Вы находитесь на странице: 1из 7

ARMA 11-252

Assessing geological vein size and intensity using Discrete Fracture

Network modelling at the El Teniente Mine, Chile.

Brzovic, A. and Herrera, S.

Codelco Chile The El Teniente Division, Rancagua, Chile.

Copyright 2011 ARMA, American Rock Mechanics Association

This paper was prepared for presentation at the 45 US Rock Mechanics / Geomechanics Symposium held in San Francisco, CA, June 2629,
This paper was selected for presentation at the symposium by an ARMA Technical Program Committee based on a technical and critical review of
the paper by a minimum of two technical reviewers. The material, as presented, does not necessarily reflect any position of ARMA, its officers, or
members. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of ARMA
is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The
abstract must contain conspicuous acknowledgement of where and by whom the paper was presented.

ABSTRACT: Using the commercial FracMan software package, discrete fracture model (DFN) of the geological stockwork veins
has been built at the El Teniente mine that honoured data collected in drill cores and mine drives. DFN models explicitly represent
key vein structures in 3D and require as a minimum data, the orientation, the size and the intensity of the veins to be defined. The
orientations of the structural data observed in oriented cores were used to bootstraps veins orientations within the modelled
volume. The correlation between observed and simulated orientation data sets was evaluated by applying Chi-Square tests and the
results were outstanding. Vein size was assessed through applying the simulated sampling technique to match the observed trace
length measurements collected in wall mine drives. The results showed the stockwork veins as having small size with a mean
around 1.0m for both lognormal and exponential size distributions. Discontinuity intensity was determined by conditioning linear
vein frequency of the three different core measurements using vein orientation and size properties determined in previous
modelling steps. Veins intensity, measured as the area of fracturing in a unit volume, was estimated as 46m2/m3, but for veins that
control fragmentation and caving propagation, intensity values were estimated between 1m2/m3 and 5m2/m3.

El Teniente mine by applying the concept of simulated

1. INTRODUCTION sampling [5], which allows issues of sampling bias to be
Discontinuity size and intensity are key parameters to accounted for. Simulated sampling techniques have been
characterize the rock structure for geotechnical and easily applied using the commercial software package
fragmentation considerations in any panel caving FracMan [6]. Through this methodology, DFN models
underground mine. This is particularly relevant at the El honoured the data collected at the mine site. DFN
Teniente mine, because it has been observed that the models explicitly represent key vein structures in 3D and
geological veins, which are the main and most common require as a minimum, the vein orientation, size and
geological structures within the ore body, play a relevant intensity to be defined.
role controlling fragmentation and seismicity during Structural data collected in oriented cores and scanlines
caving propagation [1-3]. mapping from mine drives within the dacita mine sector
Determining discontinuity size and intensity are the most have been used to build DFN models to determine
difficult tasks in rock mechanics since the geological discontinuity size and intensity of the geological veins.
structures are three-dimensional entities whereas rock This paper describes the results of this ongoing project
structure characterisation can only be undertaken at the of rock structure characterisation for geotechnical
best two dimension rock wall/surface exposure or the purposes.
one dimension drill core. Considering the limited rock
exposure, Discrete Fracture Network (DFN) modelling 1.1. El Teniente Mine Overview
has been developed and implemented as the way to The El Teniente mine is a Codelco-Chile underground
address this issue. Discrete Fracture Network modelling copper mine. It is located in the Andes Mountains in the
may not be an easy task since sampling bias must be central zone of Chile, about 70 km South-Southeast from
corrected, which sometimes implies both complex the capital city, Santiago. Since 1906 more than 1,100
mathematical solutions [4] and questionable million tons of ore have been mined using block caving
representation of true 3D occurrence of the geological methods, currently mine production is about 135,000
structures. tonnes per day (tpd), which is mainly of primary copper
Despite these shortcomings of Discrete Fracture ore. Rock masses of primary copper ore at the El
Network modelling, DFN models have been built at the
Teniente mine have been described as very competent network modelling were collected in dacite rock type
and massive during caving operation. This is because within the El Teniente ore body during recent feasibility
caving performance has been typified by larger than studies for mining (Figure 1).
expected fragmentation, and the rock mass has exhibited Data collection techniques used three oriented cores
brittle, often violent failure under high stress conditions drilled in three semi-orthogonal sampling orientations to
[7]. ensure to capture all possible discontinuity orientations.
El Teniente is the largest known copper-molybdenum Over 10 meters of detailed structural logs were
deposit in the world; main rock types include Mafic undertaken from each oriented core. Old structural
(gabbros, diabases and other called Cmet) and Felsic scanline data collected in 4m high mine drives, also in
(dacite and tonalite-diorite porphyries) intrusives, and three semi-orthogonal sampling orientations, were used
hydrothermal breccias of the Miocene era [8]. Two main to provide discontinuity trace length measurements.
structure types are observed within the primary copper Veins orientation was compared between oriented core
ore; a system of large-scale faults and a stockwork and scanline mapping to ensure that all possible
having a high frequency of small-scale vein features [2]. discontinuity orientations were captured.
Veins are mainly cemented with quartz, sulphides and FracMan was used to build DFN models of the El
anhydrite. The stockwork veins containing the original Teniente vein system. Firstly, orientation data were
mineralogy are healed. Different vein types have been replicated by using a bootstrapping [9] technique
identified, and these define alteration zones according to included in the software. Secondly, vein size was
their local abundance. These geological features agree assessed through applying the simulated sampling
with its observed competent and massive behaviour technique [5] to match the observed trace length data
during caving operations, however, it has been collected in mine drives. Lognormal and exponential
established that rock mass disassembly during caving distributions to the discontinuity size were fitted in this
propagation and fragmentation process occur thought study since previous studies at the El Teniente mine
veins having weaker infill mineral assemblages, named showed them as the best match to the observed structural
weak veins [1-3]. data [10].
Discontinuity intensity, known as P32 (fractured
area/volume; m2/m3) and introduced by Dershowitz and
Herda [11] was determined by conditioning observed
DFN models of the El Teniente stockwork veins require linear vein frequency P10 (m-1) [11] of the three
as a minimum data, the vein orientation, size and different core measurements whilst using vein
intensity to be defined. Structural data used in the. orientation and size properties determined in previous
modelling steps.

Fig. 1 Sampling location at dacita rock type within El Teniente mine (z; high in mine coordinates)
indicates a very good correlations between data sets
2.1. Assumptions considering 8 degrees of freedom (n = 9-1 spaces) for
DFN modelling assumes a Poisson distribution for 3D 95% of confidence level [if P(82 < x) = 0.95, then x
veins occurrence. It means that each vein happened 15.507].
independent of each other, i.e., vein occurrence is
assumed to be not as deterministic process as geologists
may expect. Discontinuity shape is assumed to be
similar to circular disk with neglected thickness at this
stage of the project. These assumptions are considered
reasonable with regards to a model of the vein system
Discontinuity properties such as orientation, size and
intensity are assumed the same within the modelling
volume, usually 25m x 25m x 25m. Discontinuity size
distribution is independent of the spatial location or
family set; this is also a reasonable assumption
considering that the magmatic process of hydraulic
fracturing that formed the stockwork veins should be the
same within this small volume range.

3.1. Discontinuity Orientation Fig. 3 Areas or spaces within lower hemispheric projection for
The orientation of the structural data observed in comparison of the structural data [13].
oriented cores was used to bootstrap vein orientations
within the modelled volume. To evaluate the 3.2. Discontinuity Size
bootstrapping process, simulated drill cores were
Discontinuity size assessment was carried out in two
introduced in the model, and the modelled
steps. Firstly the correlation between modelled
discontinuities were intersected within the same core
discontinuity size and their observed trace length in
section logged. Visual comparison between observed
several cross sections for a large modelled volume scale
and simulated data is shown in Figure 2.
was studied to test orientation influence of selected cross
A Chi-Square test (2) was applied to evaluate the
section. Secondly, the same was done to match observed
correlation between observed and modelled structural
and censored trace length data collected in 4m high mine
data. The lower hemispheric projection containing the
wall drifts. In both cases, the discontinuity orientations
structural data was divided in 9 similar spaces (Figure
from previous modelling steps were used.
3), and pole counting was undertaken at each space to
establish a comparison between each data set. Results of
the Chi-Square test are presented in Table 1, which

Fig. 2 Comparison between sampled data and simulated data using bootstrapping in FracMan [12]
Table 1. Chi Square Test results to the observed and simulated
orientation data.
DipDir DipDir Dip Dip Count Count 2
(Min) (Max) (Min) (Max) Real Sim. Value
a 0 45 40 90 49 58 1.653
b 45 90 40 90 61 68 0.803
c 90 135 40 90 88 104 2.909
d 135 180 40 90 213 198 1.056
e 180 225 40 90 54 52 0.074
f 225 270 40 90 70 77 0.700
g 270 315 40 90 82 71 1.476
h 315 360 40 90 95 93 0.042
i 0 360 0 40 134 129 0.187
Sum 846 850 8.900

Fig. 5 Simulated sampling results in the 100m x 100m cross

Three cross sections differently oriented were used to
sections and in the mine drive wall 4m high [12].
intersect simulated discontinuities in a volume of 100m x
100m x 100m (Figure 4) to evaluate the correlation
between discontinuity size and observed trace length. The second step in the discontinuity size assessment was
The results for lognormal size distribution, using mean to recreate the same procedure to match measured trace
equal to standard deviation (SD), are shown in Figure 5, length of veins. Data of three different scalines
which shows, as could be expected since it was assumed campaign undertaken within dacita rock type (locations
that discontinuity size is independent of orientation, that in Figure 1) indicate an average trace length between
sampling orientation is not relevant to the observe 1.84m and 2.48m and a standard deviation between
discontinuity trace length in a 2D space. The same 1.28m and 1.72m. These measurements of discontinuity
procedure was carried out considering 2 vertical cross trace length were used as the Target size for simulated
sections 4m high (as mine wall drive), and the results sampling.
were similar to the previous exercise (Figure 4 and 5) in Once again, lognormal and exponential size distributions
a simulated space of 100m x 100m x 50m. Similar were tested in the procedure with the assumption that the
results were also found for exponential size distribution. mean was equal to the SD for lognormal size
Another conclusion of this exercise is the observed distribution. Latter on it was tested for a mean being
tendency of the simulated trace length in drive mapping different to the SD for lognormal size distribution. All
4m high. In this case, the mean trace lengths slightly results of this data analysis are shown in Figure 6, and
decrease with size reduction. several conclusions can be drawn from these results:
At any case of the discontinuity size distribution, the
mean trace lengths clearly decrease with the
discontinuity size reduction. It means that the true
vein size must be small.
In order to match the observed mean and SD trace
length measurements, the size of veins should have a
mean in the range of 1m. Specifically the vein size
should have values around 0.9m for exponential size
distribution, and a mean value around 1.0m and SD
around 0.8m for lognormal size distribution.

3.3. Discontinuity Intensity

FracMan allows determining discontinuity intensity P32
by conditioning linear vein frequency P10 from standard
core logging. Using vein orientation and size properties
found in previous modelling steps, veins intensity was
determined, and the results are shown in Figure 7 and
Fig. 4 Different cross sections showing discontinuity trace Table 2.
length within the simulated space in Dacita rock type.
the sum of all observed P10. Figure 7 shows this
matching process for ten model iterations. Quantification
of this correlation is presented in Table 2, which
indicates an error lower than 10%. Chi-Square test
values considering 2 degrees of freedom (n = 3-1 cores)
for 95% of the confidence level also shows a very good
correlation between data sets [if P(22 < x) = 0.95, then x
5.991]. Although, a good correlation between data sets
should be expected, because of the conditioning process,
data analysis considering both only weak veins [2] and
only thicker weak veins [2], which also considered less
real data available (11% and 3% of the all stockwork
veins respectively) showed correlation with large error
and less confidence levels (Table 3). The calculation in
Table 3 used exponential distribution for vein size.
Table 2 Statistic of P10 conditioned simulations to determine
Log Normal
Oriented P10 P10
Coeff. Error Average 2
cores observed Simulated
Var. (%) error(%) Value
SG0320 18.8 22.4 0.02 19.3
SG0321 33.5 30.1 0.04 10.1 9.9 0.75
SG0322 23.2 23.3 0.06 0.2
Mean 46.06
2 3
(m /m ) SD 0.82
CV 0.02
Oriented P10
cores P10 Simulated Coeff. Error Average 2
observed (average) Var. (%) error(%) Value
Fig. 6 Simulated sampling results in a mine drive wall 4m SG0320 18.8 22.0 0.05 17.1
high [12]. SG0321 33.5 30.9 0.04 7.5 8.9 1.04
SG0322 23.2 22.8 0.05 1.9
Mean 45.52
(m2/m3) SD 1.13
CV 0.02
Note; Error is estimated as; |P10T-P10S|/P10T, where T; target, S; simulated.

From each realisation P32 was determined in the model.

The statistics of P32 calculations are also presented in
Table 2 and Table 3, which shows very low variability
between realisations for good confidence levels. Vein
intensity measured as the area of fracturing in a unit
volume was estimated as 46m2/m3 for all stockwork
veins, which was independent of the vein size
distribution assumed. In Figure 8 there is comparison
between observed veins (in Cmet and dacita rock types)
and simulated data for a 1 m2 mine wall section.
Although Cmet rock tape is not analysed here, the
stockwork characteristics can be considered similar to
Fig. 7 Veins lineal frequency (P10) simulation results to
estimate veins intensity (P32) for lognormal vein size
dacita (there is grater colour differences between the
distribution. white veins and the black Cmet rock type).
Veins intensity for weak veins and weak veins having
thickness grater or equal to 2mm (thicker) were
FracMan generates structures that fill up the modelled estimated as 5m2/m3 and 1m2/m3 respectively (Table 3).
volume until the P10 of all cores are satisfied. It stops
when the sum of all simulated P10 is equal or greater than
Fig. 8 Comparison between observed veins from mine drive (Cmet rock type, left) and core (Dacita rock type, right) and the
simulated 1 m2 mine drive section using P32 of 46m2/m3 (centre).

Table 3 Statistic of P10 conditioned simulations to determine

orientation data showed correlation above a 95% of
P32 for weak and thicker weak veins. confident level.
Weak Veins The size of the strockwork veins was estimated by
Oriented P10 P10 applying simulated sampling techniques, which
cores observed Simulated Coeff. Error Average 2 allowed avoiding correcting sampling bias. It was
(average) Var. (%) error(%) Value found that the veins have a small size of around 1m
SG0320 2.1 2.7 0.16 28.9 as the mean for both lognormal and exponential size
SG0321 2.2 3.0 0.21 36.8 32.7 0.953
distributions. Standard deviation for lognormal size
SG0322 4.3 2.8 0.22 34.6 distribution has also a very low value of about 0.8m.
Mean 5.05
2 3
(m /m ) SD 0.64 Veins intensity, measured as the area of fracturing in
CV 0.13 a unit volume, was estimated as 46m2/m3 for all
Weaker Veins having thickness 2mm stockwork veins, but for veins that control
Oriented P10 P10 fragmentation and caving propagation, intensity
cores observed Simulated Coeff. Error Average 2
Var. (%) error(%) Value
values were estimated between 1m2/m3 and 5m2/m3
SG0320 0.4 0.4 0.44 4.8 DFN technique that honoured observed geological data
SG0321 0.4 0.7 0.41 83.9 42.5 0.419 was found very worth to characterize the rock structure
SG0322 0.7 0.4 0.25 38.6 of veins at the El Teniente mine.
Mean 0.92
2 3 SD 0.22
(m /m )
CV 0.23
Note; Error is estimated as; |P10T-P10S|/P10T, where T; target, S; simulated. 1. Brzovic, A. (2009) Rock mass strength and seismicity
during caving propagation at the El Teniente Mine,
Chile. IN TAN, C. A. (Ed.) 7th Int. Symp. Rockbursts
and Seimicity in Mines (RaSiM7). Dalian, China,
Structural data collected in oriented cores and scanlines Rinton Press, New York. Vol. 2, pp. 838-852.
mapping from mine drives within the dacita mine sector 2. Brzovic, A. & Villaescusa, E. (2007) Rock mass
have been used to build a DFN model to determine characterization and assessment of block-forming
discontinuity size and intensity of the geological geological discontinuities during caving of primary
stockwork veins. The DFN model was built using the copper ore at the El Teniente mine, Chile. Int. J. Rock
commercial FracMan software package. Mech. Min. Sci. and Geomech, Abstr. pp. 44, pp.565-
DFN models require, as a minimum data, the orientation, 583.
size and intensity of the veins to be defined: 3. Brzovic, A., Villaescusa, E. & Figueroa, C. (2006)
Vein orientations in the model were built by Characterization of block-forming geological
bootstrapping observed veins orientations in three discontinuities during primary ore caving at the El
oriented cores places in the three semi-orthogonal Teniente mine, Chile. IN YALE, D., HOLTZ, S.,
BREEDS, S. & OZBAY, U. (Eds.) 41st US Symp. on
orientations. Chi-Square test values of the
Rock Mechanics. Golden Colorado, ARMA.
correlation between observed and simulated
4. Zhang, L. & Einstein, H. H. (2000) Estimating the
intensity of rock discontinuities. Int. J. Rock Mech.
Min. Sci. and Geomech, Abstr., 37, pp. 819-837.
5. La Pointe, P., Wallmann, P. & Dershowitz, W. S.
(1993) Stochastic estimation of fracture size through
simulated sampling. Int. J. Rock Mech. Min. Sci., Vol.
30, pp. 166-1607.
6. Dershowitz, W.S., Lee, G., Foxford, T., La Pointe, P. &
Thomas, A. (1995) FracMan Interactive Discrete
Fracture Data Analysis, Geometric Modeling, and
Exploration Simulation. User documentation. Version
2.5. Goder Associates Inc. Report 923-1089. Golder
Associates Inc, Seattle.
7. Rojas, E., Cavieres, P., Dunlop, R. and Gaete, S.
Control of induced seismicity at El Teniente mine,
Codelco Chile. In Chitombo, G., ed. Proceeding
Massmin 2000, AusIMM, 777-781 (Brisbane, 2000).
8. Skewes, M. A., Arevalo, A., Floody, R., Zuniga, P. &
Stern C. R. (2006) The El Teniente megabreccia
deposit, the worlds largest deposit. In: Porter TM,
editor. Super porphyry copper & gold depositsa
global perspective. Adelaide, Australia.
9. Efron, B. (1979) Bootstrap methods: another look at the
jacknife. In The Annals of Statistics, vol. 7, pp. 1-26.
10. Villaescusa, E. (1991) A three dimensional model of
rock jointing. PhD thesis (unpublished). University of
Queensland, Brisbane.
11. Dershowitz, W. S. & Herda, H. H. (1992) Interpretation
of fracture spacing and intensity. IN TILLERSON, J. R.
& WAWERSIK, W. R. (Eds.) 33rd US Symp. on Rock
Mechanics. Santa Fe, New Mexico, A.A.Balkena.
12. Herrera, S. (2011) Modelos Estructurales Estocsticos,
su aplicacin en la caracterizacin geotcnica en mina
El Teniente. Bachelor Geology thesis (in preparation).
Universidad de Concepcin, Concepcin.
13. Roger, S. (2011) Power point presentation. Personal