AusIMM, 2013)

Typical analyses of manganese ore grades

Element, % 47% fines 48% lumpy

Figure 9Numbered and sampled manganese stockpiles ready for

Sampling in the ferrous metals industry

Exploration and blast hole sampling

(Source: Minnitt 2010, used with permission of AusIMM, 2013)

Sampling of blending beds Sampling coal deposits

Sampling in the coal mining industry

(Source: Minnitt 2010, used with permission of AusIMM, 2013)

be collected from some of the sub-lots, but not from others,

Sampling errors at different stages of mineral development

Stage of mineral development Gold Platinum Ferrous metals Coal

Impact score 48% 51% 58% 50%

to have the higher grades. Bulk commodities require Directory D2/2009.

References GY, P.M. 1988. Heterogeneite Echantillonnage Homogeneisation: Ensemble

You might also like