You are on page 1of 12

HOME

Evaluating All the Alternatives to Select the Best Project Strategy


B Hall1
ABSTRACT

TABLE 1

It is well known that the ability to influence project value reduces


significantly as planning progresses and decisions regarding project
parameters are made. However, at the prefeasibility stage in particular,
where one of the main aims should be to evaluate a wide range of options
in order to select the best for more detailed further study, there is often a
reluctance to spend the time and money needed to do this rigorously.
This paper demonstrates, by a number of case study examples, how
significant increases in value may be obtained by fully evaluating the
combinations of alternatives for various project design parameters,
including (though not limited to) mining methods, haulage methods,
mining rates, processing methods and rates, sequencing of mining, and
cut-off grades, with changes and upgrades over time where applicable.
The additional value unlocked is typically orders of magnitude greater
than the additional cost of the more comprehensive study.
The dangers of misuse of simplistic sensitivity analyses and failure to
identify how strategies might change in response to changes in external
factors are also illustrated, and the importance of selecting the best grade
descriptor is discussed.

Summary of optimisation studies conducted.


Level of detail of study
Scoping
Prefeasibility
Feasibility

Locations

Australia
Western Europe
South Africa
East Africa
Central Asia
China
India

Minerals

Gold
Copper
Lead/zinc
Nickel
Platinum
Mineral sands

Types of operations
Single underground mine and treatment plant
As above, plus other independent ore sources (pre-existing stocks
and satellite mines)
Single open pit mine and plant with stockpiling
Single deposit and plant, with interacting underground and open pit
mines
Multiple deposits with a single treatment plant
Multiple deposits with multiple treatment plants

INTRODUCTION
The author and his colleagues at AMC Consultants Pty Ltd have
conducted a number of mine plan strategy optimisation studies in
recent years. The term mine plan is used in the widest sense,
and may include treatment and processing options, as well as
mining options such as haulage systems, mining methods and
cut-off grades.
Studies have covered many types of operations, minerals and
locations, at various levels of detail, as shown in Table 1. This
paper first describes the various stages of studies, how benefits
far exceeding the additional cost can flow from an exhaustive
evaluation of alternative strategy options, and methodologies
used for strategy optimisation studies. The importance of
selecting the best grade descriptor is discussed. Results from a
number of case studies are then used to illustrate these principles.
The type of operation is identified, but to maintain client
confidentiality the names of operations are not disclosed.

TYPES OF STUDIES AND THEIR PURPOSES

Strategies adopted at later stages of evaluation may be


significantly different from that used in the scoping study, without
implying that the scoping study was incorrect. Anecdotally there
seem to be instances where the opposite is true, and either the
credibility of those conducting the scoping study is called into
question, or subsequent studies are constrained to the conceptual
study strategy. In particular, while other strategic options may be
re-assessed, the cut-off grade in the scoping study may be used for
later evaluations without further analysis.

Prefeasibility study
The prefeasibility study (PFS) should have a two-fold purpose:

to demonstrate to a higher level of confidence (typically 20


- 25 per cent) that the project is viable, and

to evaluate an exhaustive range of options to ensure that the


best option is selected for further detailed study.

That there are a number of types of studies, with different levels


of accuracy, is generally well known, but the purposes of each
stage of the process, and hence the requirements and uses for
which the results may be employed are sometimes forgotten.
Various study classification schemes have been published
(McCarthy, 1998; Noort and Adams, 2006). This paper assumes
the following terminology and associated principal purposes, and
hence study requirements.

Many studies will evaluate a limited number of options for


various strategic design parameters, but generally only
considering variations from the base case for one parameter at a
time, rather than for all design parameters concurrently: it is
considered too costly and time-consuming to evaluate all
combinations of possible strategic options. The contention of this
paper is that this need not be a burden, and the benefits of
conducting a wide-ranging investigation at PFS stage can be
orders of magnitude greater than the extra cost.

Scoping/conceptual study

Feasibility study

Typically stated to be at a high-level, order-of-magnitude


accuracy (30 per cent), the major purpose of this initial level of
study is to demonstrate early in the resource discovery and
evaluation process that further more detailed work is justifiable.
It is therefore only necessary to find one scenario which shows
that to be so. Strategy optimisation is generally not necessary
unless the case evaluated does not demonstrate likely viability
and a better case must be found.

The feasibility study is intended to demonstrate technical and


economic viability of the project at a level of accuracy (typically
10 - 15 per cent) sufficient to justify the decision to proceed
with or to abandon the project.
Optimisation studies may be done after a major feasibility
study, but often these merely investigate ways of reducing the
costs or improving the efficiencies of the selected strategy, rather
than looking for better strategies. Also, if the study results are
positive, the identified plan will be implemented quickly; delays
for further options studies are normally not acceptable. If the best
strategy has not been identified by the PFS, it is therefore
unlikely to be found during the feasibility study.

1.

MAusIMM, Principal Mining Engineer, AMC Consultants Pty Ltd,


12/179 North Quay, Brisbane Qld 4000.
Email: bhall@amcconsultants.com.au

Project Evaluation Conference

Melbourne, Vic, 19 - 20 June 2007

53

B HALL

While the feasibility study will have done mine, plant and
infrastructure designs to a certain level of practicality and detail,
the primary purpose is to verify the technical feasibility of the
proposed designs and processes, and derive sufficiently accurate
quantities to support the predicted financial outcomes. It will
usually not produce plans that can be immediately implemented
for mining and construction.

resource/reserve, and/or to include material that should be waste.


Together these effects degrade the resource/reserve, as shown in
Figure 1. These errors are in addition to any estimation errors in
the block grades.
At many polymetallic mines, grade descriptors are calculated
by multiplying each metal grade by a multiplier and summing
these to obtain a dollar value or metal equivalent grade.

Detailed engineering

Dollar values

Detailed engineering and design are therefore required to convert


the principles in the feasibility study into construction and
mining plans for practical implementation.
It is critical that all aspects of the short-, medium-, and
long-term planning functions are implemented in the early years.
Focus only on the short-term requirements to construct and
commission the mine may lead to a reduction in output several
years into the mine life, as documented by Ward and McCarthy
(1999). The feasibility studys longer term plans will become
superseded as development, production and knowledge of the
orebody progress and improve. Longer term plans must therefore
also evolve during the commissioning phase.

THE IMPORTANCE OF THE GRADE


DESCRIPTOR
The resource model is the starting point for any strategy
evaluation. A key parameter is the grade, which identifies higher
and lower value areas in the deposit, and via the cut-off grade is
used to specify what is classed as ore and waste. There is an
implicit assumption that the stated grade is correlated with value:
higher grade blocks have higher value than lower grade blocks.
In a simple single-metal deposit, the metal grade usually fulfils
this assumption.

Polymetallic grade descriptors in homogeneous


deposits
In a polymetallic deposit, a single value is needed as the grade
descriptor. With a perfect grade descriptor, there will be rank
correlation between the true value and the calculated grade
descriptor; ie the ranking of blocks by the grade descriptor is the
same as their true ranking by value. The relationship between the
two is not necessarily linear. An imperfect grade descriptor will
show a scatter of points above and below the trend-line of a value
versus grade plot; the grade descriptor does not rank all blocks
correctly by value. The effect of this incorrect ranking is, at any
specified cut-off, to exclude material that should be ore from the

In situ dollar values, which account only for the predicted metal
prices and estimated grades, have the potential to incorrectly
rank potential ore sources with respect to the value they would
generate. Recoverable values are theoretically an improvement,
as they account for different metallurgical recoveries for different
products. Payable values are theoretically better again, as they
account for proportions of the metal price that are received by
the operation after costs associated with products have been
deducted from the prices. These include treatment and refining
charges and freight and realisation costs. They account for
different proportions of the metal price for different metals.
For in situ or recoverable values, a simple calculation to
generate equivalence multipliers is often satisfactory, as the
relationships are also typically relatively simple. For payable
values, the typical process to derive multipliers is to start with the
average head grade and work back from the net smelter revenue
(NSR) obtained to determine the contributions from each
metal. Where, for example, precious metals are recovered as a
component of a base metal concentrate, accounting-type
distributions of common costs are typical, but this can lead to
inappropriate cost allocations, which impact on the multipliers
and hence relative weightings for the various metals in the
calculated grade.
An estimate of the true payable value can be calculated by
using the operations metallurgical and financial model to
generate a notional net smelter revenue (NNSR) for mill feed
with the grades of any block of rock. With some metallurgical
recovery relationships this NNSR value will be the true
contribution made by treating it. In many other cases the true
value of an incremental tonnage depends on what it is blended
with. Such an NNSR value, while therefore not necessarily
true, will usually correctly rank blocks relative to each other.
The NNSR calculation process could be applied to all blocks
in the geological modelling software, and some mines do this.
The complexity of metallurgical processes at others can make
this impractical. However, relationships between NNSR and ore

FIG 1 - Misclassification of ore and waste using an inappropriate grade descriptor.

54

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

EVALUATING ALL THE ALTERNATIVES TO SELECT THE BEST PROJECT STRATEGY

grades can be found by calculating NNSRs for the grades of a


representative sample of blocks. A regression of the calculated
NNSRs against the grades from which they are derived can be
performed, and the identified relationship expressed as a set of
multipliers to apply to the grades to obtain dollar value
equivalent grades. These multipliers are statistically the best that
can be derived, and typically reduce the misclassification of ore
and waste when compared with using multipliers derived by the
more usual accounting-style methodology.
When dollar values are used, there is often an assumption that
various unit cost measures can be used directly as cut-offs.
However, since the dollar value derived is not necessarily the true
value, using unit costs as cut-offs may not achieve the intended
results. Dollar value grade descriptors are also typically
recalculated each time a new official set of forecast prices is
adopted, and the existing cut-offs are applied to the new dollar
values. If the cut-offs are break-evens, the potential inaccuracies
will still apply. However, if the cut-off has been optimised, this
process can change the effective cut-off and result in the applied
cut-off becoming suboptimal. It is beyond the scope of this paper
to discuss this further.

Metal equivalents
Metal equivalent may be a better grade descriptor than dollar
values. The metal equivalent multipliers are found by dividing
the dollar value multipliers by the multiplier of the major metal.
Metal equivalent tends to give a more stable grade value than
dollar value, and may not need to be updated each time metal
price forecasts change unless there are significant changes in the
ratios of payable values of each metal. Its use also breaks the
perceived but not always accurate link between dollar values and
unit costs.

Grade descriptors in non-homogeneous deposits


Grades, however defined, are generally expressed as a per tonne
measure. This implicitly assumes that every tonne is the same in
all aspects other than grade. Often this is not the case. Where
different lithologies have different throughput rates, King (1999)
has identified such grade descriptors as cash flow grades. In
this case the real constraint on production is the annual available
hours in the treatment plant, not the ore tonnes per year. Despite
having similar metal grades, the dollars generated per mill hour
by slow-milling rock will be less than that for faster-milling rock.
More generally, the aim of the operation should be to
maximise the value processed through the production bottleneck.
The grade descriptor should therefore be a parameter correlated
with value (metal grade, dollar value or metal equivalent) per
unit of the production constraint. Selecting the highest value
material by this grade descriptor will maximise the value
generated. Net revenue per mill hour has been used as the grade
for a base and precious metals deposit in several rock types
having different hardnesses and milling rates, and also different
mineralogies, metallurgical recoveries and product qualities. In a
refractory gold deposit, the production constraint was the roaster
sulfur throughput: combining the sulfur burning rate and the
gold/sulfur ratio produced a gold grams per roaster hour grade.

Grade descriptors conclusions


An essential early part of any project evaluation should therefore
be the specification of the grade descriptor to be used. As
bottlenecks in the operation change over time, the grade
descriptor used may also need to change. Analyses can be
performed to identify how well various parameters represent the
true value. More importantly they indicate the extent to which
the reserve will be degraded by exclusion of higher- and
inclusion of lower-value material resulting from incorrect
ranking using inappropriate grade descriptors.

Project Evaluation Conference

Resource geologists work to reduce grade estimation errors.


Applying simplistic inappropriate grade descriptors can negate
these efforts and increase classification errors. At best there may
be a relatively small degradation of the ore reserve. At worst,
there may be a misapprehension of the spatial distribution of
value, resulting in value-destroying mining methods and
sequences. An example of this is described in case study 3 below.

CAPABILITIES AND LIMITATIONS OF


VARIOUS EVALUATION AND STRATEGY
OPTIMISATION METHODS
Project evaluations should be focused on achieving the companys
goals, which typically include maximising shareholder value,
often with associated goals of reducing costs and improving
efficiency. Strategy optimisation studies therefore typically have
maximisation of net present value (NPV) as a major goal. Other
goals may include maximising the internal rate of return (IRR),
profits, and other rate of return measures. Minimising unit costs
and maximising the annual outputs may be important. Risk
management goals of maximising the ability to reap upside
rewards, while minimising the danger of downside risks, may be
important, though often less easy to quantify.
Companies may therefore have multiple, often conflicting,
corporate goals, plus external government and social goals, such
as minimisation of greenhouse gas emissions, maximisation of
taxation revenues and provision of local employment and
infrastructure. To provide corporate decision-makers with
adequate strategic decision-making information, the evaluations
must identify not only the strategies that deliver each goal, but
also the trade-offs required to best achieve a combination of
conflicting goals. The following sections indicate how these
requirements have been handled in studies conducted by the
author and his colleagues in recent years. A more comprehensive
discussion is given by Hall and Stewart (2004).

Single operating scenario, simple sensitivity


analysis
The simplest, and perhaps the most common, form of evaluation
used in the industry might be described as a single operating
scenario model. Mining physicals are generated by engineers
using mine planning and scheduling software. These are
imported into a financial and economic evaluation model,
typically developed in a spreadsheet such as Microsoft Excel.
Product quantities and qualities may be generated by
metallurgical parameters included in this model, or may also be
derived externally and imported. Capital costs may also be
derived within the model, or in external physical and cost
schedules. Fixed and variable costs, sales terms and metal prices
and the like are applied to generate cash flows and accounting
information, and the various value measures are reported.
In the simplest analyses, separate models are created for each
alternative production scenario. Such things as timing of
expansion capital; duration of fixed costs, sustaining capital and
depreciation; and end-of-life revenues and expenses are
calculated only for the mine life in the imported schedules for
each case. Sensitivity analyses typically take the form of
applying factors to costs, prices/revenues, ore grades and the
like; the structure of the evaluation model does not permit a more
sophisticated analysis. Changes in reserve tonnages and
production rates can only be handled by generating separate
schedules externally and creating additional single-scenario
models. If changes are made to any of the model parameters or
logic, these must be updated in all copies of the model.
The next level of complexity adds model logic to handle
different mine lives. Items that are often modelled for a fixed life
as above must now allow for timing changes: fixed costs and
sustaining capital must be incurred for the mine life; expansion

Melbourne, Vic, 19 - 20 June 2007

55

B HALL

and upgrade capital and end-of-life items must move to


the appropriate time; associated changes in, for example,
metallurgical performance must change at the right time in
response; and depreciation calculations must take account of a
non-fixed mine life. Multiple scenarios can be included in the
one model, with scenario selection switches to allow
specification of the scenario to be evaluated and reported. Project
rigour is enhanced, since changes to data and model logic need
only be updated in one place, rather than multiple models.
Sensitivities can also be applied to such things as reserves
tonnages and production rates by simple factorisation, if the logic
is also upgraded to generate new mining schedules from the
imported production scenarios.
Sensitivity analyses from these types of evaluation are
frequently presented as a star or spider diagram, as illustrated
in Figure 2. Many of these plots are linear, such as for simple
factorisation of costs, prices and grades. Others, such as for
variations in reserve tonnage, are non-linear. The results are
usually as expected eg if the prices go up or the costs go down,
the NPV increases and if relatively small unfavourable
variations in key assumptions will lead to losses, these analyses
may assist in deciding how risky a project is, and whether it
should proceed or not. These analyses typically account
mathematically for only the immediate cash effects. They do not
cater for resulting enforced or chosen changes to physical
activities.
Care should be taken in making other decisions based on these
simplistic analyses. As an example, the author is aware of a case
where a major plant upgrade failed to deliver the expected
benefits. Later investigations suggested that the capital allowed
was set at the figure which generated the companys required
IRR as determined by a simple analysis of IRR versus capital
cost. This was substantially less than the cost developed by the
engineering design team, and resulted in major changes to the
overall design. While each plant component was individually
capable of the target performance, the overall system was not.
The simple sensitivity analysis only accounted mathematically
for the change in capital cost: all other cash flows were
unchanged. In reality there should have been a downturn to low
and negative IRRs at lower capital costs, due to the lower
production rates resulting from the reduced capital.

Hill of Value multiple operating scenarios


To overcome the shortcomings of the simple single scenario
evaluation model, the author and his colleagues apply what has

been called the Hill of Value (HoV) technique. This uses an


Excel model constructed to handle all the combinations of
various strategic decision variables, or value levers that can be
independently specified. To provide the exhaustive evaluation of
all combinations of strategies required in a prefeasibility
evaluation, these will typically include such things as:

cut-offs (for the grade descriptor used), for either the whole
mine, or for underground and open pit mines, or for various
orebodies, lenses, areas or stages of the mine(s);

mining methods, which may include different sizes of open


pits, and different methods or combinations of methods
underground, all of which may vary with the cut-offs used;

production rate targets or capacities, for all or parts of the


mining operation(s), and for the treatment plant(s);

timing of the implementation of debottlenecking upgrades in


the mines or treatment plants (which may include never);
and

alternative haulage/hoisting systems.


Factors with the potential to impact on optimum strategy, but
not under the control of the company, may include:

alternative economic forecasts,


varying degrees of exploration success, and
various workforce productivity and equipment efficiency
scenarios.
The evaluation models are constructed so that all of these
parameters can be selected or specified independently. Separate
and combined effects are catered for, and various scheduling
dependencies and constraints can be defined. The model logic
ensures that realistic mining and production schedules that
honour all inputs are generated. Mining and production
schedules are therefore generated within the evaluation model,
rather than externally as for a single scenario. Revenues, capital
and operating costs, and end-of-life cash flows are modelled by
flexible-life methodologies as above, and value measures
derived.
The Hill of Value technique and its application have been
described in detail elsewhere (Hall and de Vries, 2003).
It provides a clear picture of how a mine might change its
strategy to optimise a particular goal parameter (Figure 3) or the
trade-offs between various goals. The major variables in output
plots are those over which the company has control. They are

FIG 2 - Typical simple sensitivity analysis.

56

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

EVALUATING ALL THE ALTERNATIVES TO SELECT THE BEST PROJECT STRATEGY

effectively non-linear sensitivity plots showing how value is


related to decisions rather than to values of external parameters,
though the latter may also be included in the plots so that the
sensitivities of both value and decisions to external parameters
are also apparent.
Hall and Stewart (2004) and Horsley (2005) indicate in
particular how significant value gains can be obtained by varying
cut-offs both over time (which is generally well-known to be the
case), and also by mining area, which may at times run counter
to conventional wisdom.
Figure 4 illustrates how failure to evaluate all the options can
destroy value. Once a suboptimal strategy has been decided upon
and announced or implemented, the financial cost and/or the
damage to corporate image of making the change to what could
have been a better plan had it been identified earlier may be
insurmountable.
Sufficient engineering work must obviously be done external
to the evaluation model so that relationships appropriate to the
level of accuracy of the study can be incorporated to correctly
model the effects of changes in various value lever settings. If
these relationships are not well-defined, or values of some input
data items are not accurately known, alternative relationships or
ranges of values may be included as additional axes of the HoV.
This will then indicate whether those parameters have an impact

on the decisions to be made (in which case they may require


further investigation to be specified more accurately), or whether
they merely increase the likely range of results without affecting
the strategic decisions (in which case it may be reasonable to
accept the uncertainty and not incur the cost of extra work).
The power of the HoV methodology is its ability to generate
realistic and flexible models, and to demonstrate how the values
of various goal measures for the company and other stakeholders
may vary with changes in operating strategies. Its main drawback
is the rapid increase in the number of cases to be evaluated as the
number of strategy decision parameters, and the number of
options for each of these, increases. Brute force number
crunching of all options becomes impractical, and other
techniques become necessary, though a HoV model can still be
useful to identify the sensitivity of values and decisions around
an optimum point found by other means.

Genetic algorithms
Genetic algorithms (GAs) have been described as one of the best
techniques currently available for being reasonably sure of being
reasonably close to an optimum solution, when there is no
analytical method of finding the optimum, and when the number
of cases is too great to permit exhaustive evaluation to identify
the best.

FIG 3 - Finding and climbing the Hill of Value.

FIG 4 - Potential losses through failure to conduct a full analysis.

Project Evaluation Conference

Melbourne, Vic, 19 - 20 June 2007

57

B HALL

It is beyond the scope of this paper to describe the operation of


a GA in detail. It is a hill climbing technique, but to avoid the
problem of standard hill climbing methods (which will only
climb to the top of the hill on which they start), the GA will not
only climb the hill it is on, but also look around to see if there is
a higher hill elsewhere in the solution space (Anon, 2007a).
There is no guarantee that a GA will find a higher hill, even if it
does exist, nor that it will actually find the top of the hill that it is
on. However, the authors experience has been that, if a GA
analysis starts with the best results from a HoV analysis,
improvements in value of five to 15 per cent are not uncommon.
The author uses a relatively inexpensive GA add-in for
Excel. Value levers controlled by the GA have included:

cut-offs applied to different mining stages, and to pits and


underground mining areas;

sizes of pits being mined;


sequencing of mining various deposits;
allocation of deposits to multiple treatment plants; and
timing and size of plant upgrades.

A GA is relatively quick and simple to implement in a


spreadsheet model. It may show how improvements can be made
in existing strategies, or identify new strategies. Logs of the
analyses can be used to identify not only the best solution, but
also factors common to high value and low value cases, to guide
further planning.
Unlike a HoV analysis, a GA can only optimise for one value
parameter. However, different parameters can be optimised in
separate analyses. Conducting a reasonable number of
calculations may take a long time with a complex model.
Although there is no guarantee that the best possible solution has
been found, any gain is better than none.

Dynamic programming
Dynamic programming (DP) is not a single algorithm, but rather
an underlying philosophy of approach to solving a problem
(Anon, 2007b). There is therefore no general purpose DP
software; each class of problem requires its own formulation, but
many classes of mining strategy problem will have similar
processes, and mining-specific software that can be applied
across a range of sites and problem types is available.
As a general principle, DP reduces the number of options to be
evaluated by making sequential decisions, using the results of
one decision as an input to the next decision. Determining a
life-of-mine optimum cut-off policy using Lanes methodology
(Lane, 1988) is a DP process. A simple Lane-style analysis is
limited to a single production stream with a prespecified mining
sequence, and the spreadsheet software available on most
personal computers can apply it. However, the experience of the
author and his colleagues is that, while the principles behind
Lanes methodology help to explain the features of optimum
strategies found, the complexities of real problems preclude the
simple mechanical application of Lanes formulas, which is also
noted by Lane himself in his book.
Brett Kings Comet software applies DP to a number of
mining strategy optimisation problems, in particular extending
Lanes methodology to account for optimising the timing,
sequencing, cut-offs and production rates of multiple
production streams.

Linear and mixed integer programming

CASE STUDIES
In evaluations conducted by the author and his colleagues,
increases in project NPV ranging from ten to 50 per cent of the
value of the accepted strategy at the start of the project are
typically generated by fully evaluating all the options available
and identifying better strategies. Case studies described below
illustrate the types of evaluations and strategy changes
encountered. Results have been normalised to show the value of
the base case or approved strategy at the start of the analysis as
an index value of 100, and values for other options relative
to this.

Case study 1 underground gold cut-off


optimisation
The client operated a gold mine where underground mining had
recently commenced below the exhausted open pit. The orebody
extended to over 1000 m below surface, and two mining methods
were used depending on its thickness and dip in various
locations. The mining plan called for progressively deepening
the mine by a decline access, using underground mine trucks to
haul the ore to the stockpile at the treatment plant on surface.
A number of major production options were evaluated over the
course of the study:

Shaft hoisting was proposed to replace the increasingly

Linear programming (LP) and its derivatives such as mixed


integer programming (MIP) are classical analytical techniques
for maximising or minimising an objective function subject to a
number of constraints. They apply to a number of common
mining industry problems.

58

The author has used an inexpensive LP add-in for Excel for


a strategy optimisation study where value maximisation had to
account for several quality constraints on both the ore feed and
the product, as well as plant throughput limitations. Colleagues
use AMC Consultants in-house life of business optimisation
system (LOBOS), which is based on a high-end commercial
LP/MIP software package, for complex mining-related
scheduling, blending and strategy optimisation problems.
LP can solve problems with hundreds of thousands of
variables and constraints, but cannot account for discrete
variables needed in detailed schedules. MIP can be used for
difficult scheduling and sequencing problems, but is more
limited in terms of problem size. They can solve for all time
periods simultaneously, so that decisions later in the mine life are
not constrained by earlier decisions, as is often the case in HoV
and DP models. They will find a provable optimum. LP is used
with convex surfaces, ie with only one hill and no valleys or
discontinuities in the value versus input variables surface.
Efficient algorithms will rapidly find the optimum point. MIP
can be used with hills, valleys and discontinuities in the
value versus input variables surface. The algorithms prevent
time-consuming calculations being done for regions where the
best solution is unlikely to be found.
Solution of complex problems can be time-consuming with
MIP. Algorithms are continually improving, and increasing
processing speeds are improving the likelihood of successfully
finding the optimum strategy in a reasonable time. Both LP and
MIP require strict mathematical formulation, and the
simplifications required may introduce inaccuracies. The use of
MIP requires an experienced user who understands both the
problem to be solved and modelling and solution methodologies.
Results can be generated within the LP/MIP program, or
physicals fed into an external spreadsheet model as for the single
scenario model described above. Optimum strategies from
constrained LP/MIP models have been used to guide
implementation of improved logic in HoV spreadsheet models,
which are more flexible than MIP in some aspects, but less
powerful in others.

expensive truck haulage as the mine got deeper, but there was
insufficient reserve for an ongoing reduction in ore handling
operating costs to pay for the capital outlay. The analysis
indicated the reserve tonnage needed to justify sinking a
shaft.

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

EVALUATING ALL THE ALTERNATIVES TO SELECT THE BEST PROJECT STRATEGY

Several stages of upgrades in the treatment plant were


proposed. At projected gold prices, only the first of these was
justifiable. The second stage was only justifiable at higher
than expected gold prices, where increased gold production
and a higher price together were sufficient to justify the
capital outlay.

Several stockpiles of lower-grade material remained from the


open pit mining. At the expected gold prices, fresh
underground ore (using the optimum cut-off) and reclaimed
stockpile ore had similar cash margins; the value of the
operation was therefore relatively insensitive to the
underground production rate. However, increases in the gold
price or different mine head grades resulting from
suboptimal cut-offs disturbed this balance and increased the
impact of the underground production rate on value.
The major strategic decision ultimately involved the cut-off
policy. Like many operating mines, the existing life-of-mine
(LoM) plan used a single mine-wide cut-off for the whole mine
life. Significant value was added by increasing the mine-wide
cut-off, as illustrated in Figure 5. Further value was obtained by
specifying separate cut-offs for each mine area. Cut-offs used in
this analysis were at 1.0 g/t increments, which may be adequate
to identify the shape of the HoV and an optimum mine-wide
cut-off, but are probably too coarse to fully optimise individual
mining area or time-based cut-offs. The additional value shown
for GA-optimised area cut-offs in Figure 5 is therefore believed
to be conservative.

Case study 2 underground gold production


rate and cut-off optimisation
The client had acquired a resource that had been worked as an
open pit operation for some years in a developing country. There
were two adjacent multi-lens deposits, and the title to one was
less secure than the other. An underground resource had been
identified, and a prefeasibility level study had been conducted,
but only for both deposits, at one production rate and one cut-off
grade.
An initial high-level review of this study suggested there could
be significant upside from different operating strategies, and a
more detailed study was commissioned. This was a desktop
study only, involving reworking of the PFS data and the
construction of a flexible but robust evaluation model. The
investigation involved:

definition of ore reserves at various cut-offs;


estimation of development requirements at each cut-off;
specification of ranges of the mining schedule drivers to be
evaluated:

maximum decline advance rate;


initial stock of reserves developed before full production;
ongoing development advance rate;
maximum ore lens production rates;
overall mine production target;
sublevel interval; and
production strategy equal priority for all exposed ore
sources above cut-off, or highest grade sources targeted
first (the latter requires a faster overall development rate
than the former, but has the potential to deliver higher
grades to more than offset the increased costs and thus
add value);

construction of a model to:


schedule mine development and production, accounting
for the controlling parameters above;

derive other mining and production physical quantities;


derive cash flows:
revenue,
operating and capital costs, and
taxes and government charges;
derive and report NPVs, IRR, returns to the government,
etc.
The modelling identified a number of characteristics of the
project, including different optimum sequencing strategies
depending on the relativities of various mining rate parameters,
such as decline advance rate and lens and total mine production
capabilities, and whether the less secure deposit was mined or
not. Figure 6 shows a Hill of Value for NPV as a function of
cut-off and production rate target. The value falls at high cut-offs
due to a reduction in reserve, but is relatively flat at low cut-offs
due to the block model being simplified by having low-grade
regions below the resource cut-off excised, reducing the variation
of tonnage with cut-off. Value generally continues to rise with
increasing production rate target, as upper limits to mining rates
were not well established at this stage of the evaluation.
However, high level analysis suggested that the original PFS rate
of 1.2 Mtpa could reasonably be increased to approximately
1.5 Mtpa.
Table 2 shows the relative gains due to changing the
production target, cut-off and production strategy. The results
from this study were used to set design parameters for further
PFS studies.

FIG 5 - Case study 1 value increase from cut-off changes.

Project Evaluation Conference

Melbourne, Vic, 19 - 20 June 2007

59

B HALL

FIG 6 - Case study 2 value increases from cut-off and production target changes.

TABLE 2
Case study 2 indexed NPVs for various combinations of
operating strategies.
Equal priority
Cut-off

Targeting high grade

4 g/t Au

6 g/t Au

4 g/t Au

6 g/t Au

1.2 Mtpa target

100

127

116

137

1.5 Mtpa target

127

147

139

152

Case study 3 underground base metals grade


descriptor, mining method, production rate and
cut-off optimisation
The client operated an underground lead-zinc mine which was
not generating good returns. Historically the mine had been
developed as a sequence of mining blocks approximately 100 m
high, with deeper blocks opened up as blocks closer to surface
were depleted. There were known subvertical trends in grades,
but the orebodies defined using the existing metal equivalence
formula and cut-offs were such that the grades seen within the
mineralised zone were relatively homogeneous along strike. The
bulk of the mineralisation was mined out using open stoping with
backfill to facilitate close to total extraction, with only a few
lower grade pillars being left in situ between the major lenses.
The mining sequence was therefore predominantly along strike,
with a subsidiary advance from top to bottom. Recovery of the
crown pillars created between lifts presented technical
challenges, with both lower productivity and reduced ore
recovery in those areas.
The study first evaluated the effects of the grade descriptor
used. Figure 7 shows how the true value generated by a block of
rock described by the original metal equivalent value could vary
significantly: the range was equivalent to the unit cost of mining.
A revised equivalence formula with a much lower variation in
true value was adopted.
When the mineralisation was viewed using the new grade
descriptor, the vertical zonation, particularly at higher cut-offs,
became more pronounced. An obvious change to the mining
method was to develop immediately to the bottom of the deposit
and establish stopes extending over the vertical height of the
remaining orebody, with mining in a predominantly bottomto-top sequence within lenses, and a subsidiary advance along

60

FIG 7 - Case study 3 changes in spread of true values with


different grade descriptors.

strike. This permitted both the elimination of crown pillars, and


the establishment of extra vertical rib pillars between stopes in
lower value material, allowing a significant reduction in the
amount of cemented fill required. The overall value of the
resource was enhanced by a reduction in the misclassification of
ore and waste, which in turn resulted in a more apparent
relationship between value and cut-off. Figure 8 shows value
versus cut-off for both the original equivalence formula and
mining method, and the proposed equivalence formula and
mining method. The values for the latter include the effect of
earlier capital development to drive immediately to the bottom of
the mine. Intermediate scenarios original equivalence formula
with proposed mining method, and proposed equivalence
formula with original mining method were also evaluated in
the study, but the complete change was found to be the best
strategy.
The production rate also had a significant influence on value.
Although the mines target was nominally the same as the
concentrator capacity, and the fleet size and workforce numbers
were in place for this, the mine was in practice delivering only
some 85 per cent of that target. It was identified that there were
inadequate stock levels of developed and drilled ore, exacerbated
by the reliance on a small number of large stopes in operation at
any time. The generalised Hill of Value shown in Figure 3 is
derived from this study, and indicates the value gained simply by

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

EVALUATING ALL THE ALTERNATIVES TO SELECT THE BEST PROJECT STRATEGY

FIG 8 - Case study 3 value increase from grade descriptor, cut-off and mining method changes.

producing at the target rate. This is achieved by injecting


working capital to increase ore stocks, and produce from a larger
number of smaller stopes. The change in mining method also
facilitated this change. Additional capital investment to upgrade
the capabilities of the rock handling systems was thereby
avoided.
The client elected to change the mining method and to
increase the cut-off substantially towards the identified optimum,
so as to realise the bulk of the identified potential gain, while at
the same time restricting the reduction in ore reserves to a level
where it believed the market would not react adversely and
perversely drive the value of the shares down despite the increase
in value generated for shareholders.

Case study 4 open pit base metals mining


rate and cut-off optimisation
The client operated a large base metal open pit mine. Mine staff
believed that pressure from the corporate head office to reduce
costs was resulting in a reduction in waste stripping and hence of
overall mining rates and exposure of ore above cut-off, leading to
loss of value. The study evaluated only the effects of changing
the run-of-mine (ROM) cut-off and the mining rate. The existing
ultimate pit limits and sequencing of mining over the LoM, and
the planned sequence of improvements and upgrades in the
concentrator, as defined in the approved plan, were retained.
Different rock types had different milling rates, and
combinations of mineralogies of both ore and waste minerals had
complex effects on recovery in the concentrator and on product
quality, which impacted on the price received. Net revenue per
mill hour was identified as a better grade descriptor than a simple
metal grade per tonne for ranking potential ore sources. Some
two dozen rock types were identified as having potential
influences on value. Blending of various ore types was therefore
seen during initial investigations to have potential to impact on
the value optimisation. For various reasons, blend optimisation
was not fully investigated in the main part of the evaluation,
although the recovery and quality effects were accounted for in
the evaluation model developed. The model also allowed
specification of which of the proposed concentrator upgrades
were done, which enabled identifying the optimum mining plans
for each concentrator option and hence the real value gained
from each upgrade.
Figure 9 shows the results of the study, which corroborated the
project sponsors views. Faster mining (of total rock) allowed

Project Evaluation Conference

FIG 9 - Case study 4 effects of changing total mining rate and


ROM cut-off in an open pit.

exposure of more mineralisation, feeding higher grades to the


mill in the short-term, while stockpiling the remainder for later
treatment. This adds value while the additional product pays for
the extra mining costs. Beyond this point, further mining rate
increases destroy value. Further value gains could have been
obtained by fully investigating the size of the ultimate pit,
sequencing and blending and alternative mill upgrade scenarios.
The author understands that the final decision was to save
short-term costs by reducing the mining rate, which Figure 9
shows destroys more value than the costs saved.

Case study 5 mineral sands optimising


deposit sequencing, mining selectivity and
number of primary treatment plants
The client owned a number of mineral sands deposits containing
various mixtures of heavy minerals. Each deposit typically
consisted of a strand line with spatial distributions of minerals
differing according to the heavy mineral (HM) composition and

Melbourne, Vic, 19 - 20 June 2007

61

B HALL

deposition mechanics. The values of component HMs varied


significantly, depending on prices and quality of the potential
products.
The companys policy had been to target mining and treating
of all mineralised material above a specified total HM content.
The company owned or planned to acquire a number of wet
concentrators to separate the HM from the sand mass at or
adjacent to the mining sites, with a further option to add
magnetic separators to remove relatively low-value ilmenite at
the mine. HM concentrates, with or without ilmenite, were
transported from each mine site to a single dry mill, which
separated the various heavy minerals from each other into
various products.
The company wished to identify the optimum mining and
production strategy, including such things as:

the number of wet concentrators;


the sequence of mining the deposits, accounting for
allocation of wet concentrators to deposits, and the cost of
moving concentrators between deposits, which also
depended on the sequence; and

the best plan for mining each deposit, including how much of
the mineralisation to mine, how much over-burden to prestrip
(where applicable), and whether to separate the ilmenite and
discard it as waste at the mine site.
As in case studies 3 and 4, the total HM grade descriptor was
incapable of discriminating between higher- and lower-value
portions of the deposits. A $/tonne grade descriptor was therefore
used, calculated as the net revenue after deduction of wet
concentration, transport, separation and realisation costs.
Each deposit was initially analysed individually to identify
its inherent value versus operating strategies relationships.
Whittle-4X was used to derive nested pit shells extracting
various tonnages of mineralisation. These shells tended, with
increasing size, to go rapidly to the base of the highest value
parts of the strand, initially in isolated pits, then to extend along
strike to form longer continuous pits, and across strike to include
extensions into lower-value material. Simple HoVs were derived
for each deposit, with the main value levers being the pit shell
number and the dollar value cut-off. This analysis accounted for
wet concentrator moves along the length of the deposit, and
identified whether isolated pits were better mined separately, or
with the intervening low-grade material mined to connect them.
In some deposits value was relatively insensitive to either

variable over a range of settings, but reduced significantly if


either were set outside this range. In other deposits, values were
relatively insensitive to one parameter while highly sensitive to
the other, and vice versa. The companys designs were in some
cases close to the optimum, and in others, significantly different.
While the companys ranking of deposits by traditional planning
methods was generally similar to the study rankings, there were
some differences.
An evaluation model was constructed to permit specification
of a number of strategic decision parameters for all deposits. The
number of wet concentrators was able to be varied. Each deposit,
as well as having its pit size and cut-off variable as above, was
able to be assigned to any concentrator and in any sequence, or
not mined at all. A genetic algorithm (RiskOptimizer)
controlled the settings of all these strategic decision variables,
and was run to maximise overall NPV for all deposits.
It was found that deposits mined earlier in the sequence tended
to be worked at smaller pit sizes and ore tonnages than their
standalone optimum sizes. This brings forward the value of later
deposits, so that the loss of value for an individual early deposit
is more than compensated for by the increase in NPV from
bringing forward the values of later deposits in the sequence.
This is shown schematically in Figure 10. However, shorter mine
lives in the early deposits meant that the company had to commit
to bring other deposits into production earlier, with associated
permitting and infrastructure development issues to be addressed
sooner than had been anticipated.
Significant increases in value were obtained by the strategy
optimisation study, as shown in Figure 11. It was also found that
an extra wet concentrator and earlier upgrades of the dry plant
were justified if product prices were higher, the extra capital cost
being repaid by bringing forward the net revenue from the
operating mines. Again, this would have implications for
permitting and infrastructure development, and for exploration
programs.

Case study 6 underground gold justification


of shaft haulage, optimisation of timing
This case study is for a hypothetical situation with a steeply
dipping orebody, and has been reported in more detail by Hall
(2005). It is based on the analysis conducted for evaluating the
shaft versus truck haulage option noted in case study 1. Although
specifically evaluating shaft versus trucks, the methodology can
be applied to any capital investment to be justified by reducing
operating costs.

FIG 10 - Case study 5 optimum cut-off for a deposit standalone and in a mining sequence.

62

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

EVALUATING ALL THE ALTERNATIVES TO SELECT THE BEST PROJECT STRATEGY

The hypothetical orebody has a single production front moving


vertically downwards, with truck haulage to surface. A shaft to
1000 m depth, with trucks hauling to a tip above the skip loading
station, has been proposed. Proved and Probable Reserves extend
to 900 m depth, and Inferred Resources for some distance
beyond that. The deposit is still open at depth. The analysis has
projected the known orebody characteristics to 2000 m depth to
assess the strategic options available.
Figure 12 shows the effect of deferring the shaft sinking
decision. The trucking only curve shows a maximum value at a
little over 1000 m depth, which is therefore the economic limit of
trucking. A line $5 M above this indicates a safety margin that
a shaft must have to be that much better than trucking only. With
a 1000 m deep shaft, the economic depth of the mine is about
1550 - 1575 m. The start of shaft hoisting is expressed as the
elevation of the production front at that time. Shaft hoisting lines
are plotted for its starting at 100 m intervals vertically,
representing time increments of 2.5 years with the production
assumptions made.

Deferring the shaft option results in decreasing value, as the


later the start, the less ore is hoisted at the lower shaft operating
costs. If the mineralisation extends to at least the economic depth
of say 1550 m, then the shaft option will no longer deliver the
$5 M margin over trucking if the start of shaft hoisting is
deferred beyond when the production front is at about 1030 m.
The value of the shaft falls below the maximum value with
trucking only, without the margin, if the start of hoisting is
deferred beyond when the production front is at about 1140 m. If
the mineralisation is closed off at a shallower depth, earlier shaft
starting times are needed to make the shaft option viable.
If analyses were done at one final depth of the mine only
(typically the known bottom of mineralisation at the time of the
evaluation, and often with a downgrading factor applied to
Inferred Resources), a value-destroying decision could be made.
This is similar to the situation in Figure 4, for analyses done at
one cut-off. Although the value of the shaft is greater than the
value of trucking for most final depths below the shaft starting
depth, the shaft is not always the best option.

FIG 11 - Case study 5 value increases from strategy changes.

FIG 12 - Case study 6 NPV for trucking only and for different shaft starting times.

Project Evaluation Conference

Melbourne, Vic, 19 - 20 June 2007

63

B HALL

Figure 12 also shows that deferral of the shaft results in a


range of uneconomic final depths of mining. For shaft hoisting
starting when the production front is at 1000 m depth,
mineralisation extending below the economic limit of trucking
(at approximately 1030 m depth) is uneconomic unless it extends
below 1150 m, at which point the shaft curve exceeds the
maximum trucking-only value. If the shaft is deferred to start
when the production front is at 1100 m, the uneconomic final
depth extends to approximately 1230 m. If mineralisation
extends only to these depths, its value will be lost if the shaft is
deferred too long the mine should truck only and close at the
economic limit of 1030 m depth. But if the shaft is built early
enough, the mine will be economic for any depth of
mineralisation, down to the economic depth with shaft hoisting
of 1550 m, and the value will be significantly greater than with
trucking only.
Hall (2005) also describes how these processes can be
extended to identify the reserves required to justify the proposed
shaft, and hence the exploration program needed to prove their
existence in time for a study to be conducted.

CONCLUSIONS
Numerous strategic plan optimisation case studies show that
conventional wisdom is often far from wise. Typical industry
practices for determining operating strategies, particularly
cut-offs, and the grade descriptors with which they are
associated, result in mine plans that are often far from optimal.
Significant extra value can be obtained from many operations,
often by changing the grade descriptor to identify more
accurately where true value lies in the deposits, and by
increasing the cut-off and thereby reducing the reserve and the
mine life. Sequencing of multiple deposits (or parts of large
deposits) can also have an influence, and the optimum strategy
for a deposit mined early in a sequence will not be the same as if
it were mined in isolation.
The earlier that optimum strategies are identified, the more
likely that their value can be realised. Changing strategic
direction may be politically, technically or financial unviable
once suboptimal strategies have been locked in, either by public
announcements that cannot be revoked without losing market
trust, or by physical commitments to such things as mining
methods and plant capacities. Although there is an
understandable desire to reduce the duration and cost of project
evaluations, failure to allow for the time and cost of a
wide-ranging study that rigorously evaluates all the options,
particularly at the prefeasibility study stage, carries a high risk of
destroying significantly more value than the cost saved.

64

ACKNOWLEDGEMENTS
The author wishes to thank the management of AMC
Consultants Pty Ltd for permission to prepare and present this
paper, and the secretarial staff for assistance with its preparation.
Past and present members of the AMC Mine Optimisation and
Business Improvement Groups are thanked for their input into
this paper, both directly and through their contributions to the
development of the overall concepts.

REFERENCES
Anon, 2007a. Genetic algorithm [online], Wikipedia. Available from:
<http://en.wikipedia.org/wiki/Genetic_algorithm> [Accessed: 10
April 2007].
Anon, 2007b. Dynamic programming [online], Wikipedia. Available
from: <http://en.wikipedia.org/wiki/Dynamic_programming>
[Accessed: 10 April 2007].
Hall, B E, 2005. A quantitative assessment of the factors influencing the
shaft versus trucks decision, in Proceedings Hoist and Haul 2005,
pp 53-63 (The Australasian Institute of Mining and Metallurgy:
Melbourne).
Hall, B E and deVries, J C, 2003. Quantifying the economic risk of
suboptimal mine plans and strategies, in Proceedings Mining Risk
Management Conference 2003, pp 59-69 (The Australasian Institute
of Mining and Metallurgy: Melbourne).
Hall, B E and Stewart, C A, 2004. Optimising the strategic mine plan
methodologies, findings, successes and failures, in Proceedings
Orebody Modelling and Strategic Mine Planning Conference,
pp 49-58 (The Australasian Institute of Mining and Metallurgy:
Melbourne).
Horsley, T P, 2005. Differential cut-off grades, in Proceedings 9th
AusIMM Underground Operators Conference, pp 103-109 (The
Australasian Institute of Mining and Metallurgy: Melbourne).
King, B M, 1999. Cashflow grades scheduling rocks with different
throughput characteristics, in Proceedings Strategic Mine Planning
Conference (Whittle Programming Ltd: Melbourne).
Lane, K F, 1988. The Economic Definition of Ore: Cut-Off Grades in
Theory and Practice (Mining Journal Books Ltd: London).
McCarthy, P L, 1998. Objectives of feasibility studies [online], AMC
reference library. Available from: <http://www.amcconsultants.com.
au/library/browse.asp> [Accessed: 5 March 2007].
Noort, D J and Adams, C, 2006. Effective mining project management
systems, in Proceedings International Mine Management
Conference, pp 87-96 (The Australasian Institute of Mining and
Metallurgy: Melbourne).
Ward, D J and McCarthy, P L, 1999. Startup performance of new base
metal projects, presented to Adding Value to the Carpentaria Mineral
Province Conference, Mt Isa [online]. Available from: <http://www.
amcconsultants.com.au/library/browse.asp> [Accessed: 5 March
2007].

Melbourne, Vic, 19 - 20 June 2007

Project Evaluation Conference

You might also like