You are on page 1of 12

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

net/publication/282913908

DFN generation for Mechanical Stability Analysis of Underground Works

Conference Paper · May 2015

CITATIONS READS

4 402

3 authors:

Ioannis Vazaios Nicholas Vlachopoulos


Arup Royal Military College of Canada
54 PUBLICATIONS   354 CITATIONS    182 PUBLICATIONS   1,280 CITATIONS   

SEE PROFILE SEE PROFILE

Mark S. Diederichs
Queen's University
258 PUBLICATIONS   7,199 CITATIONS   

SEE PROFILE

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

The Use of Technology in order to Enhance Student Learning View project

The Use of Fiber Optics in order to Determine the Behaviour of Ground Support Elements View project

All content following this page was uploaded by Nicholas Vlachopoulos on 23 July 2016.

The user has requested enhancement of the downloaded file.


„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

DFN generation for Mechanical Stability Analysis of Underground


Works

Author: Ioannis VAZAIOS, Queen’s University, Canada, ioannis.vazaios87@gmail.com


Co-author: Nicholas VLACHOPOULOS, Queen’s University, Canada, vlachopoulos-n@rmc.ca
Co-author: Mark S. DIEDERICHS, Queen’s University, Canada, diederim@queensu.ca

Topic: Planning and Designing Tunnels and Underground Structures

A first step in the characterization of fractured rock materials and its anticipated mechanical
response during an underground excavation is the static mapping and modelling of the subsurface
fracture system. This task is rendered difficult by the lack of truly 3D data about the subsurface
fracture geometry. Yet, it is critical to be able to establish a reliable, geologically and
geomechanically sound characterization for a tunnelling project. The use of Discrete Fracture
Networks (DFNs) is becoming increasingly more common in geomechanical practice; as such
models can serve as useful tools for estimating the interconnectedness of the discontinuities.
However, the success of reliable DFN creation and their application in mechanical stability analysis
is based on the input parameter selection and model reconstruction methodology. Due to the
significant influence of the joints with respect to tunnelling, input parameter selection (orientation,
persistence, spacing etc.) for each discontinuity set involved in an analysis has to be reasonable
and correspond to the in-situ field conditions. This depends on the reliability, precision and
measurement sample scale and the applied techniques for accumulating it, including error
correction and bias removal. After the initial discontinuity data has been synthesized, the applied
methods for the model reconstruction include a deterministic approach, matching observed
conditions at a local site where geospatial information is comprehensive and accurate, and a
stochastic approach, blending discrete observations and statistical information to fill in for the lack
of information and available data. The generation of mechanically valid DFNs for the mechanical
analysis of underground project can be rather challenging, as significant issues arise both in the
data collection and the simulation processes. The aforementioned issues and methodologies will be
discussed and compared using data sets obtained from the Brockville Tunnel located in Ontario,
Canada.

Keywords: DFN, mechanical stability, ground modelling, discontinuum

1. Introduction
Underground projects including rail and road tunnels, water conveyance tunnels, hydro-power
caverns, deep geological repositories etc., are a vital part of the modern society’s infrastructure.
However, imposed constraints due to limited budget or time have to be balanced while
maintaining high engineering standards and safety. In order to efficiently excavate in rock, both in
terms of time and cost, a thorough understanding of the rockmass structure, properties and
behaviour is required. The evaluation and estimation of the rockmass conditions allows for
versatile and efficient modifications of the temporary support or the excavation method employed
as changes occur. One of the numerical tools that have been developed in order to assist in this is
the generation of Discrete Fracture Networks (DFNs), which essentially is the numerical simulation
of the fracture network of a rockmass based on field accumulated data and deterministic and
stochastic processes.

Particularly for underground projects such as tunnels, the generation of geologically and accurate
site-specific DFNs for mechanical analysis is gaining ground over conventional equivalent material
techniques based on heuristic structural classification systems. The success of DFN creation and
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

application is based on two significant factors: a. the selection of the appropriate input parameters
and, b. the model generation methodology. Input parameter selection (mean orientation,
persistence and spacing specified along with variability descriptors) for each discontinuity set
involved in an analysis has to be reasonable and correspond to the in-situ, field conditions. This
depends on the variability, precision and measurement sample scale and resolution for field data
as well as on the applied techniques for collecting it (i.e. manual mapping, LiDAR, boreholes etc.).
However, even with the best available investigation data, it also requires a more in depth
knowledge of the geologic history and the actual genesis mechanisms of the rock formations and
their discontinuities, leading to challenges and limitations for each specific site. The applied
methods for the generation of the model, after the initial discontinuity data has been synthesized,
include a deterministic approach matching observed conditions at a local site, or a more statistical
approach, blending discrete observations and creating a stochastic model of structure [1].

Light Detection and Ranging (LiDAR) systems have been widely adopted due to their detailed
geotechnical data collected in an efficient manner. Data from laser scans produces very detailed
geometric models of the scanned surfaces with their spatial resolution being of the order of
millimetres. This paper will demonstrate ways to implement LiDAR scanners in underground
excavation as part of a methodology of blending this remote sensing system with DFN generation
in order to estimate the rockmass conditions far within the excavation of the specific site of the
Brockville Tunnel.

2. DFN Generation and the Brockville Rail Tunnel


Discontinuity orientation, fracture size and intensity are some of the critical parameters in DFN
modelling. Such data are usually collected by applying traditional techniques such as manual
mapping and borehole data or more sophisticated ones including virtual mapping of LiDAR and/or
photogrammetry scanning data. However, each of these techniques are subjected to limitations as
either the access or the exposure of the rockmass is limited [2]. This leads to significant
challenges during data collection and, depending on each specific site; this may result in data
which may be less complete or taken over smaller ranges that are not the ideal for a
comprehensive determination of rockmass characterization beyond the tunnel face.

Data extracted by LiDAR from the Brockville Railway Tunnel was used in order to determine the
discontinuity orientation, fracture length and density as input parameters as part of the DFN
generation process. Constructed between 1854 and 1860, the Brockville Tunnel has a total length
of 520m and is the oldest railway tunnel in Canada. The tunnel has primarily an arch-shaped
cross-section, measuring 4.25x4.50m. The geology of the area is predominantly Precambrian
quartzite (paragneiss) (Figure 1). Granitic gneiss or syenite is also possible as the tunnel sits in a
folded regime with these three dominant lithologies [3]. However, later bedded Ordovician
sedimentary rocks (sandstones, dolomites and limestones) are present and lie unconformably
above the tunnel or nearby to the east. A pervasive joint set running parallel to the tunnel axis
dominates the rockmass behaviour. The predominant natural fracture patterns are sub-vertical,
striking within 30 degrees of the tunnel axis. A dominant horizontal discontinuity set in the tunnel
roof also exists. The combination of these two joint sets results in the squaring of the tunnel
profile in sections.
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

Fig. 1 Unsupported rock section of the Brockville Tunnel [3]

2.1 LiDAR Scanning 3D Surface Model


LiDAR utilizes a transmitted and reflected laser beam and the scanning results in the recording of
millions of high accuracy points in space; hence creating a raw point-cloud which may be used for
visualization. However, if further analysis or measurements are necessary, data management and
processing of the point-cloud is required. A 3D surface model is the final product of processing the
point-cloud. In Figure 2, a 3D surface model of a total length of 90m of the Brockville is illustrated
and four different 10m sections were examined separately. Discontinuity data such as joint
orientation, fracture density and length was then directly measured virtually within the model
instead of applying manual mapping. Processing of the LiDAR data was conducted using
PolyWorks [4], a general purpose package for range image processing. Processing of the data
included the following steps [5]:
i. Raw point-cloud reduced to the area of interest.
ii. Polygonal models produced by meshing the point-cloud.
iii. Measurements and interpretation conducted on the polygonal model.

Fig. 2 Brockville Tunnel 3D surface model viewed from


the tunnel side

2.2 Discontinuity Orientation and Fracture Length Measurements


Polygonal models can be used to assess the structural features of a rockmass including joint
orientation and fracture trace lengths. In order to determine the orientation of the joints present
within a rockmass, discontinuity surfaces are extracted by fitting planes to selected areas of the
polygonal model after visual inspection of it (Figure 3a). The direction cosines of the normal of
these best-fit planes are used to determine the dip and direction of the discontinuities. By
repeating this process for all the planes that can possibly be identified as discontinuities,
stereographic projection techniques can be employed resulting in pole contouring of the
accumulated orientation sample. It has to be noted however, that this process is subjected to
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

limitations due to biases in the data accumulation process which have to be corrected if judged
necessary [6]. Plane orientation data extracted from PolyWorks can serve as input for Dips [7] and
the pole contouring results can be used to identify the dominant orientation for the rockmass
discontinuity sets (Figure 4a). Having determined the dominant joint sets, the mean and standard
deviation of the dip and direction serve as input parameters for the DFN generation process.

Fig. 3 ‘Virtual’ mapping of a. discontinuities and b.


fracture traces on the 3D surface model of a section
of the Brockville Tunnel [8].

While discontinuity orientation is usually easily determined, fracture size is one of the most difficult
parameters to estimate or measure, as only fracture trace lengths can be measured when
mapping an exposed surface. Fracture trace lengths are usually measured within 2D mapping
windows and the average fracture size distribution is assumed to be represented by the trace
distribution. However, truncation and length biases are rather common, since field data is
generally truncated at some lower bound and very short fractures are less likely to intersect the
mapping surface, and may have to be accounted for [2]. After determining the dominant
discontinuity sets, fracture traces belonging to specific sets can be identified and recorded within
the polygonal model (Figure 3b). Following the identification of the fracture traces [9], the number
of fractures and their length are estimated for a specific window, resulting in determining a
cumulative length distribution per discontinuity set (Figure 4b) as suggested by [10]. The
suggested methodology is based on specific length thresholds in order to determine the size and
the density of the fractures. MoFrac [11], which was the DFN generator used, employs the best-fit
line to this length distribution curve to define the size and density of the fractures of the various
joint sets within the volume of interest.

Fig. 4 a. Discontinuity structural analysis and b.


cumulative length distribution of mapped fractures
extracted from LiDAR data.
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

2.3 DFN Generation Method


Selection of representative input parameters has a significant impact on the generation of a
realistic DFN, however, also significant is the generation method applied for the creation of the
fracture network. For the purposes of this paper, the generation software MoFrac was used.
MoFrac enables the user to generate 3D DFNs which explicitly include mapped fractures (Figure
3b). These mapped discontinuity traces are used to ‘seed’ the fracture propagation algorithm
based on the statistical properties of discontinuity orientation, fracture density and length for each
joint set respectively. The DFN is then generated by the fracture propagation engine over a
volume of interest with its dimensions depending on the size of the project (Figure 5). However, in
order to generate a geologically accurate DFN model discontinuity, statistical properties may not
be adequate. Therefore, the natural joint generation sequence must be understood and depending
on the type of rockmass a unique approach must be adopted in order to generate a DFN
corresponding to the in-situ conditions in the area of interest.

3. Estimating Rockmass Conditions


One of the major challenges in the design and construction of underground projects is the limited
knowledge of the underground rockmass conditions. In urban tunnelling a site investigation
program can produce rather reliable results and provide in-depth knowledge of the ground
conditions due to the relatively low depth and limited area of interest. However, in many tunnelling
and other underground projects this is not the case due to the high overburden and the larger
area of interest which do not create the required conditions of an adequate and complete site
investigation scheme. In order to deal with these limitations numerical tools such as the
generation of DFNs can be used in order to assist in estimating the underground conditions to
potentially be encountered by producing different but realistic scenarios of the rockmass conditions.
Such a numerical tool may result in increasing the efficiency and effectiveness of a site
investigation program as well as assisting in the design-construction process in terms of
determining the potential support system required for a given portion of an underground project.
For the purposes of this paper, a specific tunnelling project will be examined in order to
demonstrate the potential of DFN generation and its utility in the design and construction of
underground infrastructure.

Fig. 5 Generated DFN for the Brockville Tunnel. Joint set


1 is coloured grey, Joint set 2 is blue and Joint set 3 is
red.
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

3.1 Determining the DFN Input Parameters


Due to the rockmass quality at the Brockville Tunnel site, a long part of the tunnel is unsupported.
Therefore LiDAR technology was employed in order to assess the rockmass and obtain a 3D
representation of its structural features along the tunnel axis. In Figure 2, the 3D surface model of
the tunnel is demonstrated and divided into specific sections which were examined. The 3D
surface model was used in order to extract the necessary input parameters for the DFN generation
assuming mapping windows of 5x10m. The average diameter (D) of the tunnel cross-section is
assumed to be D=5m. The tunnel is assumed to be unexcavated and three different excavation
faces are examined in order to investigate the rockmass conditions ahead of these. Therefore, the
following scenarios are developed:
i. Case 1: 3 Scans; Total scanned length 30m (6D); Excavation face 1; Investigated sections
ahead of face: 10m (2D)-Section 2, 20m (4D)-Section 3, 30m (6D)-Section 4.
ii. Case 2: 4 Scans; Total scanned length 40m (8D); Excavation face 2; Investigated sections
ahead of face: 10m (2D)-Section 3, 20m (4D)-Section 4.
iii. Case 3: 5 Scans; Total scanned length 50m (10D); Excavation face 3; Investigated sections
ahead of face: 10m (2D)-Section 4.

Based on the aforementioned, the input parameters were determined by adding data as the tunnel
“advances”. In Table 1, the joint orientation data are summarized per discontinuity set and
number of scans. As discussed earlier (Figure 4a), three joint sets are the dominant ones for this
given rockmass. Joint 1 corresponds to the sub-parallel to the tunnel axis pervasive joint, Joint 2
to the sub-vertical to joint and Joint 3 to the sub-horizontal discontinuity set. It can be inferred
that for Joint sets 1 and 2 both dip and dip direction do not vary significantly and even with a
limited number of scans these quantities can be measured accurately and be representative for
the total length of the tunnel. However, for Joint set 2 this does not seem to be the case. These
horizontal fractures have been created due to induced tensile stresses as a result of the unloading
process of ice melting. Therefore, with the principal stress orientation varying depending on the
topography, the orientation of these horizontal fractures varies as well. Hence, unless geological
constraints are imposed, pure use of orientation statistics may not be adequate to estimate the
rockmass conditions in terms of this specific joint set. However, for the purposes of this paper a
simple statistics approach was employed.

Table 1 Joint orientation parameters used as input in the DFN


generation algorithm

Case 1 Case 2
Dip (0) 0
Dip Direction ( ) Dip (0) Dip Direction (0)
Set Mean St.Dev. Mean St.Dev. Set Mean St.Dev. Mean St.Dev.
1A 84 3 89 6 1A 84 2 87 7
1B 84 3 262 4 1B 84 2 261 6
2A 78 8 10 7 2A 78 7 10 7
2B 78 8 181 10 2B 78 7 180 11
3 12 8 206 96 3 13 7 206 90
Case 3
1A 84 2 85 8
1B 84 2 261 6
2A 80 6 9 7
2B 80 6 181 11
3 12 7 199 85

As illustrated in Figure 4b, a cumulative length distribution is used to determine the density and
fracture size of the DFN by estimating the slope of this curve. Different cumulative length
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

distributions were determined for each joint set per advancement in order to investigate how
density and fracture trace length are affected as the exposure of the available for mapping
rockmass increases. The ‘best-fit’ slopes of the cumulative distribution curves are of the form of
Y=c*Xb due to the logarithmic scale used. The parameters c and b are summarized in Table 2 and
it can be inferred that for Joint set 1 and Joint set 2 the coefficient of variance of c and b is
between 3-7% and 3% respectively while for Joint set 2 it is 12% and 5%, showing that density
and fracture length is more difficult to estimate.

Table 2 Parameters c and b for the best-fit slope of the cumulative


length distribution per joint set.

Joint 1 Joint 2 Joint 3


Scans c b c b c b
3 2.396 -2.060 0.243 -1.872 0.286 -1.505
4 2.528 -2.008 0.238 -1.865 0.308 -1.459
5 2.126 -1.940 0.240 -1.904 0.336 -1.450
6 2.289 -1.945 0.224 -1.992 0.374 -1.350

3.2 DFN Outputs and Estimation of the Rockmass Conditions


DFN generation is a numerical tool that has been developed in order to simulate the fractures of
specific rockmasses depending on the available data and site by producing multiple
representations. However, these representations may not be valid unless geologic criteria govern
the generation process. For the purposes of this paper, multiple representations of each joint set
were produced in order to create a DFN that would be as representative as possible for the site
specific conditions.

In order to determine if the generated DFN is representative of the specific site conditions and
able to estimate the rockmass conditions, visual inspection was primarily conducted. Visual
inspection along the tunnel axis can assist in locating specific structures and determine if the DFN
‘captures’ them. However, this has to be enhanced by using a quantitative method which usually
includes ways to quantify fracture density and intensity for a given rockmass, such as linear, areal
and volumetric measures [12]. In this paper the areal fracture intensity P21, which is the length of
fracture traces per area of exposure, was used to investigate the specific sections by comparing
the simulated conditions with the in-situ ones.

The variation of P21 is examined per section depending on the number of scans used to
accumulate the necessary data. In Figure 6a, Section 2 was examined as the targeted part and 4
scans were used to create deterministic fractures that explicitly describe the in-situ conditions at
this section. It can be inferred that P21 for all joint sets, when using 3 scans, does not correspond
to the in-situ conditions. The reason is that the use of fracture traces close to the area of interest
when using 3 scans. The fracture generation is highly affected and the rockmass conditions cannot
be simulated properly. This becomes clearer when the deviation between the estimated value and
the actual, targeted value is examined, which is approximately 45% for Joint Set 1 and 85% for
Joint Sets 2 and 3 as illustrated in Figure 7a.

Continuing, Section 3 is examined and the estimated P21 values are illustrated in Figure 6b. In
Figure 7b, it can be inferred that the traces used for the deterministic fractures result in simulated
discontinuities that do not correspond to the in-situ, targeted conditions despite the greater
quantity of available data. However, away from the area of interest these deterministic fractures
have less impact on the area of interest and for Joint Set 3 the deviation between the simulated
and the in-situ conditions decreases to 30% and for Joint Set 1 is approximately 50%. This does
not apply for Joint Set 3 however. The availability of data has a great impact on the simulated
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

fractures which cannot capture the in-situ conditions and the deviation reaches approximately
90%.

Section 4 is the last part along the tunnel axis that was examined and the estimated P21 values are
illustrated in Figure 6c. The deviation between the estimated P21 values and the targeted values
per joint set for this section are illustrated in Figure 7c. For Joint Set 1 the deviation yields values
of approximately 30%, 20% and 35% for 3, 4 and 5 scans respectively. The simulated fractures in
the examined section in the 3-scans and 4-scans cases are not affected by the traces used. The
examined section is further in the rockmass and away from the traces creating the deterministic
fractures thus, minimizing the biases created by them, while for the 5 scans case this bias
increases and, hence, the deviation. Between the cases of 3 and 4 scans, it can be inferred that
since more data is being obtained as the tunnel advances, 4 scans result in capturing more
accurate input parameters and the simulated fractures capture the rockmass conditions more
accurately. The same conclusion can be drawn for Joint Set 3. For 3 scans the deviation is
approximately 50% due to the limited data. For 5 scans the traces used are close to the area of
interest; hence, they are affected more by the deterministically generated fractures creating a bias
leading to approximately 30% deviation. Alternatively, for 4 scans the accumulated data are
adequate and the bias created by the deterministic fractures is less resulting in a better simulation
and an approximately 25% deviation. However, Joint set 2 does not appear to have the same
trend, as for 3 scans the deviation is approximately 20% while for 4 and 5 scans is 40% and 70%
respectively. The bias created by the deterministic fractures in this case has a greater impact on
the simulated fractures, leading to a greater deviation despite the longer distance from the area of
interest at least for the 4-scan case as a result of the greater number of the recorded number of
the fracture traces.

Fig. 6 Areal intensity P21 for a. section 2, b. section 3


and c. section 4 along the tunnel axis. The targeted
P21 values are at the extreme right of each graph.
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

Fig. 7 Deviation of P21 from targeted value for a.


section 2, b. section 3 and c. section 4.

3.3 Visual Inspection of the DFN Model


In order to estimate the rockmass conditions ahead of the tunnel face, both qualitative and
quantitative measure have to be employed. In Section 3.2, it was discussed how areal fracture
intensity P21 can be used as a quantitative measure in order to compare the simulated with the in-
situ conditions by examining sections of specific length. Furthermore, a qualitative measure is the
visual inspection of the DFN model and its comparison with the 3D surface model in order to verify
specific geological structures and their locations. An example of a DFN model generated using data
from 3 scans is illustrated in Figure 8.

Fig. 8 Fracture traces from cross-sections created


from the generated DFN using 3 scans are
compared to the in-situ fracture traces a. top view
and b. side view from inside the tunnel.
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

The stochastic structural features can be observed along the tunnel axis with a hypothetical
excavation face located at the end of Section 5, where the stochastic fractures dominate the DFN
model instead of the deterministic ones. The tunnel is assumed to be unexcavated in both sections
and the simulated rockmass conditions are compared to the in-situ conditions. This can also
provide a better insight of possible gravity driven failures due to the intersection of the
discontinuities and depending on the accuracy of the DFN model the location of major wedges.
This can be proven rather crucial when determining the necessary support system both in the
design and construction process.

4. Conclusions
A realistic and geologically accurate stochastically generated DFN can be proven to be rather
useful in a mechanical stability analysis for underground projects. Estimating or predicting (as
accurately as possible) the rockmass in-situ conditions to be encountered during tunnelling
construction allows for effective design and construction of tunnel support systems.

LiDAR scanning is a powerful surveying tool with a wide range of applications in geological and
geotechnical engineering. It was highlighted in this paper how such state-of-the-art technologies
can be integrated in underground infrastructure design and construction along with DFN modelling
in order to improve them by providing a valuable tool in the estimation of the expected rockmass
conditions.

When estimating the rockmass conditions however, there are important factors affecting the
simulated conditions and thus their correspondence to the actual, in-situ field conditions. One of
the factors affecting the simulated discontinuities is the generation of the deterministic fractures of
the currently employed fracture generation process. These deterministic fractures, which originate
from the use of mapped fracture traces, create implications to the fracture generation process;
hence affecting adversely the stochastic fractures in the area interest close to the excavation face.
These leads to a significant deviation between the simulated and the in-situ field conditions,
despite the abundance of available data in some cases. In order to overcome this problem, a
recommendation would be the limited use of traces away from the area of interest that requires to
be examined in order to eliminate the effect of the deterministically generated fractures.

Another factor is the availability of discontinuity data including orientation, fracture length and
density. The availability of data as the tunnel advances affects the input parameters required for
the generation process and for joint sets that their properties change rapidly as the tunnel
advances, adding to the available data and updating it can be proven rather crucial in attempt to
estimate the conditions ahead of the tunnel face. However, the generation process currently
employed poses problems which obstruct determining the great significance of data abundance.

Limitations do exist, however, as the accuracy and proper interpretation of the relevant input
parameters as well as the geological accuracy of the generation process employed are crucial for
the creation of a realistic DFN. DFN modelling applications on rockmass conditions estimation and
mechanical stability of underground projects may be proven misleading unless specific criteria and
the geological constraints for a specific site are satisfied. Therefore, the nature of the joints within
the rockmass, their origins and the fracture mechanisms have to be thoroughly investigated and
incorporated within the generation sequence of in order to create a realistic DFN for a given site.

5. Acknowledgements
The authors would like to thank the Nuclear Waste Management Organization of Canada and the
National Science and Engineering Research Council who have supported this work. Dr. Matthew J.
Lato contributed to early LiDAR work at the Brockville Site.
„SEE Tunnel:Promoting Tunneling in SEE Region“
ITA WTC 2015 Congress and 41st General Assembly
May 22-28, 2015, Lacroma Valamar Congress Center, Dubrovnik, Croatia

References
[1] Vazaios, I., Diederichs, M.S., Vlachopoulos, N. 2014 Factors affecting realism of
DFNs for Mechanical Stability Analysis in Tunneling Environment. In Proceedings of
DFNE 2014, Vancouver, October 19 - 22, 2014.

[2] Palleske, C., Lato, M.J., Hutchinson, D.J., Elmo, D., Diederichs, M.S. 2013 Impacts
of limited spacing and persistence data on DFN modelling of rockmasses. In
Proceedings of GeoMontreal 2013, Montreal, 29 September-3 October 2013.

[3] Diederichs, M.S., Carter, M.A., Lato, M.J., Bennett, J.B., Hutchinson, D.J. 2013.
LiDAR surveying for liner condition, rock stability and reconditioning assessement of
Canada’s oldest railway tunnel in Brockville, Ontario. In Proceedings of GeoMontreal
2013, Montreal, 29 September – 3 October 2013.

[4] InnovMetrics. PolyWorks, V11.0.4 Quebec City, 2009.

[5] Fekete, S., Diederichs, M.S. 2012. Integration of three-dimensional laser scanning
with discontinuum modelling for stability analysis of tunnels. Int. J. of Rock Mech. &
Min. Sciences. 57 (2013): 11-23.

[6] Terzaghi, R.D. 1965. Source of error in joint surveys. Geotechnique. 15 : 287-304.

[7] Rocscience. Dips, V.6.010 Toronto, 2013.

[8] Vazaios, I., Vlachopoulos, N., Diederichs, M.S. 2015. A study of the geometrical
scale-dependency of fractured rockmasses using LiDAR scanning: The case study of
Brockville Tunnel. In Proceedings of ISRM Congress 2015, Montreal, May 10 - 13,
2015.

[9] Mauldon, M., Dunne, W.M., Rohrbaugh Jr., M.B. 2000. Circular scanlines and
circular windows: new tools for characterizing the geometry of fracture traces.
Journal of Structural Geology, 23 (2001), 247-258.

[10] Srivastava, R.M. 2006. Field verification of a geostatistical method for simulating
fracture network models. In the Proceedings of the 47th US Rock Mechanics
Symposium, Golden, Colorado, 17-21 June 2006.

[11] Mirarco. Mofrac, V1.0, Sudbury, 2013.

[12] Dershowitz, W.S., Hermanson, J., Follin, S., Mauldon, M. 2000. Fracture Intensity
Measures in 1-D, 2-D and 3-D at Aspo Sweden. In the Proceedings of the 4th North
American Rock Mechanics Symposium. American Rock Mechanics Association, 2000.

View publication stats

You might also like