Metal accounting, the core responsibility of process

engineers
Lorenzo Cappai1, Manuel González2*, Stephane Brochot1 and Pierre Vix3
1. Caspeo, Orleans CEDEX 2, France
2. Caspeo Chile, Santiago, Chile
3. Glencore, Switzerland

ABSTRACT
Metal accounting is one of the main tools for financial and technical management of metal production
industry. It is based on measurements and has to manage the uncertainty inherent to the
measurement process generating financial risk. In 2005, the AMIRA P754 project established a code
of best practice for implementing efficient and reliable metal accounting systems.
The application of the AMIRA code is not only linked to the use of given software whatever qualities
it may get or functionalities it may provide. It must be underlined that implementing an accurate
metal accounting system, first relies on the involvement of the whole company, including the top
management, the financial division, the quality assurance team and the production and laboratory
teams, all actors being ready to participate at different levels to the implementation.
The starting point of an efficient metal accounting system is a precise understanding of the process
operations, including the follow-up procedures, based on sampling and measurements. A data
reconciliation algorithm allows obtaining coherent material balance from raw data, improving the
accuracy of data by a statistical distribution of measurement errors, and allowing data quality
checking.
Citing the AMIRA code, “Accounting results must be made available timeously, to meet operational
reporting needs, including the provision of information for other management information systems,
and to facilitate corrective action or investigation”. A software solution, such as INVENTEO,
automates all the repetitive tasks, ensures traceability and workflow and rapidly provides reports.
The paper presents first the data reconciliation algorithm and highlights the need of measurement
errors estimation. Then, the main steps and actions required for the implementation of an accurate
metal accounting system are listed and illustrated with various cases. Examples such as stockpile
management, concentrator, refinery or smelter balances, demonstrate how to address specific needs.

KEYWORDS: Metal accounting, Sampling, Measurement error, Process engineering

1
INTRODUCTION
Metal accounting estimates the quantity of saleable metals produced over a defined period of time
(Morrison, 2008a). It is widely used to quantify the performances of production plants and to estimate
as accurately as possible the metal inventory. Precise information is critical to decrease the financial
risk linked to a wrong estimate of the stocks and to quickly identify losses and process issues. In this
perspective, the AMIRA P754 project established a code of best practice for implementing efficient
and reliable metal accounting systems. Such systems often use a statistical data reconciliation
algorithm to obtain coherent material balance from raw data and improve the accuracy of data by a
statistical distribution of measurement errors. Different software solutions are currently proposed to
help in fulfilling compliance with the AMIRA code. Nevertheless, the use of a given software alone,
whatever qualities it may get or functionalities it may provide, is not sufficient to apply completely
the code of best practice. In fact, the implementation and the maintenance of a reliable metal
accounting system is first of all a project that involves the whole company, from the top management
to the laboratory team, and each actor shares the responsibility for the project to succeed. Amongst
others, the role of the process engineer is particularly important due to his functions linked to the
process management on a daily basis and his precise understanding of the process operations. His
proximity to the process and his know-how allow him a very strong contribution in setting up the
metallurgical accounting system, in maintaining it up-to-date as the process evolves with time, and
last but not least, in using the metal accounting results to adjust the production in order to optimize
it.
The paper presents first the data reconciliation algorithm and how to improve its results, highlighting
the need for measurement errors estimation. Then, the main steps and actions required for the
implementation of an accurate metal accounting system are listed and illustrated with various cases.
For each step, the main responsibilities associated to the process engineer will be highlighted. Real
cases managed with INVENTEO, a metallurgical accounting solution, such as stockpile management,
concentrator, refinery or smelter balances, demonstrate how specific needs can be addressed.

METHODOLOGY
In metal accounting, statistical data reconciliation is often considered as necessary since data are
incoherent regarding the material conservation laws (Hodouin & Everell, 1980; Herbst, Metha & Pate,
1988; Fourniguet et al, 1997). To appreciate such incoherence, it is necessary to perform enough
measurements, which is to say to have a redundant system. Data redundancy means having more
data than the minimum required to calculate the material balance. In fact, when we consider a whole
plant or a single operation unit, the sum of the measured masses of input streams will never be
exactly equal to the sum of the measured masses of output streams, neither for the global mass nor
for single masses of metals, even in steady state conditions. This is because measurements of masses,
moisture and metal contents are random processes, and therefore subject to uncertainty. Such
uncertainty can be evaluated for every measurement by its associated measurement error (Morrison,
2008b). The overall measurement error includes many components that can be divided, following the
Pierre Gy’s classification (Gy, 1979), into two main ones: the total sampling error and the analytical
error. The analysis of the uncertainty budget of the overall measurement error highlights the main
components on which the efforts of improvement can be done (Brochot, 2011). The data reconciliation
algorithm hence finds a set of estimated values so that they are as close as possible to the measured
values according to their errors, and that they verify the material conservation laws (Brochot &

2
Durance, 2012). The information redundancy allows delivering a set of coherent estimates more
accurate than the initial measurements (Ragot et al, 1990; Ragot & Maquin, 2006). Therefore, data
redundancy and the related measurement errors estimation are the necessary starting point for a data
reconciliation that enables to improve the knowledge of the process (Durance, Brochot & Mugabi,
2004). Moreover, the accuracy of data reconciliation results increases with the level of detail and the
quantity of information used to calculate the material balance. When sufficient data redundancy is
available it is hence interesting to take into account intermediate process streams (Lyman, 2005), to
account for the wet mass and related moisture content instead of the calculated dry mass alone, to
include non-saleable components in the material balance calculation and operate metallurgical
accounting also over periods of time longer than one month (Brochot & Durance, 2012). The quality
of data management, processing and checking before data reconciliation is indeed of primary
importance to avoid the “garbage in, garbage out” effect.
To perform the metal accounting calculation over a given period, raw data such as masses, moisture
and metal contents obtained by measurements have to be gathered. Once collected, the total quantity
of material can be calculated as the sum of many mass measurements such as truck loads or
production weights per shift, and mean moisture or metal contents as the weighted average of the
contents of many samples. The aggregation of these raw data gives the “basis data”. Before being fed
to the reconciliation algorithm, measurement errors are attached to the basis data. The accurate
estimation of these errors using the error propagation calculation rules is necessary to correctly
eliminate the incoherencies during the statistical reconciliation. The preparation of a basis data set is
a key operation and can be very time-consuming for plants that have numerous streams. The use of
a metallurgical accounting solution such as INVENTEO can answer the need for automatic data
aggregation, providing the traceability from starting raw data to final basis data, while ensuring the
respect of a well-defined procedure for the calculation of every variable and timely results.
All these considerations must be taken into account during the entire life cycle of a metal accounting
system, from the first implementation of the system to its adjustments due to the evolution of the
production process, and when performing the regular metal accounting operations.
The skills and functions of the process engineer put him in a core position during the entire life cycle
of the metal accounting system. In the life time of a company, different life cycle levels can be
considered for metal accounting (Brochot & Durance, 2012) and the process engineer participates to
all of them. One life cycle concerns the metal accounting system itself, its implementation up to its
validation, the phase of metal accounting reports production, and finally the last reconciliation model
corresponding to the end of the production plant activity.
During the implementation, the process engineer is the one best suited to define the objectives of the
metal accounting and the delimitation of the system, according to the availability of raw data, which
also requires the review of process documentation and of the existing measurement system, and in
case of missing information, choosing the best locations to implement additional points of
measurement. During tests and tuning of the initial reconciliation model, thanks to his proximity to
the process he can study the reconciliation results to check if they are realistic and satisfactory.
After the validation of the metal accounting system, he manages the production of metal accounting
reports. Moreover, he has to identify any possible improvement of the reconciliation model and
evaluate possible enlargements or contractions of the delimitation of the system. Changes may be
due to improvements in the measurement system and data management, but also to process
modifications, new raw materials, new products or units of operation. After the identification of the
needs for updating the reconciliation model and their implementation, a new phase of tests and

3
tuning is necessary to validate the updated version of the reconciliation model. On the long run, the
process engineer may also identify actions to improve altogether the process efficiency and its
accountability, for example modifying the stock management using smaller parallel stockpiles when
possible.
When the choice to stop the production and close the site has been taken, the configuration of the
final reconciliation model is necessary to account for the stock exportation, tailings disposal and
management of the wastes generated by the dismantlement.
During the life cycle linked to the calculation of each metal accounting report, performed in the frame
of the production phase as defined above, the process engineer is responsible for checking the
availability and correctness of raw data, identify exceptional events that could affect the material
balance, and this, very often during the considered period. Some operation units or measurement
points may be down, or a stock may be managed with an uncommon manner due to an unexpected
problem. Again, his proximity to the process is a major advantage. Once all raw data are available,
verified and fixed he has to manage the reconciliation process and generate intermediate reports
before their validation by the management staff. A metallurgical accounting solution such as
INVENTEO can help in flagging mistakes in raw data thanks to automated diagnostic tools. It can
manage the workflow offering a complete traceability from the raw data extraction, up to their
aggregation, including their possible manual adjustment and approbation.
The next section presents a series of real cases encountered during the implementation of
metallurgical accounting systems with INVENTEO.

RESULTS AND DISCUSSION

This section presents several approaches to reach the best results with data reconciliation by material
balance. It is illustrated with real cases encountered during INVENTEO implementations. As shown,
the same principles can be applied to a variety of processes dealing with production of different
metals.

Material balance of the flotation circuit of a gold ore concentrator

In the flotation plants it is usual to lack the measurements concerning the amounts of produced
concentrates and tailings. It is generally the case when concentrate is immediately processed
downstream. Similarly, flowrate and solid percent of tailings are rarely or inaccurately measured.
Very often, such amounts are calculated using the “two products” formula or the “three products”
formula when two concentrates are produced. The formula bases its calculation on the amount of the
material feeding the flotation plant, and the contents of one of the components in feed, concentrates
and tailings. The formula is based on the assumption that there is no accumulation of material in the
plant (the tank levels are constant). When only one concentrate is produced, the formulae to calculate
the amounts of concentrate (1) and tailings (2) are:
(𝑓−𝑡)
𝐶=𝐹 (1)
(𝑐−𝑡)
(𝑐−𝑓)
𝑇=𝐹 (2)
(𝑐−𝑡)

Where F, C and T are the quantities of feed, concentrate and tailings; and f, c and t their corresponding
contents of one of the components. It must be reminded that such procedure is considered an
exception by the AMIRA code, that must be clearly justified in the metal accounting protocols and

4
used only if direct measurements are lacking. Therefore, the addition of flowmeters or belt weighers
to have better figures must be considered when possible. Nevertheless, when the formula is used
some choices allow to improve the calculation results. First of all, the metal (or component) used in
the formula has to be well defined: it could be one of the metals of economic interest, but also another
major component that can be accurately measured. In fact, a major component, even if non-saleable,
can enhance the estimates. It must be kept in mind that the larger the measurement errors associated
to the contents, the larger the errors associated to the calculated quantities. Following is an example
related to a flotation circuit treating gold ore, where solid quantities of concentrate and tailings are
calculated using the three gold contents and the dry mass of the feed by means of the “two products”
formula. As seen in Table 1, the estimates of the standard deviations for the gold contents of feed,
tailings and concentrate that take into account sampling and analytical errors highlight a noteworthy
uncertainty, a problem frequently encountered when working with gold contents. Corresponding
measurement errors, calculated as twice the standard deviation, range between 12 and 19 percent.

Table 1 Gold contents of feed, tailings and concentrate and the estimates of their standard
deviations

Relative standard Absolute standard

Stream Au (g/t) deviation (%) deviation (g/t)

Feed 2 6 0.12

Concentrate 29 7 2.03

Tailings 0.4 9.5 0.04

Table 2 shows the standard deviations associated to the calculated quantities of concentrate and
tailings, deduced by means of the error propagation formula. Results are not accounting for the
standard deviation of the feed quantity, and yet, in Table 3 it is possible to observe that the use of
such approach leads to a non-negligible uncertainty in the amount of gold recovered in the
concentrate. The variability coming from the “two products” formula calculation method is
represented by an absolute standard deviation equal to 1.27 kg of gold per day. The amount of gold
recovered in the concentrate is then estimated as 16.22 kg ± 2.54 kg every day, with very relevant
financial implications. This calculation is assuming that samples are correctly selected and managed,
avoiding biases. If bias affected just a single one of the considered contents, the situation would be
even worse.
It is interesting to study how the “two products” formula can be exploited in data reconciliation, in
which, its result represents only an initialization for the values of masses of concentrates and tailings,
while their most probable values will be calculated thanks to the subsequent reconciliation by
material balance. The simultaneous material balance calculation of many components adds
redundancy to the system and improves the accuracy of the metal accounting (Morrison, Gun &
McCallum, 2002; Guerney, Dunglison & Cameron, 2005). In the event that only the component used
in the formula is considered in the material balance calculation, the redundancy is not sufficient, since
there is just enough information to calculate the material balance.

5
 Table 2 Calculated concentrate and tailings quantities and their associated standard deviations
calculated with the error propagation formula
Relative standard Absolute standard
Stream Mass (t/d) deviation (%) deviation (t/d)

Feed 10000

Concentrate 559.4 10.57 59.1

Tailings 9440.6 0.63 59.1

Table 3 Quantities of gold in the concentrate and the tailings and their associated standard
deviations calculated with the error propagation formula

Relative standard Absolute standard

Stream Mass of Au (kg/d) deviation (%) deviation (kg/d)

Feed 20

Concentrate 16.22 7.83 1.27

Tailings 3.78 9.65 0.36

In the flotation circuit presented above there is no redundancy of data. Therefore, improvements of
this node balance are impossible without measuring at least another component for each stream.
These considerations were taken into account during the INVENTEO implementation. Iron, arsenic
and silica were added to the material balance calculation to increase redundancy. Unfortunately,
sulfur contents were not available for all the streams on a regular basis, not allowing a stable
redundancy with respect to this element and hence to consider it for a periodical accounting.
Moreover, since these contents can be measured more accurately than gold ones, silica was used
instead of gold in the “two products” formula for a more accurate initialization. Silica was preferred
to the others because it reaches considerably higher contents, ranging from 23 up to 66 percent of the
dry masses, representing alone more than half of feed and tailings dry masses. Moreover, as fed silica
mainly reports to the tailings its contents allow to better describe the feed to tailings mass ratio. For
data reconciliation, all the elements are taken into account. Gold and arsenic contents will contribute
to a better estimate of the concentrate, while silica content will contribute to a better estimate of the
tailings. This example shows the importance of redundancy and of measurement errors calculation
necessary to exploit it, and how they can enable to get the most out of data even under unfavorable
conditions.

Stockpile management and batch processes in a copper smelter plant

Smelter plants receive large amounts of concentrates from various sources and the operations of
dispatching and blending require a large number of stocks. Accumulation of material concerns also

6
WIP (Work in progress) and intermediate stocks of reverts. Stocks are very important concerning
metal accounting as they contain large amounts of metals, especially in the last processing steps of
smelters and refineries. Stocks lock up much value but their estimate are generally less accurate than
the material movements (Wortley, 2008; Holmes, 2008; Connelly, 2009). To add redundancy and
improving results when using data reconciliation, it is interesting to account each stock individually
through a specific devoted node in the material flow diagram. One or more input streams can feed
the stock and one or more output streams can reclaim from it. The quantity and the quality of the
stock are accounted for at the beginning (opening) and at the end (closing) of the considered period.
The beginning inventory can be seen as the first input stream of the period and the ending inventory
as the last output, the ending inventory becoming the beginning inventory of the following period.
Such approach not only adds redundancy, it also gives an estimate of the stock amount and quality
at a given time.
When direct measurements of a stock are difficult or non-representative (for example, it is impossible
to correctly sample a stockpile), it can be preferable to back-calculate the quantity and/or the contents
of an inventory from the data concerning the feed and the reclaim of the stock. Such data are
measured in material flow conditions. In that case, a “stock model” considers the feeding and the
reclaiming streams with their timetable (quantity and quality of material versus time) to draw a
picture of the consecutive quantities and qualities of materials that have been fed to or reclaimed
from the stock. In order to calculate a weighted average, masses can be linked to their corresponding
contents by means of this timetable or lot numbers. The stock model can be considered as a perfect
mixer (such as an agitated tank or a blending bed), a FIFO (first input, first output such as a silo) or a
LIFO (last input, first output such as a box). The main risks of this approach are the possible
accumulation of the inaccuracies linked to the material movements into the stock and the fact that
the stock does not operate exactly as a mixer, FIFO or LIFO model. This calculation alone can hence,
after many months, create a figure far from reality. It must therefore be considered as an initialization
for the data reconciliation algorithm, to be coupled with the appropriate calculated error. Different
approaches are compared with a numerical example considering a stockpile of copper concentrate,
fed and reclaimed during the period. Table 4 reports the known quantities and copper contents of
movements and inventories. The content of Reclaim 1, the quantity and the content of the ending
inventory have to be estimated.

Table 4 Quantities and average copper contents of feeding movements and inventories

Movement/Inventory Mass (t) Cu (%) time

Beginning inventory 73 20 t0

Feed 1 52 28 t1

Reclaim 1 54 Not measured t2

Feed 2 56 29 t3

Ending inventory Not measured Poorly sampled t4

7
Table 5 reports the results of different calculation approaches. The first approach considers a global
material balance over the entire period, and that the contents of the reclaim and the ending inventory
are identical. It is wrong because it does not take into account the time when movements occurred:
the content of Reclaim 1 at t2 also considers Feed 2, happening later at t3. The second approach
considers a stock model with perfect mixing: feed streams are perfectly mixed with the pre-existing
stock. Feed 2 is not considered when calculating the content of Reclaim 1, and is accounted in the
calculation of the ending inventory content, leading to the calculation of two different contents. This
approach can be considered more accurate than the first one. The problem is that considering a
stockpile as a perfect mixer remains a bold assumption. Unfortunately, there is no alternative since
getting a representative sample is impossible. Consider the stockpile at t2 perfectly shaped as an
equilateral cone, with a “true” average copper content equal to 23.33 %. Feed 2 is fed in such a way
as to ensure that the final cone is still an equilateral one. At t4 a weighted average defines a “true”
average copper content equal to 25.83 %. As shown in Figure 1, if we tried to sample such a cone with
an auger from the middle of the apothem and perpendicularly to the surface, the resulting sample
would present an important increment weighing error. Due to the geometry of the system, the
increments of the two qualities would not be proportional to their real mass proportion in the
stockpile. This sample would lead to a very different estimate of the ending inventory copper content,
that based on geometrical considerations can be estimated as 24.83 %. This simplified example does
not consider the segregation phenomena depending on density and size classes of particles and yet
it shows how figures could get strongly affected by a non-representative sample. At the end of the
period, depending on the chosen approach, the copper contained in this single stockpile can be
estimated with a difference of more than one ton.

Table 5 Missing values obtained with different methods

Approach Movement/Stock Mass (t) Cu (%) Cu (t)

Ending inventory 127 25.08 31.86

Global mass balance
Reclaim 1 54 25.08 13.54

Ending inventory 127 25.83 32.80

Perfect mixer stock model
Reclaim 1 54 23.33 12.60

Ending inventory 127 24.83 31.53

Sampling the stockpile
Reclaim 1 54 25.68 13.87

8
 Figure 1 Simplified geometry of a stockpile

Above all, much can be done practically in the plant in terms of stock management, such as building
small stocks with frequent depletion. This can be achieved by building small parallel stockpiles, one
being reclaimed while the other is being fed. This allows to frequently “set the clock to zero” and
avoid huge build-ups of value in the stocks, reducing financial risk. When volumes of material are
relatively small, it can also be interesting to avoid mixing different lots of production. This practice
allows the production of small stockpiles for which metal contents are well known. Also, when the
remaining volume of a stock is sufficiently small, it is the opportunity to directly weight it, instead of
calculating its dry mass from a volume measurement and assuming its density. These practices are
dictated by common sense and can really make the difference in increasing the knowledge and
accuracy of plant stocks.
The implementation of INVENTEO in this copper smelter plant was particularly successful thanks
to the coupling of good practices together with well-designed data management and data
reconciliation. Thanks to automatic filters, data concerning feeding and reclaiming streams are
automatically aggregated and used to calculate the inventory mass and contents with a stock model
chosen on case-by-case basis, depending on the stock management. The subsequent data
reconciliation improves the knowledge of the inventories. The role of process engineers on the field
must be underlined as it allowed the improvement of the process efficiency and its accountability
altogether.

Hydrometallurgy and liquid accounting in a zinc smelter plant

In mineral processing, the definition of a wet mass with corresponding solid and metal contents is
sufficient to track the metals contained in solids along the process. In hydrometallurgy, metals can
be present in both liquid and solid phases and they can pass from one to another. A more detailed
phase definition is hence necessary. In a zinc smelting process, leaching and electrolysis operations
must be accounted together with roasting and melting ones. Figure 2 reports a simplified flowsheet.
Some streams are wet solids (roaster feed, cements and cake from leaching…), others are completely
dry (calcine, cathodes, ingots…) or completely liquid (spent electrolyte, purified solution…). The
possibility to define several phases is a major feature to allow a detailed metal accounting for such
process.

9
Depending on the process conducted in each considered process step, and on the data availability,
different conservation laws should be defined for each node of the flow diagram. In the node
associated to the roasting operation the loss of sulfur and moisture and the gain of oxygen modify
the total wet mass. Liquids and solids are not conserved. Instead, metals in solids must have the same
amount in input and output streams. For leaching and cementation operations the conservation of
the total amount of liquids can be assumed. Moreover, metals can pass from solids to liquids and vice
versa. Hence their total amount must be conserved whichever the phase. Approaching the end of the
process, metals other than zinc are present in low concentration and often poorly measured. For
instance, in the electrolysis node, zinc is mainly present as other metals have been previously
removed by cementation. In this case, only the zinc conservation constraint is considered in the
material balance calculation since not enough information is available for the other metals. For
melting of cathodes and casting of ingots operations the conservation of the total dry mass and the
zinc contained in it can be assumed.

Figure 2 Simplified flowsheet of a zinc smelting process

Table 6 reports an example of possible imbalances for some conservation laws considered in the
material balance. These imbalances are due to measurement errors and can be eliminated thanks to
data reconciliation. Positive values indicate that the input masses are larger than the output ones,
negative values show that output masses exceed the input ones.

10
 Table 6 Example of imbalances before data reconciliation

Operation Conservation law Imbalance

(kg)

Conservation of zinc in solid phase 8291

Conservation of lead in solid phase -1217

Roasting
Conservation of copper in solid phase 96.34

Conservation of silver in solid phase 2.793

Conservation of liquid mass -146700

Global conservation of zinc in solid and liquid phases -80560

Leaching and cementation Global conservation of lead in solid and liquid phases 1056

Global conservation of copper in solid and liquid phases -93.43

Global conservation of silver in solid and liquid phases -8.586

Electrolysis Global conservation of zinc in solid and liquid phases 28390

Conservation of solid mass -12100

Melting and casting
Conservation of zinc in solid phase -12100

As shown, the definition of conservation laws has to be very flexible and can be different depending
on the process conducted in each node and on the data availability. For example, one metal can be
conserved in one node and not considered in another one if not enough information is available to
close a balance around that node.

CONCLUSIONS
During the implementation of a metallurgical accounting system, the review of the available data can
give important hints about the flow diagram to be considered. Similarly, the measurement error
estimation points the way forward the improvement of raw data, already before statistical data
reconciliation. The article reports examples of how to define the material flow diagram, the
conservation laws to be considered and the number of considered components, to improve the
accuracy of the final metal accounting by means of data reconciliation by material balance. The use
of a metallurgical accounting solution such as INVENTEO can answer the need for automatic data
aggregation, providing the traceability from starting raw data to final reconciled data, while ensuring
the respect of a well-defined procedure for the calculation of every variable and timely results. The
success of the implementation depends strongly on the implication of the company personnel. In
particular, process engineers have a core role thanks to their functions linked to the process

11
management on a daily basis and their precise understanding of the process operations. In fact, their
proximity to the process is crucial in setting up the metallurgical accounting system, in maintaining
it up-to-date as the process evolves with time, and last but not least, in using the metal accounting
results to optimize the production.

REFERENCES
AMIRA International (2007) Metal Accounting: Code of Practice and guidelines, Release 3, AMIRA
International Africa.

Brochot, S. (2011) ‘The application of sampling theory in metallurgical accounting process – Inventeo
methodology implementation’, Proceedings Fifth World Conference on Sampling and Blending, M. Alfaro,
E. Magri, F. Pitard, Santiago de Chile, pp. 185—193.

Brochot, S., and Durance, M.V. (2012) ‘A new approach to metallurgical accounting’, Proceedings 11th
AusIMM Mill Operators’ Conference, The Australasian Institute of Mining and Metallurgy, Melbourne,
pp. 217—224.

Connelly, D.E.G. (2009) ‘Metallurgical accounting standards in process plants’, Proceedings Tenth Mill
Operators’ Conference, The Australasian Institute of Mining and Metallurgy, Melbourne, pp. 281—287.

Durance, M.V., Brochot, S., and Mugabi, M. (2004) ‘Material balance approach for parameter
determination in Bioleaching process’, Proceedings 9th International Symposium on Computer Applications
in Biotechnology (CAB9), Nancy.

Fourniguet, G., Villeneuve, J., Rocchia, L., Védrine, H., Brochot, S., and Guillaneau, J.C. (1997) ‘Analyse
par bilans matière d’un procédé hydrométallurgique de recyclage de déchets d’aciéries électriques’,
Proceedings Sixth French Congress of Process Engineering, Société Française de Génie des Procédés, Paris.

Guerney, P., Dunglison, M., and Cameron, P.M. (2005) ‘Leveraging accuracy and precision — Multi-
phase mass balancing and reconciliation as a tool for quality data management’, Proceedings Second
World Conference on Sampling and Blending, The Australasian Institute of Mining and Metallurgy,
Melbourne, pp. 121—125.

Gy, P.M. (1979) Sampling of particulate materials – Theory and practice, Elsevier, Amsterdam-Oxford-New
York.

Herbst, J.A., Mehta, R.K., and Pate, W.T. (1988) ‘A hierarchical procedure for mass balance closure and
parameter estimation: application to ball mill grinding’, Proceedings XVI International Mineral Processing
Congress, K.S.E. Forssberg, Elsevier, Amsterdam, pp. 1719—1732.

Hodouin, D., and Everell, M.D. (1980) ‘A hierarchical procedure for adjustment and material balancing
of mineral process data’, International Journal of Mineral Processing, vol. 7, pp. 91—116.

Holmes, R.J. (2008) An Introduction to Metal Balancing and Reconciliation, Julius Kruttschmitt Mineral
Research Centre, Indooroopilly, pp. 141—170.

Lyman, G.J. (2005) ‘Sampling, metallurgical accounting and reduction of balance estimation variance’,
Proceedings Second World Conference on Sampling and Blending, The Australasian Institute of Mining and
Metallurgy, Melbourne, pp. 127—135.

Morrison, R.D. (2008)a An Introduction to Metal Balancing and Reconciliation, Julius Kruttschmitt Mineral
Research Centre, Indooroopilly, pp. 3—23.

Morrison, R.D. (2008)b An Introduction to Metal Balancing and Reconciliation, Julius Kruttschmitt Mineral
Research Centre, Indooroopilly, pp. 27—75.

12
Morrison, R.D., Gun, Y., and McCallum, W. (2002) ‘Metal balancing from concentrator to multiple ore
sources’, Proceedings Value Tracking Symposium 2002, The Australasian Institute of Mining and
Metallurgy, Melbourne, pp. 141—147.

Ragot, J., Darouach, M., Maquin, D., and Bloch, G. (1990) Validation de Données et Diagnostique, Hermès
Science Publications, Paris.

Ragot, J., and Maquin, D. (2006) ‘Validation et réconciliation de données. Approche conventionnelle,
difficultés et développements’, Les Techniques de l'Industrie Minérale, vol. 29, pp. 22—30.

Wortley, M. (2008) An Introduction to Metal Balancing and Reconciliation, Julius Kruttschmitt Mineral
Research Centre, Indooroopilly, pp. 77—139.

13