Ninth International Mining Geology 2014 Conference

Paper Number: 37.00

Optimum Dig Lines for Open Pit Grade Control
Isaaks, E, Ph.D., MausIMM, Geostatistical Consultant, Isaaks & Co., 1042 Wilmington Way,
Emerald Hills, CA, USA, 94062, Email [email protected]
Treloar, I, Manager Mine Technical Services, Newmont Mining Corporation, Newmont
Boddington Gold, Boddington, WA, 6390, Email [email protected]
Elenbaas, T, P.E., Consulting Geostatistician, Newmont Mining Corporation, Newmont
Boddington Gold, Boddington, WA, 6390, Email [email protected]
ABSTRACT
Critical reviews of grade control generally focus on blast hole sampling and the estimation of ore
control block model (OCM) grades with little or no attention given to dig line design. For
example, the estimation of block grades is deemed important as estimation errors lead to the
misclassification of ore types and subsequent dollar loss. Recognition of the problem that under
estimation of block grades may be more or less costly than over estimation led to the application
of loss functions to the classification of individual OCM blocks by ore type. The method is
known as maximum profit selection or MPS. However, each OCM block may be misclassified a
second time by dig line design. And although each block may initially be assigned an optimum
ore type by MPS, OCM blocks are seldom individually minable by ore type. As a result, a
common solution to this problem is the design of minable polygons or dig lines loosely
constrained by a minimum mining width. All material within a properly designed dig line
polygon can be mined and sent to a single destination such as the mill, stockpile or waste dump.
These dig lines are often designed manually or by simple variations of computer generated
contour lines. They are generally suboptimal and result in additional misclassifications of ore
types. This paper provides an algorithm for constrained optimum dig line design where dig line
misclassifications are also evaluated through loss functions or MPS. Constrained optimum dig
line designs minimize the dollars lost by dig line misclassifications for a given minimum mining
width constraint. A case study is provided illustrating constrained optimum dig line design and
subsequent benefits.
INTRODUCTION
Open pit grade control generally entails the sampling and assaying of blast hole cuttings
followed by the estimation of ore control block model (OCM) grades. The estimated OCM block
grades are then used in turn to design surface polygons or dig lines that outline and separate
various ore types and waste material for the purposes of mining. Unfortunately, the estimation of
block grades and the design of dig lines often results in the misclassification of ore types. For
example:

True block grades are not known and must therefore, be estimated. If the classification of
ore type based on the true but unknown block grade is different from the ore type based
on the estimated block grade, then the ore type of the block is misclassified.

The locations of the true ore type contacts are not known and therefore their locations
must also be estimated. Dig lines must be designed so that the contained ore type can
actually be mined as designed by project mining equipment. If the dig line classification
of ore type is different from the ore type classification based on block grade, then the
block is misclassified. For example, a dig line containing a majority of blocks classified
by grade as mill ore may include a small proportion of blocks classified by grade as
waste. The waste blocks are misclassified by the dig line and as a result, will be sent to
the mill.

Misclassification errors reduce revenue. They do not cancel one another. They simply
accumulate. Interestingly, the misclassification of OCM blocks resulting from the estimation of
block grades has received considerable attention. For example, several authors have shown that
the classification of ore types based on the best1 estimates are sub-optimal (Schofield, 1997;
Srivastava, 1987; Deutsch, Magri, and Norrena, 2000). The following example illustrates this
apparent paradox:

Consider a truck that has just been loaded with muck pile material and the mill and waste
dump are the only two possible destinations.
If the truck is sent to the mill, and the true grade of the truckload is below the waste/mill
cutoff grade, a relatively small dollar loss will occur because although the truckload is
waste, it does contain some metal that will be recovered which will offset processing
costs.
If the truck is sent to the dump, but the true grade of the truckload is greater than the
waste/mill cutoff grade, the full potential dollar value of the truck load of ore is lost
nothing is recovered.

In other words, the loss resulting from an underestimation of grade and misclassification of ore
type may be greater than the loss resulting from an overestimation of grade and the ensuing
misclassification of ore type. Srivastava (1987) points out that minimum error variance
estimates1 such as kriged estimates may not be the best for grade control because minimum error
variance estimates are calculated by assuming the loss resulting from over and underestimation is
symmetrical. But, as the example above illustrates, the losses due to underestimation may be far
greater than those due to overestimation.
One solution to the problem of asymmetric losses consists of assigning loss functions to
estimation errors. Equation 1 and Figure 1 provide an example of an asymmetric loss function
where the loss due to underestimation is greater than the loss due to overestimation.
*

Z Z

*
c2 * Z Z

Z Z

L(Z , Z)

c1 * Z* Z

11\* MERGEFORMAT ()

1 Geostatistical estimators that minimize the variance of the distribution of estimation errors are often
referred to as B.L.U.E estimators or best linear unbiased estimators.

where

Z are Z*

are the true and estimated grades,

constants. The estimation error

L(.) is the loss function, c1 and c2 are

e is calculated as e Z * Z .

Figure 1: An example of an asymmetric linear loss function where

c 198 and c 66
1

The estimate minimizing the expected loss for linear loss functions is the quantile of
corresponds to the ratio

c1 c1 c2

Z * that

. For the example in Figure 1 the estimate that minimizes the

expected loss is the 0.75 quantile of the probability distribution of

Z * . The probability

distribution of Z can be obtained by conditional simulation methods (Davis, 1987; Deutsch,

Magri, and Norrena, 2000; Isaaks, 1990). In all likelihood, the 0.75 quantile of the distribution of

Z * is greater than the mean. This would tend to increase the number of blocks classified above
cutoff grade which would in turn increase the profit given that underestimation is costlier than
overestimation. If this seems counter intuitive, consider a block whose kriged estimate is slightly
less than the waste/ore cutoff grade. The kriged estimate would classify the block as waste,
*

however, the 0.75 quantile of the probability distribution of Z may be above the cutoff grade,
which would render the block as ore and reduce the dollar loss.
The ore type classification of OCM block grades by using loss functions of the block grades
rather than by the block grades themselves is sometimes called MPS or maximum profit
selection. Misclassification errors made by assigning an ore type to a block based on either the
block grade or loss functions of the block grade are referred to as grade misclassifications.
The efficacy of MPS is frequently demonstrated by calculations based on free selection
(Schofield, 1997; Deutsch, Magri, and Norrena, 2000; Dagdelen and Coskun, 1999). Free
selection presupposes that each OCM block is equivalent to a selective mining unit (SMU) and
each SMU can be mined and sent to the correct destination according to its classified ore type. In
other words, none of the SMU will be misclassified or sent to the wrong destination at the time
of mining. This presumption is not true in practice. The spatial configurations of SMU blocks

with the same ore type are often very complex, rendering the mining of an individual SMU
impractical. Moreover, many OCM models consist of blocks much smaller than an individual
SMU.
A practical method for the layout of minable ore types is the design of dig lines constrained by a
minimum mining width (MMW). In fact, most open pit mining operations have a designated ore
control engineer or geologist who is tasked with the design of dig lines for each blast.
However, the design of dig lines is typically charged with additional misclassification errors.
Figure 2 shows an example of an ore control block model overlain with dig lines. The various
ore types are symbolized by different shades of grey. The two arrows illustrate the MMW.
Admittedly, the blocks are smaller than the MMW or SMU, however, the misclassification of
many OCM blocks by dig line design is obvious. These misclassification errors are referred to as
dig line misclassification errors as opposed to grade misclassification errors.

Figure 2: An example of an ore control block model. The white lines represent dig lines constrained by a minimum
mining width (MMW). The various shades of grey indicate very high grade (VHG), high grade (HG), medium grade
(MG, and low grade (LG) ore types.

Experience suggests that the revenue lost by dig line misclassification errors is far greater than
that lost through grade estimation errors. First, there is considerable uncertainty associated with
the actual location of the ore/waste contacts. Second, this uncertainty is compounded by the
imposition of a MMW constraint on the dig line design. Thus, a significant number of OCM
blocks may be misclassified by dig line design.
A recent literature search on dig line design fails to identify references that describe methods for
open pit dig line design constrained by a MMW. References are found for dig line design, but
without a MMW constraint (Norrena and Deutsch, 2002; Srivastava, Hartzell and Davis, 1992).
Rendu, (1982) discusses the application of MMWs to the mining of veins, but the discussion
pertains to underground mining methods and does not describe an algorithm for open pit
constrained dig line design.
OPTIMUM DIG LINE DESIGN
Isaaks & Co. developed an algorithm and computer program (Digger) for the design of dig lines
constrained by a MMW. The algorithm is revenue based and minimizes dollars lost due to dig
line misclassifications through loss functions. The general problem is illustrated by Figure 3.

Figure 3: This figure illustrates the minimum mining width problem. All five blocks will be mined together and sent
to either the waste dump, the leach pad, or the mill. The correct destination is the one that minimizes the dollars lost
due to the misclassification of ore types, thereby maximizing net revenue.

The Digger algorithm resolves the correct destination problem by application of loss functions
to the blocks within the MMW. For example, the potential revenue of each OCM block is
calculated for each ore type using equation (2). Loss functions can then be evaluated for each
block and for each possible destination. The correct destination is the one where the combined
dollar loss of all 5 blocks is minimum.

Pi $* Z * Ri Bi , i 1, n ore types

22\* MERGEFORMAT ()

where

P = net revenue in dollars

$ = metal price
Z = block grade
R = metallurgical recovery rate
B = break even cost in dollars
Zc = cutoff grade

Cutoff grades are almost always pre-defined by mining personnel, so the calculation of
corresponding break even costs can be calculated using Equations (3). Note that each ore type is
featured by a unique recovery function,

Ri , i 1,n ore types.

B1 $* Zc1 * R1
Bi 1 $* Zci 1 ( Ri 1 Ri ) Bi i 1, n 1 ore types

33\* MERGEFORMAT

()
An example of the loss functions employed by Digger is shown in Figure 4. The superscript * on
the symbols M, S, and W indicate mill, stockpile, and waste ore types that are assigned to a block

by the encompassing dig line while M, S, and W without a superscript * symbolize the ore types
assigned to the block by the OCM block grade. The symbols within the table cells are the
calculated dollars lost (using equation 2) as a result of dig line misclassifications. For example,
the table cell containing

PM

conveys an OCM block that is waste (W) by its block grade, but
*

the block is within a mill dig line

(M )

where it will be mined as mill ore and sent to the mill.

The symbol M is the dollar loss which is calculated using mill recoveries and break even costs.
Because the block is waste grade, it contains insufficient metal to pay for the mill processing
costs and will be processed at a dollar loss.

PM

PM PS

PS

PS PM

PS

PM

W*

S*

M*

Figure 4: This figure tabulates the loss functions for 3 ore types, waste (W), stockpile material (S), and mill
material (M). The symbols in the table cells are the calculated dollars lost (using equation 2) as a result of dig line
misclassifications. See text for additional definitions.

The application of loss functions to groups of OCM blocks defined by a MMW is the key to
Digger. Digger initially defines a mosaic of SMU type dig lines. Based on simulated annealing,
Digger repeatedly iterates through the OCM testing the existing dig line locations. If a new dig
line location can be found that reduces loss and does not violate the MMW, then the dig line is
re-located to the new location. Digger continues iterating through the OCM until either no new
dig line locations can be found or until a pre-defined number of iterations is reached.
CASE STUDY
The efficacy of constrained optimum dig line design is easily evaluated by comparing potential
Digger revenues to those generated by other methods given an identical OCM. A simple
summation of the ore type tonnes and grade generated by each set of dig lines is all that is
needed. Differences between tonnes, grade, and revenue can only be due to differences between
dig line designs.
In addition to providing an example comparing manually designed dig lines to optimum
computer dig line designs, the case study also provides examples that focus on the impact of
varying the MMW. Generally, once the MMW is defined it is seldom altered by grade control
operations. Thus, the benefit of evaluating potential recoverable tonnes and grade by ore type as
a function of MMW is typically not known. The MMW examples clearly illustrate the benefits of
being able to rapidly evaluate potential dig line recoveries for various MMWs.

Although this study does not account for the uncertainty associated with the estimation of block
grades, any accounting such as MPS would likely have very little impact on dig line
misclassifications. Increases in net revenue as a consequence of assigning initial ore types to the
OCM blocks by MPS would simply be additional to revenue increases as a consequence of
optimum dig line design.
The examples provided are based on actual mine data. However all grades, costs, and recoveries
have been altered to preserve confidentiality. The case study considers data from a small open pit
gold mine where oxide, transition and sulfide ore types exist. Ore type destinations are waste,
leach pad, and mill.
Case study data
The OCM under study consists of 4,183 blocks, each measuring 2 m x 2 m x bench height. The
OCM blocks are classified into 4 ore types plus some minor waste based on the oxide, transition,
and sulfide classification rules described below.
Ore types

Sulfide mill (SuM)

Transition mill (TrM)
Oxide mill
(OxM)
Oxide leach
(OxL)
Waste
(W)

Processing costs

Sulfide mill
$ 30.00/tonne
Transition mill $ 15.10/tonne
Oxide mill
$ 14.95/tonne
Oxide leach
$ 3.15/tonne

Metal prices

Au $1500.00/oz
Ag $30.00/oz
Cu $3.00/lb

Metal recovery
The metal recovery functions are complex non-linear functions of ore type and metal grades.
Oxide, transition, sulfide classification rules
If soluble copper > 1500 ppm or sulfide sulfur > 10% then
ore type = sulfide
else if soluble copper > 250 ppm then
ore type = transition
else if sulfide sulfur < 1.5% then
ore type = oxide
else if ratio of soluble gold to total gold < 0.6
ore type = transition

else
ore type = oxide
Computer program input (OCM block grades)

Total Au grade
Soluble Au grade
Ag grade
Soluble copper grade
Sulfide sulfur grade

Computer program output

Dig line tonnes and grade by ore type.

Dig line revenue by ore type as a function of recoverable Au, Cu, and Ag.
Dig line polygons

Manually designed dig lines versus Digger

This section compares dig line recoveries from a set of manually designed dig lines to those from
a set of optimum dig lines designed by Digger. The manually designed dig lines by the company
geologists were not available for this study, so we challenged an outside Ph.D. geologist with the
task of constructing dig lines that minimize the misclassification of the OCM blocks while
maintaining a MMW of 12 m or 6 blocks. The manual dig lines are shown in Figure 5 on the left.
The Digger dig lines are shown by the map on the right side of Figure 5. Although the minimum
mining width is 12 m or 6 blocks for both sets of dig lines, the white Xs drawn on the manual
set of dig lines indicate MMW violations. The tiny squares in each map represent the original ore
control block model. Ore types are identified by the grey scale. The labels in the maps indicate
the dig line ore type.

Figure 5: This figure compares a set of manually designed dig lines (on the left) to a set of computer generated
optimal dig lines (on the right). The grey scale indicates the initial ore type by block grade; black is SuM ; dark grey
is TrM; medium grey is OxM; light grey is OxL; and white is waste. The labels on the maps indicate the dig line ore
type.

The relative dig line tonnages, Au and Ag ounces, and dollar revenues by ore type are given in
Table 1. All of the entries in Table 1 are percentages of the manual dig line quantities. For

example, the total in situ dollar value of the material within the optimum dig lines is 102.8% of
the in situ value of material within the manual dig lines. In other words, the optimum dig lines
potentially provide a 2.8% increase in net revenue over the manual set of dig lines. Although the
recoverable ounces of Au and Ag compare very closely, the manual dig lines contain
approximately 116% more SuM ore tonnage than the optimum dig lines. Note that the processing
costs of SuM ore are approximately double the cost of milling either TrM or OxM. Thus, the
misclassification of material within the SuM dig lines is very costly and is likely the major
contributor to the manual dig line dollar loss. This example clearly illustrates the value of loss
functions or MPS.

Ore
Type
OxL
OxM
TrM
SuM
Total

Table 1: Comparative Dig Line Statistics.

Table entries are % of manual dig line quantities.
Au
Ag
Revenue
Tonnes
(wt % of
(wt %
(% of
(% of manual)
manual)
of manual) manual )
85.4
83.3
78.8
88.9
111.3
114.9
108.6
100.0
110.5
103.0
109.1
108.5
87.2
84.5
84.5
98.1
99.2
100.5
100.8
102.8

The 2.8% increase in net revenue provided by the optimum computer dig lines over manual
designs is typical of what has generally been observed at other mines. Increases in net revenue
generally range between 2% to 5%.
Minimum mining width examples
This section compares dig line recoveries for 4 different MMWs, namely 10 m, 12 m, 15 m, and
18 m. All dig lines were generated by Digger and the employment of loss functions. In other
words, the dig lines are physically located by minimizing the dollar loss due to the
misclassification by dig lines constrained by a minimum mining width.
The 4 sets of dig lines are shown in Figure 6. Dig line misclassifications are clearly visible in
each map. For example, near the center of the MMW-12 map, a number of waste blocks (white)
are located within a SuM dig line, e.g., these are waste blocks that will be sent to the SuM
process.

Figure 6: This figure shows 4 sets of dig lines constrained by 10 m, 12 m, 15 m, and 18 m MMWs. The grey scale
and labels indicate the ore types as explained in Figure 5. Although the 10 m MMW appears to be more selective,
with less dilution and ore loss, the relative simplicity of the 18 m MMW likely reduces dilution and ore loss as a
result of digging error at the time of mining.

Summary statistics by MMW

The dig lines shown in Figure 6 are very efficient. For example, the dollars lost due to dig line
misclassification are computed by comparing the total OCM revenue before dig line design to
the total revenue after dig line design. For example:

MW10 dig line loss = 1.36% of total OCM revenue;

MW12 dig line loss = 1.99% of total OCM revenue;
MW15 dig line loss = 2.56% of total OCM revenue; and
MW18 dig line loss = 2.12% of total OCM revenue.

10

Table 2 is interesting because it shows that the total revenues attributable to each MMW compare
very closely in spite of the increasing quantities of metal lost with larger MMWs. Although more
ounces of Au and Ag are lost with larger MMWs, the recoverable metal grades are almost
equivalent across all four MMWs. This is evident because revenue and tonnes are nearly
equivalent across the four MMWs.
Table 2: MMW Statistics
Table entries are % of OCM quantities (no dig lines).
Rev
Tonnes
Au
Ag Loss
MMW
(% of OCM)
(% of OCM)
Loss
(wt%)
(wt%)
10 m
98.6
97.7
1.59
2.69
12 m
98.0
98.1
1.68
3.94
15 m
97.4
99.4
2.54
7.05
18 m
97.9
99.5
2.02
4.71

MMW summary statistics by ore type

A comparison of the dig line dollar revenues, tonnes, Au ounces, and Ag ounces by ore type is
provided by Figure 7. A noteworthy observation in Figure 7 is that maximum recoverable
quantities for each ore type2 are delivered by different MMWs. For example, suppose a monthly
mine plan calls for the maximum production of TrM tonnes. From Figure 7, it can be seen that
the 18 m MMW is the best option. Alternatively, if the mine plan calls for the maximum
production of OxL tonnes, then a 10 m MMW is the best option. The MMW corresponding to the
maximum recoverable quantity of metal also varies with ore type. For example, if the monthly
mine plan calls for the maximum production of OxM ounces of Au, then the 12 m MMW is the
best option. Alternatively, if the mine plan calls for the maximum production of SuM ounces of
Au, then a 10 m MMW is the best option, and finally, a 15 MMW is the best option for
maximizing the production of TrM ounces of Au.

In Figure 7, the total quantity of metal recovered across all ore types including waste is
identical for each MMW. Thus, the comparisons of MMW percentages within an ore type are
relative to one another.
2

11

Figure 7: A comparison of recoverable quantities attributable to each MMW by ore type. All comparisons are given
as the percent of total. For example, the total potentially recoverable ounces Au associated with the 10 m MMW is
4.4% + 27.4% + 22.7% +45.3% + 0.2% = 100%.

Evidently, the complexity of the in situ ore type contact boundaries influences the efficiency of
dig line design. Extremely complex spatial configurations of ore type boundaries increase the
number of dig line misclassifications. Consequently, the benefit of optimum dig line design will
increase. On the other hand, in deposits where the in situ ore types are relatively continuous with
reasonably well defined contacts, the problem of dig line misclassification should be minimal. At
this point, an obvious question might be, What about blast movement? Good question, but that
is a topic for another paper.
CONCLUSIONS
Computer design of optimum dig lines constrained by a MMW is a new and relatively unknown
tool available to the mining industry. As a consequence, the benefits of constrained optimum dig
line design are largely unrealized. Constrained optimum dig line design is a low risk high
return investment with huge potential gains:

Optimum dig line design maximizes net revenue. Material sent to the wrong destination
is costly. At $1500US/oz, the dollars lost by sending a single 600 tonne OCM block of
gold averaging 0.5 g/t to the dump exceeds $14,000.00. Over the course of a year, suboptimal dig line designs can cost hundreds of thousands of dollars. These losses can
easily be avoided by using optimized computer designed dig lines. No additional
equipment, personnel, or samples are required.
Computer dig line design enables one to quickly evaluate different dig line designs. The
benefits of different MMWs or cutoff grades can be evaluated in a few minutes enabling
12

the selection of a dig line set that best satisfies current short term mine plan targets or
changes in the local patterns of mineral continuity as mining progresses through the
deposit. This is not possible with manually designed dig lines. The numerous calculations
required within a few seconds are too many for the human brain.
Computer dig line design enables rectangular MMWs. For example, the MMW in the
direction of shovel travel may be wider or narrower than the MMW in the perpendicular
direction. Rectangular dig lines may reduce loss where the patterns of mineralization
form narrow bands or veins with a reasonably consistent strike and dip. Dig lines are
easily designed in directions parallel to the strike of a vein system or a contact trace by an
internal rotation of the OCM before dig line design.
Computer dig line design can be applied to resource/reserve risk assessment. For
example, the uncertainty associated with the predictions of an annual production schedule
is often studied by calculating the in situ recoverable tonnes and grade from multiple
conditional simulations of drill hole data across the deposit. But these calculations are
typically based on free selection which is unrealistic. Computer generated optimum dig
lines permit realistic simulated recoveries of tonnes and grade for different MMWs which
not only provide the means to explore various mining options, but also increase the
accuracy of the study.

FUTURE WORK
Currently, Digger works with the OCM block grades as they are provided, no matter which block
grade estimator is used. However, work is underway with the full implementation of MPS. Given
this option, Digger will simulate a block probability distribution for each OCM block based on
the LU (Lower/Upper) decomposition of the covariance matrix (Davis, 1987). This will enable
Digger to assign initial MPS ore types to each block which will subsequently be used for
optimum dig line design. The combination of MPS applied to block grades and dig line design
will further increase the value of recoverable ore at the time of mining.
The estimation of conditionally unbiased OCM grades can be difficult if not impossible to
achieve in deposits where the distribution of sample grades is highly skewed. Conditionally
biased estimates contribute further to ore loss and dilution inherent in grade control. Work is
underway with the development of an additional option that corrects conditionally biased OCM
grade estimates (see Guertin, 1987). The corrected block grade estimates are conditionally
unbiased thereby reducing ore loss and dilution which further increases net revenue.
REFERENCES
Dagdelen, K,. and Coskun, B., 1999, Optimizing grade control in open pit gold mining, SME
Preprint Number 99-189, 25 p.
Davis, M. W. 1987. Production of conditional simulations via the LU triangular decomposition
of the covariance matrix. Mathematical Geology, Vol. 19, No.2, pp 91-96.
Deutsch, C., Magri, E., and Norrena, K., 2000, Optimal grade control using geostatistics and
economics: methodology and examples. SME Transactions, Vol. 308, pp 90-106.

13

Guertin, K., 1987, A correction model for conditional bias in selective mining operations,
Mathematical Geology, Vol 19, No. 5, pp 407-423.
Isaaks, E. H., 1990, The application of Monte Carlo methods to the analysis of spatially
correlated data: Ph.D. Dissertation, Stanford University, 213 p.
Norrena, K. P. and Deutsch, C. V., 2002, Optimal determination of dig limits for improved grade
control, 30th International Symposium on Computer Applications in the Mineral Industries
(APCOM)., Phoenix, Arizona, pp 329-339.
Rendu, J., 1982, Computerized estimation of ore and waste zones in complex mineral deposits,
SME-AIME Annual Meeting, Dallas, Texas, pp 82-95.
Schofield, N. and Rolley, P., 1997, Optimization of ore selection in mining: method and case
studies. Mining Geology Conference Proceedings. AusIMM, Melbourne, pp 93-97.
Srivastava, R.M., 1987, Minimum variance or maximum profitability?, CIM Bulletin, vol. 80,
no. 901, pp 63-68.
Srivastava, M., Hartzell, D., and Davis, B., 1992, Enhanced metal recovery through improved
grade control. 23rd Application of Computers and Operations Research in the Mineral Industry,
Society for Mining, Metallurgy, and Exploration, Tucson, AZ, pp 243-249.
FIGURE CAPTIONS

Figure 1: An example of an asymmetric linear loss function where

c 198 and c 66
1

Figure 2: An example of an ore control block model. The white lines represent dig lines constrained by a minimum
mining width (MMW). The various shades of grey indicate very high grade (VHG), high grade (HG), medium grade
(MG, and low grade (LG) ore types.
Figure 3: This figure illustrates the minimum mining width problem. All five blocks will be mined together and sent
to either the waste dump, the leach pad, or the mill. The correct destination is the one that minimizes the dollars lost
due to the misclassification of ore types, thereby maximizing net revenue.
Figure 4: This figure tabulates the loss functions for 3 ore types, waste (W), stockpile material (S), and mill
material (M). The symbols in the table cells are the calculated dollars lost (using equation 2) as a result of dig line
misclassifications. See text for additional definitions.
Figure 5: This figure compares a set of manually designed dig lines (on the left) to a set of computer generated
optimal dig lines (on the right). The grey scale indicates the initial ore type by block grade; black is SuM ; dark grey
is TrM; medium grey is OxM; light grey is OxL; and white is waste. The labels on the maps indicate the dig line ore
type.
Figure 6: This figure shows 4 sets of dig lines constrained by 10 m, 12 m, 15 m, and 18 m MMWs. The grey scale
and labels indicate the ore types as explained in Figure 5. Although the 10 m MMW appears to be more selective,
with less dilution and ore loss, the relative simplicity of the 18 m MMW likely reduces dilution and ore loss as a
result of digging error at the time of mining.
Figure 7: A comparison of recoverable quantities attributable to each MMW by ore type. All comparisons are given
as the percent of total. For example, the total potentially recoverable ounces Au associated with the 10 m MMW is
4.4% + 27.4% + 22.7% +45.3% + 0.2% = 100%.

14

TABLE CAPTIONS
Table 1: Comparative Dig Line Statistics.
Table entries are % of manual dig line quantities.
Table 2: MMW Statistics
Table entries are % of OCM quantities (no dig lines).

FIGURES

Figure 1: An example of an asymmetric linear loss function where

c 198 and c 66
1

Figure 2: An example of an ore control block model. The white lines represent dig lines constrained by a minimum
mining width (MMW). The various shades of grey indicate very high grade (VHG), high grade (HG), medium grade
(MG, and low grade (LG) ore types.

15

Figure 3: This figure illustrates the minimum mining width problem. All five blocks will be mined together and sent
to either the waste dump, the leach pad, or the mill. The correct destination is the one that minimizes the dollars lost
due to the misclassification of ore types, thereby maximizing net revenue.

PM

PM PS

PS

PS PM

PS

PM

W*

S*

M*

Figure 4: This figure tabulates the loss functions for 3 ore types, waste (W), stockpile material (S), and mill
material (M). The symbols in the table cells are the calculated dollars lost (using equation 2) as a result of dig line
misclassifications. See text for additional definitions.

Figure 5: This figure compares a set of manually designed dig lines (on the left) to a set of computer generated
optimal dig lines (on the right). The grey scale indicates the initial ore type by block grade; black is SuM ; dark grey
is TrM; medium grey is OxM; light grey is OxL; and white is waste. The labels on the maps indicate the dig line ore
type.

16

Figure 6: This figure shows 4 sets of dig lines constrained by 10 m, 12 m, 15 m, and 18 m MMWs. The grey scale
and labels indicate the ore types as explained in Figure 5. Although the 10 m MMW appears to be more selective,
with less dilution and ore loss, the relative simplicity of the 18 m MMW likely reduces dilution and ore loss as a
result of digging error at the time of mining.

17

Figure 7: A comparison of recoverable quantities attributable to each MMW by ore type. All comparisons are given
as the percent of total. For example, the total potentially recoverable ounces Au associated with the 10 m MMW is
4.4% + 27.4% + 22.7% +45.3% + 0.2% = 100%.

TABLES
Table 1: Comparative Dig Line Statistics.
Table entries are % of manual dig line quantities.
Au
Ag
Revenue
Ore
Tonnes
(wt
%
of
(wt
%
(% of
Table 2: MMW Statistics
Type Table(%
of manual)
manual)
manual)
entries
are % of OCM
quantitiesof(no
dig lines).manual )
OxL
85.4
83.3
78.8
88.9
Rev
Tonnes
Au
Ag
Loss
OxM
111.3
100.0
MMW
(% of
OCM)
(%114.9
of OCM) 108.6
Loss
(wt%)
TrM
110.5
103.0
109.1
108.5
(wt%)
SuM
87.2
84.5
84.5
98.1
10
m
98.6
97.7
1.59
2.69
Total
99.2
100.5
100.8
102.8
12
m
98.0
98.1
1.68
3.94
15 m
97.4
99.4
2.54
7.05
18 m
97.9
99.5
2.02
4.71

18