You are on page 1of 8

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/303988565

Sampling Technique Biases and their Effect on Discrete Fracture Network


Generation for Underground Works using LiDAR Scanning

Conference Paper · October 2016

CITATIONS READS

4 209

3 authors:

Ioannis Vazaios Nicholas Vlachopoulos


Arup Royal Military College of Canada
52 PUBLICATIONS   184 CITATIONS    171 PUBLICATIONS   888 CITATIONS   

SEE PROFILE SEE PROFILE

Mark S. Diederichs
Queen's University
215 PUBLICATIONS   5,370 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Laboratory direct shear testing of rock fractures View project

Monitoring, Instrumentation, Analysis and Design for Geotechnical Engineering Projects View project

All content following this page was uploaded by Ioannis Vazaios on 28 April 2017.

The user has requested enhancement of the downloaded file.


Sampling Technique Biases and their Effect on
Discrete Fracture Network Generation for
Underground Works using LiDAR Scanning
I. Vazaios, N. Vlachopoulos, M.S. Diederichs
Department of Geological Sciences and Geological Engineering,
Queen’s University, Kingston
Ontario, Canada, K7L 3N6

ABSTRACT
Structural mapping is a key element that provides valuable data such as discontinuity orientation, density, persistence,
fracture trace length etc. which serve as input to the geometrical modelling of a rockmass. However, in an active
underground environment, traditional joint survey techniques, and more specifically, structural mapping of the tunnel walls,
may become difficult. LiDAR scanning in an operational underground environment can be proven to be more efficient, as
the required collection data time and impediment to the construction is minimized as processing of the structural rockmass
data can be conducted at a later time. However, similar to traditional mapping, “virtual” mapping is subjected to the same
biases due to the inherent limitations of the sampling techniques employed. This affects the discontinuity data which is
then used as input to the geometrical modelling of the fracture network (such as the use of Discrete Fracture Networks
(DFNs)), which is critical in order to establish a reliable, geologically and geomechanically sound “picture” of any
underground project. This paper will focus on the procedures associated with utilizing LiDAR data in order to aid in the
characterization of the rockmasses past the face of tunnelling projects. The workflow and results will be discussed utilizing
data sets obtained from a tunneling project in Norway.

RÉSUMÉ
La cartographie structurale est un élément clé qui fournit des données précieuses telles que l’orientation des discontinuités,
la densité, la persistance, la localisation des fissures et fractures, etc. L’ensemble de ces données est utilisé et nécessaire
à la modélisation d’une masse rocheuse. Cependant, dans un environnement souterrain actif, les techniques de relevés
conventionnelles, et plus spécifiquement, la cartographie structurale des parois d’un tunnel peuvent s’avérer difficiles. À
l’opposé, un balayage LiDAR dans un environnement souterrain opérationnel peut s’avérer plus efficace en raison de la
rapidité de la collecte des données qui diminue les entraves aux opérations. La collecte de données est rapide puisque le
traitement des données de la masse rocheuse peut être effectué ultérieurement. De plus, semblable à la cartographie
conventionnelle, la cartographie virtuelle est soumise aux mêmes inconvénients en raison des limites inhérentes aux
techniques d’échantillonnage utilisées. Cela entraine une interruption dans le flux de données qui sont ensuite utilisées
dans la modélisation géométrique du réseau de fracture (telles que l’utilisation d’un réseau de fractures discrètes), ce qui
est critique à la création d’une image sonore fiable géologiquement et géo mécaniquement de tout projet souterrain. Cet
article porte sur les procédures associées à l’utilisation et à la gestion de données LiDAR afin d’optimiser la caractérisation
d’une masse rocheuse une fois l’acquisition des données complétées. Des méthodes et résultats seront présentés en
utilisant des données provenant d’un projet antérieur de tunnel en Norvège.

KEYWORDS: DFN, LiDAR, Underground Projects, Sampling Biases

1 INTRODUCTION the numerical tools that has been developed in order to


assist in this is the generation of Discrete Fracture
Exposed rockmass surfaces are used to characterize the Networks (DFNs).
fracture patterns present at a specific site with the However, DFN modelling relies highly on the selected
expectation that obtained discontinuity data will provide an input parameters and therefore, the adequate
insight and enhance the understanding of the fracture characterization of the rockmass is critical in order to
conditions of the rockmass under study. Especially when it generate models that are geologically reasonable and
comes to underground projects, the evaluation and correspond to the is-situ conditions of the investigated site.
estimation of the rockmass conditions can allow for Light Detection and Ranging (LiDAR) systems can be
versatile and efficient modifications in the proposed support rather efficient in such cases due to their ability to produce
system or the excavation method employed. detailed 3D models of exposed rockmass surfaces and
In-situ characterization combined with techniques such assist in obtaining geotechnical data in an efficient manner.
as the geometrical modelling of a rockmass can be However, this approach is subjected to biases similarly to
particularly useful, as different generated models based on traditional discontinuity sampling techniques which affect
the rockmass characterization can provide a wide range of the DFN realizations, and therefore the estimate of the
different expected fracture conditions and therefore rockmass conditions and the design of the underground
different support systems in order to manage these. One of project.
This paper will focus on the procedures associated with surface (Mauldon & Mauldon 1997). All of these factors
utilizing LiDAR data in order to aid in the characterization influence the measured quantities and may lead to skewed
of the rockmasses past the face of tunnelling projects. The input data, not appropriate for DFN modelling.
workflow and results will be discussed utilizing data sets Furthermore, during DFN generation the size of the
obtained from a tunneling project in Norway. fractures has to be determined. However, this is a rather
difficult task as measuring or estimating directly is not
2 DFN GENERATION possible. However, indirect estimations can be made by
measuring the trace lengths on the exposed rockmass
DFN model realizations serve as explicit representations of surfaces. In a 2D mapping window or plane, the length
finite size fractures in 2D or 3D space, used to simulate distribution of the traces is usually assumed to be
discontinuities in a synthetic rockmass. In order to representative of the average fracture size distribution.
extrapolate, modify and expand collected data to produce However, field data is usually truncated at some lower
a synthetic fracture network, the generation of DFNs bound, as below that it is either considered unimportant or
usually relies on statistics derived from accumulated field difficult to determine, which may result in skewed input
data for a given project. Discontinuity orientation, fracture parameters for the generation process. Additionally, very
size, intensity, spatial variation persistence and aperture short fractures are less likely to intersect the tunnel wall
are typical input parameters in DFN modelling and are surfaces, even when more, than larger fractures skewing
usually collected by applying traditional techniques, the data towards the larger ones.
including manual joint mapping and boreholes, or more As stated above, it becomes evident that the area of the
sophisticated techniques, including virtual mapping of 3D exposed rockmass is crucial as it controls the size of the
surface models created by LiDAR or photogrammetry data. sampling windows and therefore the availability of data
However, the access to the rockmass or its exposure may which can be used in a DFN generation process, especially
be limited; hence, each of these techniques are subjected for discontinuity sets that that are more widely spaced
to limitations leading to significant challenges during the (Palleske et al. 2013). Furthermore, due to construction
data collection, which also highly depends on the reasons and restrictions depending on the cycle of works
specifications of the investigated site, resulting in less in a tunneling environment, access to the exposed
complete, taken over smaller ranges or even skewed data. rockmass may be limited. Moreover, tunnels are long
This creates conditions which are not ideal for a underground structures advancing towards a specific
comprehensive rockmass characterization. direction and consequently the exposed rockmass
surfaces. Therefore, data collection is impacted by two
2.1 DFN Input Parameters and Related Biases in major factors: a. the total area extent of the surface to be
Underground Works mapped; and b. the maximum size of the surface in a given
direction.
As mentioned above, discontinuity data collection is critical
for DFN modelling; however, significant challenges may 3 LIDAR SCANNING IN UNDERGROUND WORKS
rise depending on the employed sampling techniques and
the site specific requirements. Especially in a tunneling LiDAR has already been proven as a real asset for slope
operation environment, this may lead to misleading data surveying and hazard assessment and has been
during the collection process, resulting in DFNs that are not extensively used for rockmass structural characterization
representative of the rockmass simulated. at the rock outcrop scale. The same principles can be
Regarding orientation, the spacing of joints of all sets employed at a certain extent to underground applications
observed in a tunnel relative to the tunnel diameter can as well, at unsupported sections of tunnels where the
significantly influence the number of observed joints, as if rockmass is exposed (Fekete & Diederichs 2013).
the spacing of blind-zone joints is close to or greater than Utilizing a transmitted and reflected laser beam, the
the tunnel diameter, there will inevitably be a scanning results in the recording of millions of high
disproportionately small number of observations of the accuracy points in space with co-ordinates relative to the
orientation of blind zone joints (Terzaghi 1965) and position of the scanner known as “point-cloud”. Further
therefore the collected data that will be used as an input for process of the point cloud results in a 3D surface model as
the DFN generation process. illustrated in Figure 1. One of the major advantages of such
Fracture density and fracture intensity also serve as a model is that discontinuities and fracture traces can be
input parameter for DFN modelling and can be derived by mapped explicitly within the model and therefore
employing window mapping techniques on the tunnel walls. discontinuity data including joint orientation, fracture
The combination of number of fractures combined with a density and fracture trace length can be extracted and used
total fracture length in an area, organized by set in total, in DFN modelling (Vazaios et al. 2015).
provides a less subjective and a more complete description One of the key strengths of using LiDAR scanners in
of the rockmass in 2D (Palleske et al. 2013). However, underground projects is that they can function without
window mapping is subjected to limitations due to sampling underground lighting, since the laser acts as its own “light
problems such as censoring and truncation and therefore source”. This can be rather advantageous when compared
depends on the excavated length of the tunnel. to photogrammetric or traditional method mapping
Additionally, if the tunnel diameter is sufficiently large in techniques. Additionally, high-speed phase-based
relation to the size of the fracture, a fracture may be scanners have led to decreased scanning times which
contained inside the tunnel, without intersecting the tunnel allow their use in active tunnels (Fekete et al. 2010); hence
a better integration into the construction cycle of a tunnel 3.2 Deriving Input Parameters for DFN Modelling from
than conventional mapping may be achieved. LiDAR Data

After identifying the discontinuity surfaces and fracture


traces within the 3D surface model, discontinuity related
data can be extracted and used for DFN modelling serving
as input parameters. Discontinuity orientation can be
determined by fitting planes to selected areas (Figure 2a).
The normal of these planes determines the dip and dip
direction of the joint surfaces. Pole contouring of the
mapped discontinuities results in identifying the joint
“families” present and an average orientation or an
orientation distribution can be derived and used as input for
the DFN generation process. In Figure 3a, plane orientation
data from the 3D surface model of the Bånkall Tunnel is
used as input to Dips (Rocscience 2015) in order to
Figure 1. 3D Surface model created using LiDAR scanning determine the dominant discontinuity groups for that
data of the Bånkall Tunnel, Norway. specific rockmass.
Fracture intensity is another discontinuity property
3.1 Virtual Mapping of Discontinuities in Tunneling usually serving as an input parameter for DFN modelling
Environments and provides information regarding the fracture frequency
and the fracture size. Fracture intensity in 2D (areal fracture
Surface models are used to assess a rockmass as its intensity P21) is estimated by employing window mapping
structural features can be identified and mapped explicitly techniques from fractures traces. As illustrated in Figure
within them including joint surfaces and fracture traces and 2b, fracture traces can be explicitly mapped within the 3D
have been successfully used in slope and outcrop surface model and therefore can be used to determine
characterization. Similar principles can be employed in fracture intensity. MoFrac (Mirarco 2014), a stochastic DFN
underground applications where the unsupported span of generator used for the purposes of this paper, employs the
a tunnel can be scanned, create the surface model and best-fit line of a fracture intensity distribution in order to
“map” the discontinuities (Fekete & Diederichs 2013, Vöge define the size and density of the fractures created within a
et al. 2013) (Figure 2a) and the fracture traces on the tunnel volume of interest (Figure 3b).
walls (Figures 2b) using Polyworks (InnovMentrics 2009),
a general purpose package for range for range image
processing.

Figure 3. Discontinuity and fracture properties derived for


Figure 2. Virtual mapping of a. discontinuity surfaces the Bånkall Tunnel from LiDAR data including a.
(highlighted with a red ellipse and a blue rectangle), and b. Discontinuity orientation and pole contouring (Set 1-red,
fracture traces (red) within the 3D surface model of the Set 2-blue, Set 3-yellow) and b. Fracture intensity (Set 1).
Bånkall Tunnel, Norway.
4 THE BÅNKALL TUNNEL CASE For the purposes of this paper, the DFN generator
MoFrac, which enables the user to generate 3D DFNs
The Bånkall Tunnel located in Oslo area, Norway was using explicitly mapped fractures (Figure 2b) seeding the
excavated in a massive plutonic rock, Permian in age. More fracture propagation algorithm based on collected
specifically, the tunnel was excavated in quartz-bearing discontinuity data. Therefore, creating a dataset that will
with grey plagioclase syenite (Figure 4). The tunnel is an serve as input for the generation process is of great
approximately 65m long and of a diameter of 6m testing importance since it will control the number and size of the
facility excavated by employing the drill and blast method fractures which are going to be generated. In Figure 5, one
(Fekete et al. 2010). of the DFN models generated for the purposes of this paper
is illustrated.

Figure 5. Generated DFN model of the Bånkall Test Tunnel


for a volume of interest 30mx60mx30m. Three different
joints sets can be identified based on different coloring (Set
1-red, Set 2-blue, Set 3-yellow).

4.2 Fracture Data Sampling and Input Parameter


Selection

Four main sampling strategies for collecting discontinuity


and fracture data are the usual state-of-practice and
reported in the literature including the linear scanline
method (Priest and Hudson, 1981; Priest, 1993), areal
sampling (Wu and Pollard, 1995), rectangular window
sampling (Pahl, 1981; Priest, 1993) and the circular
scanline method (Mauldon et al., 2001; Rohrbaugh et al.,
2002). These field techniques can be transferred and used,
Figure 4. The Bånkall Test Tunnel in massive quart-bearing based on the same principles, along with LiDAR data in
syenite located in the wider Oslo area, Norway. order to obtain input parameters for DFN models. In this
particular case for the Bånkall Tunnel a rectangular window
4.1 DFN Generation for the Bånkall Tunnel Site sampling strategy is applied, as illustrated in Figure 6.
Rectangular window sampling utilizes a rectangular area,
The selection of representative input parameters has a which is placed on the exposed rockmass (outcrop, tunnel
significant impact on DFN generation, and equally wall etc.) and selected discontinuity attributes are
important is the generation method applied for the creation measured within the area of the rectangle (Priest, 1993;
of a fracture network. In order to examine the effect of the Watkins et al., 2015). Rectangular window sampling
selected input parameters and therefore the discontinuity reduces orientation bias and allows for a simple estimation
mapping and sampling biases on the generation of DFNs, of mean trace length (Pahl, 1981). However, the data is
the Bånkall Tunnel case is examined and demonstrated in subjected to censoring, due to the size of the exposed
this section. For these purposes the total length of the rockmass area, and the quality of the exposure (Watkins et
tunnel was scanned and the generated point cloud was al., 2015). Within the indicated windows in Figure 6, joint
used. Processing of RGB images or intensity values was surface orientation and fracture trace length
not conducted and the results rely solely on the geometrical measurements were taken. The main objective is to
information obtained from the scanning process. examine how the collected data from windows of 10m,
Figure 6. Rectangular window sampling along the right wall
springline of the Bånkall Tunnel. Five mapping windows of
10m (red), 20m (yellow), 30m (green), 40m (blue), and 50m
(orange) are employed in order to estimate the
discontinuity properties of the 10m-Target Window (white).

20m, 30m, 40m, 50m assist in the estimation of the


discontinuity properties of the Target Window (highlighted
with white colour in Figure 6). It has to be noted that the
mapping windows have a constant height at all times which
is the height of the tunnel and equal to h=5.60m.

4.2.1 Orientation Measurements

Orientation measurements were taken for every window in


order to determine the orientation of the mean planes that
would be used as input for the DFN generation. From these
mean planes three joint sets are dominant for this specific
rockmass, with Set 1 being approximately parallel to the
tunnel axis (850/2720), Set 2 being perpendicular to the
tunnel axis (810/300), and Set 3 being approximately
horizontal (210/1220). In order to take into account possible Figure 7. Deviation of a. Dip and b. Dip Direction between
orientation biases due to the orientation of the tunnel axis the mean planes obtained from orientation measurements
the Terzaghi correction method was also applied (Terzaghi, from the mapping windows and the Target Window.
1965). However, the obtained orientation values did not
differ significantly and therefore the originally estimated 4.2.2 Fracture Intensity Measurements
ones were used. Mapping of discontinuities on the other
sidewall revealed that one more discontinuity set is Following the orientation measurements and the
present. However, its existence was not evident on the right identification of the dominant discontinuity sets within the
wall of the tunnel which was under investigation and was rockmass, fracture intensity measurements were also
not included in the modelling process. conducted. In this particular case, the areal fracture
In Figure 7a and 7b, the collected orientation data are intensity as defined by Dershowitz and Herdra (1992), P21,
demonstrated in order to examine how the accumulation of was measured for each joint set.
data, as the mapping window length increases, affects the For the determination of the intensity, the total length of
dip and dip direction of the mean planes of all three joint the fracture traces per set was measured within the
sets when compared to the mean planes of the Target sampling area and the results are demonstrated in Figure
Window. From these it can be observed that for Set 1 the 8. From this it can be observed that for Set 1 the fracture
accumulation of additional data, as the mapping window intensity does not fluctuate a lot for all different window size
length increases, helps improve the estimation of its dip cases. For Set 2 it seems that for a sampling lengths equal
and dip direction as the deviation between the accumulated to or greater than 20m, the intensity remains approximately
data and the data of the Target Window decreases. The constant. Therefore, P21 can give relatively good
same can be inferred for Set 2 at some extent as well. For estimations for the fracture intensity of these two joint sets.
Set 3, the dip direction shows a similar trend. However, the However, for Set 3 the areal fracture intensity is strongly
dip of Set 3 does not have a similar trend and in its case affected by the size of the sampling area due to its
the additional data seem to add more uncertainty in the orientation (Set 3 is an approximately horizontal fracture
estimation of the dip for Set 3, which can be explained by set) and due to the fact that P21 is a fracture intensity
taking into account the geological and geotechnical measure that is direction dependent (Dershowitz and
conditions of the specific site. Herdra, 1992).
For Set 2 in Figure 10, P21 has a downward trend as the
sampling length increases for lengths greater than 30m it
becomes consistent and has a value of P21=1 m-1. The
deviation between the simulated and the observed in-situ
conditions also shows a downward trend, and it gets
consistent at around 25% for sampling lengths greater than
30m. In this case, the orientation of Set 2 relative to the
tunnel axis favors the sampling process and enhances the
input data as more of it is sampled by increasing the
window mapping size. While the deviation can still be
Figure 8. Areal fracture intensity as measured from considered a little high, this is probably the result of
mapping windows of 10m, 20m 30m, 40m and 50m length assigning a lower bound fracture size for our numerical
along the tunnel axis of the Bånkall Tunnel. analysis. This favors the creation of larger fractures which
increases the areal fracture intensity and may not reflect
4.3 DFN modelling in Estimating Fracture Conditions and exactly the in-situ fracture intensity. Hence, an additional
the Effect of the Selected Input Parameters bias may be introduced in the numerical model, associated
with the generation process though.
In order to investigate the impact of the input parameters
derived from LiDAR data for different sampling windows in
the estimation of the fracture conditions, different DFN
models based on the prescribed sampling windows were
generated. Based on the numerical modelling that was
conducted, cross-sections were created from the DFN
models and the Target Window section was examined in
order to compare the fracture intensity of the numerical
models with the in-situ observed fracture intensity. The
results are demonstrated in Figures 9, 10 and 11 for Set 1,
Set 2 and Set 3 respectively.
For Set 1 in Figure 9, it can be observed that P21 varies Figure 10.Simulated areal fracture intensity P21 for Set 2
within the range of 1.5 and 2.5 m-1 as the sampling length and its deviation from the observed in-situ intensity for the
increases and it becomes more consistent for that specific Target Window.
section. However, the intensity of the models does not
correspond to the in-situ conditions, as the deviation from Following the aforementioned, in Figure 11 the results for
the targeted value is fluctuates between 70% and 80% for Set 3 can be observed. In this case the areal fracture
sampling lengths greater than 30m. In this case, the update intensity P21 has an upward trend for sampling lengths
of the model did not contribute in making a better estimate between 10m and 20m. However, again for sampling
for the expected fracture conditions. On the contrary, it can lengths greater than 30m P21 becomes approximately
be inferred that the fracture conditions within the Target consistent at a value around P21=0.70m-1. Regarding the
Window section change locally. This cannot be taken into deviation between the simulated intensity and the in-situ
account by the model based on the previously collected intensity of the Target Window, a downward trend can be
data, and therefore, the estimation of the fracture observed for sampling lengths of 10m and 20m
conditions for this specific set is rendered difficult. respectively. For lengths greater than 30m it ranges
Additionally, due to the relative orientation between Set 1 between 5% and 20%. As in the case of Set 2, the
and the tunnel alignment, sampling within the specific orientation of the set relatively to the tunnel alignment
section can be challenging and the observed fracture allows for an adequate sampling process, and the increase
traces may underestimate the in-situ fracture intensity. in the mapping length assists in updating the input
parameters for the DFN generation process resulting in the
better estimation of the in-situ fracture conditions of the
Target Window section.

Figure 9.Simulated areal fracture intensity P21 for Set 1 and


its deviation from the observed in-situ intensity for the
Target Window.
Figure 11.Simulated areal fracture intensity P21 for Set 3
and its deviation from the observed in-situ intensity for the
Target Window.
5 CONCLUSIONS Fekete, S. and Diederichs, M.S. 2012. Integration of three-
dimensional laser scanning with discontinuum
Estimating fracture conditions during the design of an modelling for stability analysis of tunnels. Int. J. of Rock
underground project is a really important step in Mech. & Min. Sciences. 57 (2013): 11-23.
determining the required support system. In order to Mauldon, M. and Mauldon, J.G. 1997. Fracture Sampling
achieve this, a process of combining remote sensing on a Cylinder: From Scanlines to Boreholes and
techniques along with numerical tools was demonstrated. Tunnels. Rock Mechanics and Rock Engineering,
More particularly, DFN modelling can assist in 30(3): 129-144.
estimating the rockmass conditions, as discontinuities can Mauldon, M., Dunne, W.M., Rohrbaugh, M.B., 2001.
be modelled and multiple geometrical models can be Circular scanlines and circular windows: new tools for
created for a specific site in order to examine possible characterizing the geometry of fracture traces. J. Struct.
fracture patterns and their effect on the tunnel design. Geol., 23: 247-258.
However, DFN modelling highly relies on the selected Pahl, P.J., 1981. Estimating the mean length of
input parameters used for the generation of the models. discontinuity traces. Int. J. Rock Mech. Min. Sci.
LiDAR can be used to extract discontinuity related data Geomech. Abstr., 18: 221-228.
which can serve as the basis of the required modelling Palleske, C., D.J. Hutchinson, D. Elmo, M.S. Diederichs.
input parameters. The “virtual mapping” of joints within 2013. Impacts of limited data collection windows on
LiDAR models is subjected to sampling biases, similar to accurate rock simulation using discrete fracture
traditional mapping techniques. This has an impact on the networks. In Proceedings of the 47th US Rock
generated DFN models, and therefore it affects the Mechanics/Geomechanics Symposium, San
estimation of the rockmass conditions when using these Francisco, 23 – 26 June 2013.
numerical models. In order to highlight this the case of the Priest, S.D. 1993. Discontinuity Analysis for Rock
Bånkall Tunnel was examined. Engineering. Chapman & Hall, London, United
The results of the DFN modelling for different sampling Kingdom.
windows were illustrated and the effect of the geological Priest, S.D. and Hudson, J.A. 1981. Estimation of
conditions, the abundancy of structural data and the related discontinuity spacing and trace length using scanline
input parameters, and the generation process on the surveys. Int. J. Rock Mech. Min. Sci. Geomech. Abstr.,
estimated fracture conditions was discussed. Unfavorable 18: 183-197.
orientation and locally changed fracture conditions for Set Rohrbaugh, M.B., W.M., Dunne, M., Mauldon 2002.
1 in this specific case introduced biases that affected the Estimating fracture trace intensity, density, and mean
simulation process of the fractures, and therefore the length using circular scan lines and windows. AAPG
estimation of the fracture conditions was not judged Bull. 86 (12): 2089-2104.
satisfactory, even when the DFN model was updated with Terzaghi, R.D. 1965. Sources of error in joint surveys.
the addition of new data. On the contrary, for Sets 2 and 3, Geotechnique, 15(3): 287 - 304.
the favorable orientation conditions relative to the tunnel Vazaios, I., N., Vlachopoulos, M.S. Diederichs. 2015. A
axis, created the conditions for adequate and accurate study of the geometrical scale-dependency of fractured
sampling. This resulted in the additional data obtained, due rockmasses using LiDAR scanning: The case study of
to the increase of the mapping window size, to assist in the Brockville Tunnel. In Proceedings of the 13th
estimating the fracture conditions as the DFN model was International Congress on Rock Mechanics: ISRM
updated by modifying the used input parameters. This Congress 2015, Montreal, 10 – 13 May 2015.
highlighted how DFN modelling is sensitive to the selected Vöge, M., M.J., Lato, M.S., Diederichs 2013. Automated
input parameters and how specific factors introduce biases rockmass discontinuity mapping from 3-dimensional
that affect these, and the therefore the DFN generation surface data. Engineering Geology, 164: 155-162.
process. Watkins, H., C.E., Bond, D., Healy, R.W.H., Butler 2015.
Appraisal of fracture sampling methods and a new
ACKNOWLEDGEMENTS workflow to characterise heterogeneous fracture
networks at outcrop. Journal of Structural Geology, 72:
The authors would like to thank the Nuclear Waste 67-82.
Management Organization of Canada, the National Wu, H., Pollard, D.D., 1995. An experimental study of the
Science and Engineering Research Council and the relationship between joint spacing and layer thickness.
Department of National Defence who have supported this J. Struct. Geol., 17: 887-905.
work. InnovMetrics. PolyWorks, V11.0.4 Quebec City, 2009.
Mirarco. Mofrac, V1.0, Sudbury, 2014.
REFERENCES Rocscience. Dips, V6.016 Toronto, 2015.

Dershowitz, W.S., Herda H.H. 1992. Interpretation of


Fracture Spacing and Intensity. Proc. 32nd US Rock
Mech. Symp., Santa Fe, NM, 757-766.
Fekete, S., M.S., Diederichs, M.J., Lato 2010. Geotechnical
and operational applications for 3-dimensional laser
scanning in drill and blast tunnels. Journal of Tunneling
and Underground Space Technology, 25, 614-628.

View publication stats

You might also like