You are on page 1of 7

Blast Fragmentation Modelling of the Codelco-Andina Open Pit Expansion

F. Mardones
GeoBlast S. A., Chile
C. Scherpenisse
GeoBlast S. A., Chile
I. Onederra
The University of Queensland, Sustainable Minerals Institute, W H Bryan Mining and Geology Research
Centre, Qld 4072 Australia
ABSTRACT: In large scale metalliferous mining, there is significant evidence to suggest that by providing an
appropriate size distribution to crushing and grinding circuits, a measurable increased throughput and/or re-
duced power draw can be obtained. Tailoring blast designs to suit specific fragmentation requirements is now
common place at both the pre-feasibility and feasibility study stages. This is particularly the case when sig-
nificant increases in ore production rates are being considered. As part of the feasibility study of the open pit
expansion project of Codelco’s Andina operation, a comprehensive blast monitoring program was conducted
in the currently mined secondary ore domains of the “Don Luis” pit. The objective of this program was to
calibrate and implement a site specific blasting model to enable the prediction of fragmentation trends in the
deeper, more competent ore zones, also referred to as primary rock domains. This paper gives a brief descrip-
tion of the blast fragmentation monitoring program conducted and discusses the calibration and application of
a stochastic blast fragmentation modelling framework. Results from several simulations have highlighted the
key differences in fragmentation if current blast designs are applied in the more competent primary rock do-
mains. A number of blast design options have been evaluated and recommendations made in order to achieve
specific ore handling and processing targets.

1 INTRODUCTION 2 DESCRIPTION OF BLASTING DOMAINS

Blasting activities in major mining operations The main geotechnical units forming the core of
have been placing significant emphasis on the ability the mining environment of current and future opera-
to tailor fragmentation to improve downstream proc- tions at Andina have been grouped into Primary
esses. In many of these operations, the impact of Rock, Secondary Rock and Riolite and Dacite chim-
fines has been clearly identified. neys. This study is mainly concerned with the cali-
At the conceptual and feasibility stages, fragmen- bration of an empirical fragmentation model in sec-
tation modelling studies which support future Mine ondary rock and simulations in primary rock.
to Mill strategies can be conducted through the cali- Primary rock masses have been described as
bration of empirical models using existing data; and “hard” and competent with well healed gypsum or
if need be, through the implementation of specific Anhidrite filled fractures, typical RMR values are in
trials. An important pre-requisite is the adequate the range of 60 to 80. Secondary rock masses can
classification and characterisation of the blasting also be described as “hard”, however fractures are
domains of interest. As part of the feasibility study generally open and hence reduce the competency of
of the open pit expansion project of Codelco’s the rock mass with characteristic RMR values in the
Andina operation, a comprehensive blast monitoring range of 42 to 50.
program was conducted in the currently mined sec- Relevant to blast fragmentation modelling is the
ondary ore domains of the “Don Luis” pit. The ob- degree of in situ fracturing. As shown in Figure 1,
jective of this program was to calibrate and imple- total spacing statistics derived from fracture fre-
ment a site specific blasting model to enable the quency data show that the degree of fracturing is
prediction of fragmentation trends in the deeper, clearly more intense in the secondary rock mass do-
more competent ore zones mains. Results from the available core logging data
indicated that total fracture spacing may be as wide
as 0.4 m in the secondary domain and 0.91 m in the
primary rock domain. The analysis shows that the
variability in fracture intensity appears to be greater
in the primary rock domain.
From a drilling and blasting perspective the sec- material data collected by the geology and geotech-
ondary and primary domains can be classified as nical department of Andina provided the necessary
fractured and blocky rock masses respectively. Frac- input to reliably implement a stochastic modelling
ture spacing statistics were used to provide first pass approach.
estimates of the potential range of the mean size of
in situ blocks. This is a required input parameter that
is further defined through back-analysis or the model
calibration process.

Table 1. Summary of intact rock properties in secondary and


primary domains.

Figure 1. Total spacing statistics of secondary and primary rock


masses. 3 OVERVIEW OF MONITORED BLASTS

The rock types or lithological units of concern to The detailed monitoring of production blasts in
this study include Granodiorita Cascada (GDCC) secondary rock masses has been an important and
and Brecha de Turmalina (BXT). Both rock types are necessary component of the model calibration and
found in the secondary and primary rock domains. In verification process. As summarised in Table 2, a to-
both cases, the degree of alteration appears to be the tal of four production blasts were monitored in the
main factor that affects strength and stiffness. Analy- “Don Luis” pit of the Andina operation, three were
sis of the geotechnical information provided by located in the GDCC domain and one in the BXT
Andina allowed the definition of the average intact domain.
rock material and rock mass parameters used in the
calibration and modelling of different blasting sce-
narios. Table 1 gives a summary of the domain and
properties of the GDCC and BXT rock types.
Table 1 shows that the intact rock material in the
primary domain is slightly stiffer than in the secon-
dary domains. In terms of compressive strength,
there are no significant differences between the
GDCC rock in the secondary and primary domains
with mean values of 161 and 156 MPa respectively.
A more pronounced difference is observed in the
BXT rock with mean values of 135 and 158 MPa re- Table 2. Design parameters of monitored production blasts.
spectively. From a drilling and blasting perspective, * The explosive Apex 150 is a Heavy ANFO (50% Emulsion)
all rock material can be classified as hard and com- product supplied by Orica Chile.
petent. In these hard competent conditions incipient
damage defined by peak particle velocity is esti-
mated to be in the range of 900 to 1100 mm/s; and 4 FRAGMENTATION ASSESSMENT
breakage is expected to be in the range of 3600 to
4400 mm/s. The overall breakage and fragmentation A detailed fragmentation assessment program was
potential is expected to be driven by the degree and conducted during this study. The detailed program
condition of fracturing; and therefore differences in included the acquisition of images during the exca-
the intermediate and coarse end of the fragmentation vation of muckpiles as well as the sieving of a lim-
distribution are expected between secondary and ited number of samples taken from selected regions.
primary rock masses. It should be noted that the rock As illustrated in Figure 2, the assessment procedure
consisted of sampling lines and profiles taken at dif-
ferent stages of extraction. This procedure is consis-
tent with best practices in fragmentation assessment
using image analysis methods.
Table 3. Summary of fragmentation images samples taken dur-
ing the monitoring of blasts in the GDCC and BXT domains.

5 BLAST FRAGMENTATION MODELLING

The expected distribution of fragments in the


fines and coarse regions is modelled by two separate
distributions based on the recently published Swe-
brec function (Ouchterlony, 2005). The Swebrec
function has recently shown to be far superior in fit-
ting fragmented rock in the intermediate and finer
end of the fragmentation curve than previous mod-
els. The main modelling framework includes the
ability to consider a range of values to key input pa-
rameters through the explicit definition of distribu-
tion functions. In this way stochastic simulations can
be conducted to determine a predictive fragmenta-
tion envelope that takes into account the variability
of rock material, rock mass, blast geometry and ex-
plosive performance parameters. The current ap-
proach also incorporates modelling parameters that
can simulate the impact of inter-hole delay timing on
fragmentation (Onederra, 2008).

5.1 Calibration results in secondary ore


As has been extensively discussed in the literature,
one of the main limitations of empirical fragmenta-
tion models is their requirement for site specific
calibration. This necessary process generally in-
volves the back analysis or prediction of fragmenta-
tion based on measured data and monitored prac-
Figure 2. Example plan view of sampling lines of blast
tices. As mentioned earlier, four production blasts
3724_12. covering GDCC and BXT secondary rock domains
were used to calibrate the proposed fragmentation
Detailed analysis included both manual editing modelling framework. The calibration process in-
and the definition of site specific fine correction fac- volved the refinement of estimates associated with
tors. These factors were determined directly by the rock mass parameters likely to impact on uniformity,
sampling and sieving of fragments in the areas of in- mean fragmentation outcomes and the propensity of
terest. Blasting literature shows that reliable esti- the rock fabric to generate fines during the fracturing
mates of Run of Mine (ROM) fragmentation can be process.
obtained following procedures similar to those in- Figure 3 summarises the results of comparisons
corporated in this study (Latham et al 2003 and San- between predicted and measured fragmentation out-
chidrian et al 2005). comes for one of the blasts in the GDCC domain
The data obtained from the monitored areas were (3724_10). In this analysis, statistics associated with
used in this study to calibrate and verify the blast rock material input parameters, pattern geometry and
fragmentation models implemented in this study. explosive performance were included to generate an
The total number of samples taken in both the expected fragmentation bounded by envelopes of
GDCC and BXT domain are summarised in Table 3. minimum, maximum and 95% confidence. It is im-
portant to note that the fragmentation envelope given
by each simulation is a function of the level of un-
certainty or variability assigned to the available input 6 RESULTS AND DISCUSSION
data.
Fragmentation modelling results for GDCC and
BXT primary ore domains are summarised in Fig-
ures 4 and 5 respectively. Note that only the ex-
pected size distribution curves are shown for com-
parison purposes. Modelling results demonstrate the
influence that changes in pattern geometry may have
on fragmentation, particularly in the intermediate
and finer size fractions. Differences between do-
mains and designs are also summarised in Table 5.
For similar pattern geometries and corresponding
powder factors, modelling results suggest that blast-
ing in the GDCC domain has the potential to gener-
Figure 3. Summary of calibration results based on monitored ate more fines than in the BXT domain. Relative dif-
production blasts. ferences may be of the order of 3% to 5% between
these two domains. As expected, designs D3 and D6
5.2 Fragmentation modelling of primary rock give the finest fragmentation in the GDCC domain;
and designs D9 and D12 give the finest fragmenta-
A total of 14 simulations were conducted to quan- tion in the BXT domain. By comparing designs D2
tify relative changes in ROM fragmentation in pri- and D2A, modelling results suggest that by decreas-
mary rock. Table 4 gives a summary of the pattern ing stemming lengths by approximately 1 m, a 1%
geometries and powder factor ranges investigated. gain is expected in the amount of fines generated in
As shown, powder factors reflect the use of pattern the GDCC domain. In the BXT domain however the
geometries similar to those currently implemented at gain is only approximately 0.5%, as shown by com-
Andina, as well as more aggressive designs which paring designs D8 and D8A. It is important to note
include both reductions in burden, spacing and that model calibrations in the BXT domain were
stemming lengths. All simulations have maintained only based on a single production blast. More data
the use of 270 mm diameter blastholes using Apex may be required to improve the predicted capabili-
150 and Apex 165 as the base case explosive prod- ties of the model in this particular domain.
ucts. It should also be noted that a single hole firing
mode was assumed with inter-hole delays of 10 ms.
As discussed earlier, the adopted stochastic ap-
proach has allowed the inclusion of distribution
functions to rock material and rock mass input pa-
rameters as well as design specific parameters such
as hole and charge lengths. The Latin Hypercube
sampling technique was used with simulations set to
500 iterations.

Table5. Summary of fragmentation modelling results in pri-


mary ore conditions.

Table 4. Summary of pattern configurations for GDCC and


BXT primary ore domains.
* The explosive Apex 165 is a Heavy ANFO (65% Emulsion)
product supplied by Orica Chile.

Figure 4. Comparison between designs in GDCC primary ore.


Figure 5. Comparison between designs in BXT primary ore.

As discussed earlier, single hole firing conditions


were adopted in the modelling calculations, in this
case, a 10 ms inter hole delay was assumed based on Figure 6. Modelling results showing the potential influence of
short inter-hole delay times on fines for design D2 (GDCC
estimations of minimum response time (Onederra,
Domain).
2007). The current modelling framework was used
to investigate potential gains in fines generation by
introducing shorter delays (e.g. 2 ms to 10 ms). De- It is important to note that simulations are indica-
signs D2 and D8 were used as base cases for the tive of what may be achieved if all measured and as-
GDCC and BXT domain respectively. Results of the sumed modelling conditions are met. Actual meas-
analysis for the expected values in GDCC are sum- urable results will undoubtedly be influenced by the
marised in Figure 6. As shown, for the ½” and 1” field implementation process. For this reason, the
size fractions with the use of inter hole delays of 2 implementation of a Quality Assurance / Quality
ms, gains of approximately 2 % and 3.5 % may be Control strategy (QA/QC) was strongly recom-
achieved in the GDCC. The use of very short inter mended, particularly as improved designs are im-
hole delays (e.g. 2 ms) demonstrates gains in the in- plemented in both current and future domains (Sec-
termediate and fine fractions, however as shown in ondary and primary rock). The impact on
Figure 6, these gains may not be significant if one is
to consider the variation associated with modelling fragmentation outcomes given by variations in pat-
predictions, and in particular the lower limit predic- tern geometry is demonstrated for design D2 in Fig-
tive envelope. It should also be noted that the inter ure 7. In this case, a standard deviation of 0.5 m was
hole delay adjustment factors proposed in the current assumed for the mean values of burden and spacing.
modelling framework (Onederra, 2008) are based on Results show a widening of the predictive envelope,
limited data and further validation will still be re- which can translate into coarser or more bi-modal
quired in primary rock conditions. fragmentation outcomes.
Although fragmentation may be improved, it is
important to note that “high intensity” blasting with
the use of short inter hole delays may be counter
productive if the risk of rock mass damage is in-
creased and loading productivity is influenced by the
lack of muckpile looseness. Preliminary modelling
results have highlighted the need to further quantify
the potential impact of short delays on near field
damage and downstream loading productivity. This
should be considered a priority if short inter hole de-
lays are to be used in primary rock production blast-
ing.

Figure 7. Potential impact on fragmentation outcomes given by


simulated variations in pattern geometry
7 CONCLUSIONS required to improve the predicted capabilities of the
model in this particular domain
A comprehensive production blast monitoring
program was conducted in secondary ore domains of The current modelling framework was used to in-
the “Don Luis” pit at the Codelco-Andina operation. vestigate potential gains in fines generation by intro-
The objective of this program was to calibrate and ducing shorter delays (e.g. 2 ms to 10 ms). Designs
implement a site specific blast fragmentation model D2 and D8 were used as base cases for the GDCC
to predict fragmentation outcomes in primary rock and BXT domain respectively. Results showed that
domains. The rock types or lithological units of main for the ½” and 1” size fractions with the use of inter
concern to this study included Granodiorita Cascada hole delays of 2 ms, gains of approximately 2 % and
(GDCC) and Brecha de Turmalina (BXT). A total of 3.5 % may be achieved in the GDCC domain. The
four production blasts were comprehensively moni- use of very short inter hole delays (e.g. 2 ms) dem-
tored in secondary rock, three were located in the onstrates gains in the intermediate and fine fractions,
GDCC domain (i.e. 3724_10, 3724_12 and however these gains may not be significant if one is
3724_09) and one in the BXT domain (i.e. 3708_3). to consider the variation associated with modelling
The calibration process allowed the definition and predictions, and in particular the lower limit predic-
refinement of estimates associated with key rock tive envelope.
mass indices which impact on the expected uniform-
ity, mean fragment size and the propensity of the
Preliminary modelling results have highlighted the
rock fabric to generate fines.
need to further quantify the potential impact of short
The calibrated model used in this study can be delays on near field damage and downstream loading
best described as a two component model utilising productivity. This should be considered a priority if
the Swebrec fragmentation distribution function. The short inter hole delays are to be used in primary rock
adopted approach is stochastic and therefore allows production blasting.
the inclusion of distribution functions to rock mate-
rial and rock mass input parameters as well as design
specific parameters such as hole and charge lengths. It is important to note that proposed changes in
The modelling framework also incorporated model- blasthole configurations and geometry (i.e. tighter
ling parameters that can simulate the impact of inter patterns) may be restricted by operational matters.
hole delay timing on fragmentation. These types of constraints should be reviewed and
assessed at the operational level.
14 simulations were conducted to quantify relative
changes in ROM fragmentation in primary rock
(GDCC and BXT rock types). A range of pattern ge-
ometries and corresponding powder factors were in-
vestigated. The analysis indicated that for similar
pattern geometries, blasting in the GDCC domain
has the potential to generate more fines than in the
BXT domain. Relative differences may be of the or-
der of 3% to 5% between these two domains.
As expected, designs D3 and D6 gave the finest
fragmentation in the GDCC domain; and designs D9
and D12 gave the finest fragmentation in the BXT
domain. By comparing designs D2 and D2A, model-
ling results suggested that by decreasing stemming
lengths by approximately 1 m, a 1% gain is expected
in the amount of fines generated in the GDCC do-
main. In the BXT domain however the gain was only
approximately 0.5%, as shown by comparing designs
D8 and D8A.
Model calibrations in the BXT domain were only
based on a single production blast. More data may be
REFERENCES

C. V. Cunningham: Fragmentation estimations and


the Kuz-Ram model - Four years on. Proceedings of
the second international symposium on rock frag-
mentation by blasting, Keystone, Colorado, 1987,
475-487.

I. Onederra, S. Esen, and A. Jankovic: Estimation of


fines generated by blasting - applications for the
mining and quarrying industries. IMM transactions,
Mining Technology, Vol 113, 2004, No.4:237-247.

F. Ouchterlony: The Swebrec function: linking


fragmentation by blasting and crushing. IMM trans-
actions, Mining Technology, Vol 114, March 2005,
No1:A29-A44.

J. P. Latham, J. Kemeny, N. Maerz, M. Noy, J.


Schleifer and S. Tose: A blind comparison between
results of four image analysis systems using a photo-
library of piles of sieved fragments. FRAGBLAST -
International Journal of Blasting and Fragmentation
2003, 7: 105-132

I. Onederra: Empirical charts for the estimation of


minimum response time (Tmin) in free face blasting.
IMM transactions, Mining Technology, Vol 116 ,
March 2007, No 1: 7-15

I. Onederra: A delay timing factor for empirical


fragmentation models. IMM transactions, Mining
Technology, February 2008. Vol 116, No 4, 176-
179.

J. A. Sanchidrian, P. Segarra and L. M. Lopez: A


Practical Procedure for the Measurement of Frag-
mentation by Blasting by Image Analysis. Rock Me-
chanics and Rock Engineering, November 2005.

GeoBlast S:A, Final Report PRO–ASP–246/07-E,


“Modelamiento de la Fragmentación Resultante de
Tronadura en Roca Primaria''. Emitido para:
Proyecto Expansión Andina, CODELCO, Abril 9,
2008.

You might also like