ARTICLE IN PRESS

International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219– 228

Contents lists available at ScienceDirect

International Journal of
Rock Mechanics & Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

Estimating the potential for spalling around a deep nuclear waste repository
in crystalline rock
C.D. Martin a,, R. Christiansson b
a
Department of Civil and Environmental Engineering, University of Alberta, 3-071 Markin/CNRL Natural Resources, Engineering Facility, Edmonton, Alberta, Canada T6G 2W2
b
Swedish Nuclear Fuel and Waste Management Company, Stockholm, Sweden

a r t i c l e in fo abstract

Article history: Stress-induced brittle failure (spalling) is probable at a deep geological repository for nuclear waste in
Received 26 October 2007 crystalline rock. In the early stages of repository design it is unlikely that orientation and magnitudes of
Received in revised form the principal stresses and the rock mass strength will be accurately known. A simple methodology is
19 February 2008 developed for estimating if spalling will occur and the severity of the hazard. The methodology is
Accepted 1 March 2008
calibrated to case studies and applied to a site in Sweden. Results from the methodology are expressed
Available online 23 May 2008
in terms of a factor of safety for the mean input values and the probability of spalling based on input
Keywords: parameter distributions. It is shown based on the calibration studies that a factor of safety of 1.25 using
Brittle yielding
the mean values should be adequate to reduce the probability of yielding to negligible levels. The
Spalling
methodology is proposed as a screening tool in the early design stages of a project to identify potential
Probability of yielding
Hazard assessment
spalling problems.
Nuclear waste disposal & 2008 Elsevier Ltd. All rights reserved.
Crystalline rock

1. Introduction Brittle failure of underground openings in hard rocks is a

A major objective in the site investigation and early design
phase of an underground repository for used nuclear fuel is to
estimate if there is sufficient space for the construction of
repository that meets the safety and operational requirements.
One of the conditions that could limit the space available is
stability of the underground openings, i.e., deposition tunnels and
deposition boreholes. The two major stability issues commonly
found in hard rocks are (see [1]): (1) structurally controlled
gravity-driven instability, referred to wedges or falls-of-ground,
and (2) stress-driven instability referred to as brittle failure or
spalling (see Fig. 1).
In the hard rocks of the Fennoscandian Shield, wedge-type
failures can occur but practical experience has shown that these
wedges are effectively controlled by traditional underground
support methods such as rock bolts and reinforced shotcrete [2].
Experience at the Äspö Hard Rock Laboratory suggests that at the
depth of 400–450 m the occurrence of significant wedges is
relatively rare. Hence, at a repository site and at depths between
400 and 700 m, the proposed depth of a repository, wedge-type
failures are not considered to be a significant factor in repository
design.
 Corresponding author. Tel.: +1 403 492 2332; fax: +1 403 492 8198.
E-mail address: derek.martin@ualberta.ca (C.D. Martin).
function of the in situ stress magnitudes and the rock mass
strength. In the Fennoscandian Shield the stress magnitudes
increase with depth and hence the potential for brittle failure
also increases. In addition because the magnitudes of the principal
stress vary with direction the orientation of the underground
openings can influence the rock mass response. In the early stages
of design it is unlikely that orientation and magnitudes of the
principal stresses and the rock mass strength will be accurately
known. Hence there is a need to have a relatively simple method
for estimating the potential for brittle failure that accounts for the
variability in the unknowns. In this paper we present a
methodology for estimating the potential risk for brittle failure
and apply it to a site in the Fennoscandian Shield.
2. Risk-based methodology
Geotechnical engineering is fundamentally about managing
risk. Morgenstern [3] summarised risk assessment concepts using
the framework for risk management adopted by the Canadian
Standards Association (Fig. 2). Fig. 2 is a generalised framework
for risk management that would need to be modified for
individual projects. But in all cases risk assessment requires
identification of the hazard and quantifying the risk associated
with each hazard. Morgenstern noted that while quantitative risk
analysis (QRA) is one aspect of the framework, qualitative risk
assessment is also a valuable component of risk management.

1365-1609/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijrmms.2008.03.001
 ARTICLE IN PRESS

220 C.D. Martin, R. Christiansson / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219–228

geotechnical design. However, Stille et al. [5] suggest that this

approach is too restrictive and probability-based methods should
be used throughout the whole design and construction process. In
their report they use the following definitions:

Falling
Hazard: a source of a potential damage event.
Sliding Damage event: an event that may cause damage to humans,
properties or the environment.
Risk: the combined effect of the probability of a damage event
and the consequences caused by the damage event.

In the following sections a risk-based methodology is proposed

that follows the general principles suggested by Stille et al. [5] for
assessing risk in rock engineering. Firstly the hazards identified in
Fig. 1 are evaluated as potential damage events and the severity of
the hazard assessed by its magnitude, e.g., size of the wedge or
depth of spalling failure. This methodology to assess the wedge
and spalling hazard is illustrated in Fig. 3. The methodology is
restricted to that noted above the ‘‘risk control options’’ in Fig. 3.
For each hazard a criterion is required that can be used to
determine the frequency of occurrence. Once the frequency of the
hazard has been determined the next step is to assess the severity
of the hazard. In this paper we focus only on the spalling hazard
and describe the methodology and its application in the following
sections.

RISK ASSESSMENT FOR UNDERGROUND STABILITY

HAZARD IDENTIFICATION
Fig. 1. Illustration of the two significant modes of underground instability * Fall-of-ground (Wedge)
anticipated at a repository. (a) Wedge failure. (b) Stress-induced brittle failure. * Spalling (Stress induced)

Risk Management FREQUENCY ANALYSIS SEVERITY OF HAZARD

* Probability distribution * Size of Wedge to Support
Risk Assessment Risk control - In-situ stress - Elevation
- DFN model - Tunnel orientation
- Strength
Risk Risk Decision- * Frequency of occurrence * Depth of Spalling
Monitoring
analysis evaluation making - % of tunnel length - Elevation
- vs Depth & tunnel orientation - Tunnel orientation
Hazard Risk Risk Risk
identification estimation acceptance estimation

Fig. 2. A general framework for risk management proposed by the Canadian RISK CONTROL OPTIONS
Standards Association [4].
* Depth of Repository * Tunnel orientation * Tunnel support
* Tunnel Size * Tunnel shape * Exposure time
With risk defined as the combination of the probability of * Excavation procedures * Organization of the tunnelling work
occurrence of an undesired event and the possible extent of the
event’s consequence, risk can, in principle, be calculated. The full
RISK EVALUATION
potential of QRA is best met with the establishment of acceptable
risk criteria. Establishing acceptable risk criteria in geotechnical * Compare to levels of acceptable risk for underground construction
engineering is a challenge that is evolving. Relating consequences - Capacity
to cost/benefit analysis provides a simpler basis for evaluating - Servicablity
acceptable risk. The link between risk and benefit must be
balanced and within the context of geotechnical engineering the
risks are usually reduced to an acceptable standard by the best Need for
Underground Risk
practical means. Yes No additional
Repository Acceptable
information
Design & analysis
2.1. Terminology
Fig. 3. Illustration of the steps involved in assessing the risk for wedge instability
Swedish regulations state that probability-based methods, and spalling. This report is focused on developing the methodology for assessing
such as the partial coefficient method, should be used in the severity of the hazard and does not address ‘‘risk evaluation’’.
 ARTICLE IN PRESS
C.D. Martin, R. Christiansson / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219–228
3. Rock mass spalling strength
When the stresses on the boundary of an underground
excavation reach the rock mass strength failure occurs. In good
quality hard rock, such as that found in the Scandinavian Shield,
the failure process is described as ‘‘spalling’’ and the associated
rock mass strength as the ‘‘spalling strength’’. The term spalling is
purposely used to indicate that the failure process involves
extensional splitting/cracking first analysed by [6] and is a
different failure process from that of shearing which is commonly
observed in weak rocks.
Spalling is generally defined as the formation of stress-induced
slabs on the boundary of an underground excavation (see Fig. 4a).
It initiates in the region of maximum tangential stresses and
results in a v-shaped notch that is local to the boundary of the
opening (see Fig. 4b). The slabs can vary in thickness from a few
millimetres to a few centimetres, for circular underground
openings that range in diameter from 1 to 5 m. The progressive
nature of the slabbing process results in v-shaped notches that are
normally located diametrically opposite each other for circular
openings. In order to identify if spalling will occur the rock mass
spalling strength must be established. Two extensive in situ
experiments have been carried out to investigate brittle failure:
AECL’s Mine-by Experiment and SKB’s Äspö Pillar Stability
Fig. 4. Examples of spalling observed around a 1.8-m-diameter borehole in the
ÄPSÖ Pillar Stability Experiment [7]. Photos provided by C. Andersson. (a) Spalling
slabs. (b) v-Shaped notch.
221
Experiment (APSE). The results from those tests are briefly
described below.
3.1. AECL’s Mine-by Experiment
The first phase of AECL’s Mine-by Experiment, the excavation
of a 3.5-m-diameter tunnel, was carried out at the 420 Level of
AECL’s Underground Research Laboratory from 1990 to 1995 to
investigate progressive failure around a circular opening in brittle,
unfractured Lac du Bonnet granite. The tunnel was excavated
without explosives, and state-of-the-art instrumentation com-
prising both geomechanical and geophysical instruments was
used to monitor the failure process. The in situ experiment
was supplemented by extensive laboratory tests. The major
conclusion from the Mine-by Experiment was that brittle
failure initiated when the maximum tangential stress on the
boundary of the tunnel reached 120 MPa [9]. The mean uniaxial
compressive strength of Lac du Bonnet granite was given as
212 MPa, therefore the onset of brittle failure can be expressed as
120=212 ¼ 0:56UCSm.
This 120 MPa value was also confirmed by Read et al. [10] by
excavating tunnels with various shapes and various orientations
relative to the in situ stress state at the 420 Level. Their work
concluded there was little variation in the tangential stress
required to initiate brittle failure in the medium-grained Lac du
Bonnet granite suggesting that the local variation in the granite
rock mass strength and in situ stress magnitudes were minor.
However, Read et al. [10] noted that the tangential stress required
to initiate brittle failure in the fine-grained granodiorite was
approximately 150 MPa, despite the granodiorite only having a
modest increase in the laboratory mean uniaxial compressive
strength. Only a few observations were made at the URL, but it
was clear that the grain size of the rock had a significant effect on
the stress magnitude required to initiate brittle failure.
3.2. Äspö Pillar Stability Experiment
The APSE is reported in this volume and relevant results are
reported here. The APSE was excavated at the 420-m depth of
Äspö Hard Rock Laboratory and examined spalling in a fractured
rock mass subjected to thermo-mechanical loading. The excava-
tion of two 1.8-m-diameter boreholes, separated by 1-m-thick
pillar, was purposely over-stressed, using excavation- and ther-
mally induced stresses, to induce spalling in the pillar [11]. The
stresses in the pillar were controlled and the onset of spalling
monitored using acoustic emission instrumentation and deforma-
tions. The major difference between the APSE and the Mine-by
Experiment is that APSE was located in slightly fractured
heterogeneous rock mass but the intact uniaxial strength of Äspö
Diorite is similar to that of Lac Du Bonnet Granite, i.e., 211 and
212 MPa, respectively.
3.3. Summary
Table 1 compares the rock mass spalling strength determined
from APSE and the Mine-by Experiment. In Table 1 the rock mass
spalling strength ðssm Þ is normalised to the mean laboratory
uniaxial compressive strength. The mean uniaxial strength, rather
than the range of uniaxial strength values, was found to provide
better agreement with field observations. This is discussed further
in the section on calibration. This ratio provides a method for
evaluating the rock mass spalling strength for other sites when
only the laboratory uniaxial compressive strength is known.
In the early stage of a project the rock mass spalling strength
cannot be back-calculated from in situ observations provided in
 ARTICLE IN PRESS

222 C.D. Martin, R. Christiansson / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219–228

Table 1
Summary of the onset of dilation (crack initiation, CI) measured in laboratory tests and the rock mass brittle strength from in situ tests

Underground facility Laboratory mean values In situ spalling strength

CI (MPa) CI/UCS UCSm (MPa) ssm ðMPaÞ ssm =UCSm

Äspö Pillar Stability Experiment

Äspö Diorite [8] 94 0.44 211 122  9 0:58  0:04
Mine-by Experiment
Lac du Bonnet granite 97 0.45 212 120 0.56
Granodiorite 120 0.52 228 150 0.65

The mean peak laboratory uniaxial compressive strength (UCSm) is used as the reference strength.

Table 1. In such situations, the onset of extension cracking -0.1

2
(dilation) measured in laboratory uniaxial samples can be used 
+0.1
to indicate the boundary for the initiation of damage in brittle Df
rock [12–16]. Comparison of the crack initiation values measured 1.8 a
in laboratory uniaxial tests to the spalling rock mass strength
suggests that crack initiation provides a lower bound limit for the
1.6
rock mass spalling strength (see Table 1) and in crystalline rocks
frequently occur between 0.4 and 0.6 of the mean uniaxial Df / a
compressive strength [17]. For the purposes of this paper the rock 1.4
mass spalling strength established for the APSE and given in Table
1 will be used in following sections. 1.2 Df 
= 0.48 ± 0.1 + 0.5 
a sm

4. Spalling assessment methodology

1.25 1.5 1.75 2 2.25 2.5 2.75 3
Once the rock mass spalling strength is established the next Maximum boundary stress 
step is to determine if spalling will occur. Because spalling is a Rock mass spalling strength sm
stress-induced failure process, the stresses on the boundary of the
excavation must be determined. Fig. 5. Empirical relationship used to establish the severity of the hazard, i.e., the
depth of spalling. Data from [18]. Also shown are the new data from the Äpsö Pillar
Stability Experiment (APSE) [7].
4.1. Spalling criterion

The potential for spalling is assessed using a traditional factor

of safety approach that compares demand to capacity. If the
magnitude of the maximum tangential elastic stresses ðsyy Þ on the
boundary of an underground opening (Demand) reaches the rock
mass spalling strength (Capacity) the factor of safety can be
expressed as
ssm
factor of safety ¼
syy
The maximum tangential elastic stress on the boundary of circular
opening in a continuous, homogeneous, isotropic and linearly
elastic rock can be expressed using the well known Kirsch
equations for plane strain:
syy ¼ 3smax  smin
where smax and smin are the maximum and minimum far-field
principal stresses, respectively, in the plane of analysis.
In the criterion above, determining the maximum tangential
stress is essential for evaluating the potential for spalling. It
should be noted that the maximum tangential stress is a function
of the tunnel profile shape and the orientation of the tunnel
relative to the in situ stress tensor. Numerical analyses are
required to determine the maximum tangential stress on the
boundary of non-circular tunnels, e.g., horse-shoe-shaped tunnels.
4.2. Severity of spalling
Once the stresses on the boundary of the excavation reach the
rock mass spalling strength (Factor of safety for spalling p1) and
spalling initiates, the severity of the hazard must be assessed, i.e.,
how deep will the spalling extend. The depth of spalling can be
estimated using the empirical correlations described in [18] and
given in Fig. 5. These data were compiled from published case
histories in a wide range of rock mass conditions and in situ
stresses. The results from the APSE have been added to Fig. 5 and
(1) Fig. 5 shows that the empirical correlations predicted the depth of
spalling for the APSE reasonably well.
In Fig. 5 the depth of spalling is normalised to the tunnel radius
and is measured from the centre of the tunnel. The expression for
the depth of spalling given in Fig. 5 for approximately circular
openings can be rewritten as
(2)  
syy
Sd ¼ a 0:5  0:52 for syy 4ssm (3)
ssm
where the Sd is measured from the boundary of the tunnel.
Andersson [7] observed that the depth of spalling in the APSE
was influenced locally by existing fractures. However, as shown by
Fig. 5 it is not anticipated that existence of fractures will
significantly affect the average depth of spalling calculated using
Eq. (3).
4.3. Probability for spalling
As with any geotechnical analysis there is uncertainty in the
input parameters. This uncertainty or variability varies with the
stage of the project. In the spalling assessment the uncertainty is
 ARTICLE IN PRESS

C.D. Martin, R. Christiansson / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219–228 223

Step 1: Stress and strength distribution for the in situ stresses. The triangular distribution is typically used
as a subjective description when there is only limited sample data
mean and the user wishes to provide the most likely value. Other
sm = kUCS mean = x UCSmean
distributions can be used if sufficient data are available.
k The factor of safety was calculated using the @Risk Monte
min max Carlo simulations and hence the output is expressed as a
probability distribution, shown graphically as Step 2 in Fig. 6.
For these simulations there is no correlation between stress and
mean mean strength, i.e., the highest stress can be associated with the lowest
 = 31 - 3 = 3 -
strength. In reality, this may not be the case as the highest in situ
1 3 stresses are often found in the most competent rock mass and the
min max min max lowest stresses in highly fractured rock masses. This level of
Step 2: Spalling Factor of safety information could be considered if available.
The output in Fig. 6 for the spalling factor of safety provides the
1.0 mean factor of safety as well as the probability of the factor of
Mean = 1.11
0.8 safety being less than 1. If there is a 10% probability that the
sm spalling factor of safety is less than 1, this implies that for a 100 m
0.6
FOS = = long tunnel, there is the potential that 10 m of the tunnel length
 0.4 will experience spalling. If there is 0% probability of spalling there
0.2 is no depth of spalling. However, once the probability of spalling is
greater than zero, a depth of spalling is determined. For the depth
0.7
0.7 0.9 1 1.1 1.3 1.5 of spalling, only the mean depth of spalling was used for assessing
the severity of the hazard.
14.2% 85.8%
Probability of Spalling 1.5
Step 3: Depth of spalling 5. Calibration of methodology
sm
Sd = a 0.5  - 0.52
 The 420 Level (depth of 420 m) of AECL’s URL provided an
a = radius of the opening in metres opportunity to calibrate the methodology described above. The
16 Mean = 0.043 m
extent of spalling, indicated by the size of the v-shaped notch,
14 varied for tunnels in different directions [19]. Fig. 7 shows the
12 location of various tunnels on the 420 Level of AECL’s URL, the
10 observed spalling and the orientation of the in situ stress
Sd = 8 magnitudes. The extent of spalling can be readily seen in Fig. 7
6 by comparing the excavation profiles for Rooms 413 and 405. Note
4 that for Room 413, which is oriented parallel to the maximum
2 horizontal stress, only very minor spalling occurred as the half-
0
0 0.045 0.09 0.135 0.18 barrels associated with the blast holes were still visible for nearly
all of the perimeter profile. While Room 405 which was excavated
90% 5%
perpendicular to the maximum principal stress, showed extensive
0.0036 0.0987 spalling, i.e., large v-shaped notch. In addition to the AECL data, an
Depth of Spalling (m) additional calibration analysis was carried out for the 5-m-
diameter tunnel used in SKB’s Prototype Repository located at a
Fig. 6. Illustration of the calculation flow chart used to establish spalling factor of
depth of 450 m at the Äspö Hard Rock Laboratory. No spalling was
safety, the probability of spalling and the average depth of failure.
observed during the excavation of this tunnel. In fact no spalling
had ever been observed at the Äspö Hard Rock Laboratory prior to
associated with the mean uniaxial compressive strength and the the APSE (see Andersson this volume).
in situ stress magnitudes. There is also uncertainty with the stress Note that [20] established the confidence for the in situ stress
orientation but for the purpose of this paper that will be ignored. tensor for the URL 420 Level through large scale back-analyses and
To incorporate this uncertainty a probabilistic approach is used to provided the likely range in stress magnitudes. These ranges were
assess the potential for spalling. Fig. 6 illustrates the calculation used in conjunction with the approach given in Fig. 6 to determine
logic used to establish the spalling factor of safety, the probability the probability of spalling for the AECL’s Mine-by test tunnel,
of spalling and the average depth of spalling. It should be noted in Room 413 Level and AECL’s 1-m-diameter borehole drilled from
Fig. 6 that the mean uniaxial compressive strength is used to the floor of Room 405 (Fig. 8a). The same approach was also used
establish the rock mass spalling strength ðssm Þ, and the factor k is for the Äspö Prototype Repository tunnel using the stress tensor
used to establish the expected variability in the rock mass spalling determined from convergence back-analysis by [21]. The com-
strength. puted mean factors of safety for spalling are in agreement with
An EXCEL spreadsheet was developed for Fig. 6 using Eqs. (1) the observations. For example, in AECL’s Room 413 the methodol-
and (3), i.e., spalling factor of safety and the depth of spalling, ogy indicated a mean factor of safety of 0.99, which is in
respectively. The probability for spalling and the potential range agreement with the very minor spalling observed. However, the
in the depth of spalling was determined using the software @Risk1 tunnels oriented perpendicular to the maximum stress, which
which consists of a series of macros for EXCEL that conducts always exhibited spalling, had a mean factor of safety of 0.72. In
Monte Carlo simulations. In Fig. 6 a triangular distribution is used addition there was no evidence of spalling in the Äspö Prototype
Repository tunnel which is also consistent with the computed
mean factor of safety of 1.43. The range in the factors of safety
1
Available from www.palisade.com. computed in Fig. 8a have been summarised in Fig. 8b.
 ARTICLE IN PRESS

224 C.D. Martin, R. Christiansson / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219–228

11 MPa 59 MPa that computed for the Äspö Prototype Repository tunnel. By
modifying the tunnel geometry to increases the tangential
59 MPa stresses on the tunnel rook and floor the factor of safety was
45 MPa 45
SE NW 48 reduced to 1.3. Though the factor of safety was significantly

m f
oo o
decreased no spalling was observed in the 50-m-long tunnel.

R xis
A
However, by drilling the 1.8-m-diameter borehole in the floor of
204 the tunnel the tangential stresses on the boundary of the borehole
252 27 locally exceeded the spalling strength and the factor of safety was
reduced to approximately 1. By drilling the second 1.8-m-
225° diameter borehole the tangential stresses were increased again
and spalling was observed over 25% of the borehole depth (see
1-m-diameter borehole
Fig. 4b). The computed factor of safety based on elastic stresses
Mine-by Test tunnel in floor of Room 405
would have decreased to approximately 0.8 at the end of drilling
the second hole. While the APSE was not used in the development
Extensive 11 MPa
11 MPa of Fig. 8b the spalling observations and the computed elastic
Minor Spalling Spalling
stresses reported by [7] are consistent with the trend in Fig. 8b.
44 MPa 59 MPa
Half barrel 5.1. Spalling for non-circular openings
Half barrel

The methodology used to assess the spalling hazard given in

1m 1m the previous sections was for a single circular opening in a
continuous, homogeneous, isotropic, linearly elastic rock. For non-
circular openings or when the interaction of multiple openings
Room 413 Room 405 must be evaluated, a numerical analysis is required. Martin [9]
described the development of the criteria that should be used for
numerical analysis to simulate the spalling failure and Hajiabdol-
VENT
majid et al. [22] illustrated the implementation of the methodol-
RAISE SHAFT ogy using the Mine-by Experiment. This criteria uses a
Mohr–Coulomb failure envelope with the peak friction angle
ðfp Þ set to zero, to indicate that the strength of the rock is entirely
N E cohesive and the peak cohesion ðcp Þ of the rock set to 12 the spalling
Room 409
W rock mass strength (cp ¼ 60 MPa). Once spalling occurs and the
S
5
40

slabs form, the strength of the broken rock slabs is mainly

RM

frictional and hence the cr is set to a nominal 5 MPa reflecting the

σ1
Mine-by Test Tunnel
σ2
Room 413
σ1 σ2
Magnitudes (MPa) 60±3 45±4
Trend/Plunge (°) 145/11 054/08
Fig. 7. Location of AECL’s Mine-by test tunnel, Room 413, Room 405 and 1-m-
diameter borehole drilled from the floor of Room 405.
Fig. 8b gives the mean factor of safety versus the range
(minimum–maximum) in the factor of safety for the cases
investigated. Inspection of Fig. 8b reveals that if the mean factor
of safety is greater than approximately 1.15, the probability of
spalling is essentially zero. In these analyses the in situ stress
tensor is known with confidence. During the early stages of a site
investigation program the in situ stress tensor may not be known
with the same degree of confidence as the examples used to
develop Fig. 8b. Hence it is suggested that when the stress tensor
is not well constrained a factor of safety of 1.25 be used for
preliminary design screening purposes. Based on Fig. 8b a factor of
safety of 1.25 should give an extremely low probability of spalling.
If the analysis using the mean stress magnitudes and the mean
uniaxial compressive strength gave factors of safety for spalling of
less than 1.25, probabilistic analyses could be carried out.
A circular access tunnel for the APSE while not shown in Fig. 8b
would have had a factor of safety of approximately 1.5, similar to
small strain nature of the problem, and fr is set to 45 . The peak
σ3
tensile strength ðst Þ goes to zero once spalling initiates. Fig. 9
shows the spalling-failure envelope that could be used to predict
the depth and extent of spalling using this approach. Once the
depth of spalling has been assessed using this approach, the
length and capacity of support system can be determined
assuming that the entire weight of the spalled region must be
σ3 retained by the support. Obviously it would be more time
consuming to conduct probability analyses using this approach.
11±4
However, if required the Point Estimate Method, developed by
290/77 [23], could be used for the calculation of the mean and standard
deviation of the factor of safety. This latter approach is discussed
in detail in [24] and is beyond the scope of this paper.
Fig. 9 shows the spalling-failure envelope that could be used to
predict the depth and extent of spalling using this approach. Once
the depth of spalling has been assessed using this approach, the
length and capacity of support system can be determined
assuming that the entire weight of the spalled region must be
retained by the support.
6. Application of methodology
SKB in Sweden are currently conducting a site investigation for
a deep geological repository for the disposal of nuclear waste. The
design of that repository will be carried out using the reference
KBS3 concept [25]. The site is several square kilometres in area
and boreholes have been drilled to depths of 1000 m. A major
objective in the site investigation and early design phase is to
estimate if there is sufficient space for the construction of
repository that meets the safety and operational requirements.
One of the operational concerns that arise is will spalling occur at
 ARTICLE IN PRESS

C.D. Martin, R. Christiansson / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219–228 225

Probability of Spalling - Aspo Probability of Spalling -

Prototype Repository Room 413
1.000 1.000
Mean=1.43 Mean=0.99
0.800 0.800
0.600 0.600
0.400 0.400
0.200 0.200
0.000 0.000
1.25 1.35 1.45 1.55 1.65 0.75 0.875 1 1.125 1.25
5% 5% 5% 5%

1.319 1.550 0.853 1.090

Spalling FOS Spalling FOS

Probability of Spalling - URL 420 Level

1-m-diameter Borehole Probability of Spalling - Mine-by
1.000 1.000
Mean=0.89 Mean=0.72
0.800 0.800
0.600 0.600
0.400 0.400
0.200 0.200
0.000 0.000
0.8 0.845 0.89 0.935 0.98 0.65 0.695 0.725 0.753 0.8

5% 5% 5% 5%
0.848 0.935 0.695 0.753
Spalling FOS Spalling FOS

1.6

Minimum Mean
Aspo Prototype
Mean Factor of Safety

1.4
No
Spalling
1.2
Maximum

1 Very Minor Spalling

Room 413
Extensive 1-m-diameter borehole
0.8 Spalling

Mine-by

0.6

0.6 0.8 1 1.2 1.4 1.6

Factor of Safety

Fig. 8. Comparison of the probability of spalling and the factor of safety for the Äspö Prototype Repository tunnel, AECL’s Mine-by test tunnel, Room 413 and 1-m-diameter
borehole drilled from the floor of Room 405. (a) Probability of spalling. (b) Spalling factor of safety.

the proposed repository horizon between 400- and 700-m depth.
If yes, how severe will the spalling be. In this section the
methodology that has been outlined previously is applied to the
data obtained from the site investigation program [26] to assess
the potential for spalling for the circular vertical deposition hole
and circular horizontal deposition tunnels (Fig. 10). In these
analyses the deposition hole is analysed separately from the
deposition tunnel. In reality there could be additional stress
concentrations around the deposition hole caused by the deposi-
tion tunnel. However, in the KBS3 concept, the top portion of the
deposition hole is not considered a barrier and hence for the
purpose of this study the two excavations are analysed separately.
6.1. In situ stress
In situ stress data were obtained from triaxial overcore
measurements using the Borre Probe [27] and hydraulic fracturing
[26,28]. The results for the stress magnitudes, expressed as
horizontal and vertical components, are given as stress gradients
in Table 2. These gradients were used to determine the stress
magnitudes at the 350, 450, 550 and 650 m depths in Table 2.
Table 2 presents the mean stress gradient as well as the
uncertainty in the gradient based on all available data from the
site. For example, it is highly unlikely the vertical stress gradient
will have a large variation with depth while there is considerable
 ARTICLE IN PRESS

226 C.D. Martin, R. Christiansson / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219–228

250
Peak Parameters
φp = 0°
cp = 60 MPa
200 σt = 10 MPa
Spalling Residual Parameters

Shear stress (MPa)

φr = 45°
150 Failure envelope cr = 5 MPa
φr = 45°

100

50

σsm σci
σt
50 100 150 200 250
Normal stress (MPa)

Fig. 9. The spalling-failure criterion developed by Martin et al. [17] and Hajiabdolmajid [22] for implementation in numerical continuum software such as Phase2 or Flac.

Table 2
Stress gradients used for spalling analysis

sHmax shmin svert

Gradient (MPa/m) 35 þ 0:02z 19 þ 0:025z 0:0265z
Variability 10% 20% 0:0005z
Depth, z (m) (MPa) (MPa) (MPa)

350 42 28 9.3
450 44 30 11.9
550 46 32 14.6
650 48 35 17.2

6.3.1. Vertical deposition holes

Clay buffer Crystalline bedrock
Fig. 10. The proposed layout for the 1.8-m-diameter deposition boreholes and
deposition tunnels in a KBS3-type repository.
uncertainty in the minimum horizontal stress gradient with
depth, expressed as 0:005z MPa=m and 20%, respectively, in
Table 2.
6.2. Uniaxial compressive strength
The uniaxial compressive strength for the granite was taken
from laboratory test results from [26]. The data are summarised in
Fig. 11. Fig. 11a shows two rock types but for this analysis only the
uniaxial compressive strength for granite is considered in the
analysis.
The mean factor of safety and probability for spalling for the
1.8-m-diameter vertical deposition boreholes at repository depths
of 350, 450, 550 and 650 m are given in Fig. 12. These results are
based on 10,000 simulations using the @Risk software. A
probability of failure of 0% implies that none of the 10,000
simulations had a factor of safety less than 1. At a depth of 650 m
the probability of spalling is less than 1%. Hence these results
suggest that spalling will not be encountered along the deposition
holes, regardless of the repository depth. These analyses are for
the end-of-construction case. Similar analyses could be carried
out to assess the impact of thermal loading on the potential for
spalling.
It should be noted that these analysis assumed a uniform stress
distribution along the deposition hole. In reality the deposition
holes are 8-m long and connected to a deposition tunnel. Hence
the stress magnitudes along the deposition hole will not uniform.
For such geometry a three-dimensional stress analysis using the
tunnel geometry and deposition hole spacing would be carried
out in a later design phase for those cases where the probability of
spalling is significant.

6.3. Spalling assessment

The spalling analysis was carried out using the steps shown in
Fig. 6. The detailed results from the analyses for both the
deposition boreholes and the deposition tunnels are presented
in terms of the mean factor of safety (see Eq. (1)), the probability
in percent that the factor of safety is less than 1, and the
associated mean depth of failure.
6.3.2. Deposition tunnels
The factor of safety, probability for spalling and associated
depth of failure in metres for 5-m-diameter circular deposition
tunnels at repository levels of 350, 450, 550 and 650 m are given
in Fig. 13. The tunnels are evaluated parallel (01), perpendicular
(901) and at 451 to the orientation of the maximum horizontal
stress. Note that only the tunnels aligned perpendicular to the
 ARTICLE IN PRESS

C.D. Martin, R. Christiansson / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219–228 227

The tunnels analysed were assumed to be circular to assess the

350 impact of tunnel orientation relative to the stress field. Once a
Uniaxial compressive strength (MPa)

suitable orientation is selected it can be further optimised for the

300 actual tunnel shape, which for the KBS3 concept is currently
horse-shoe-shaped.
250

200
7. Conclusions
150
The construction of a nuclear waste repository in crystalline
Tonalite

100 rock will likely encounter spalling on the boundary of the

excavations in the regions of the maximum tangential stress. A
50 probabilistic methodology was developed to assess if spalling will
occur and determine the severity of the hazard defined as the
0
0 100 200 300 400 500 600 700 800
1.8
Depth (m) Deposition Tunnels
1.7

Mean Factor of Safety

3 1.6
X <= 189.0
Normal (225, 22) X <= 261.7 0deg
5.0% 95.0% 1.5
2.5
1.4
45deg
Values x 10^-2

2 1.3

1.5 1.2
90deg
1.1
1
1
200 300 400 500 600 700
0.5 Depth (m)

0
160 180 200 220 240 260 280 300 45
90 deg
40
% Probability of Spalling

Uniaxial Compressive Strength (MPa)

35
Fig. 11. Uniaxial compressive strength (UCS) versus depth from ground surface
(top figure) and normal distribution fit to all the unconfined compressive strength
30
data for Forsmark rock unit. Tonalite has been excluded in the normal distribution. 25
(a) UCS versus depth. (b) UCS histogram. 20
15
1.4 100
10
90 5
0 deg & 45 deg
80 0
Probability of Saplling (%)
Mean Factor of Safety

1.3 200 300 400 500 600 700

FOS 70
Depth (m)
60
1.2 50 0.007
Mean Depth of Spalling (m)

40 90 deg
0.006
30
1.1 0.005
20
0.004
10
%PofS 0.003
1 0
300 400 500 600 700 0.002
Depth (m) 0.001
0 deg & 45 deg
Fig. 12. Deposition holes—mean factor of safety for spalling and the probability 0
for spalling for repository levels at 350, 450, 550 and 650 m. Note that below a 200 300 400 500 600 700
depth of 500 m there is less confidence in the in situ stress magnitudes used in the Depth (m)
analysis.
Fig. 13. 5-m-diameter circular deposition tunnels—factor of safety, probability for
spalling and associated depth of failure for repository levels at 350, 450, 550 and
650 m. The notation of 01, 451 and 901, refers to the orientation of the deposition
maximum horizontal stress has a significant risk for spalling. This
tunnel relative to the orientation of the maximum horizontal stress, i.e., 01 implies
risk increases with depth and the probability for spalling is the tunnel is parallel to the maximum horizontal stress and 901 implies that the
approximately 40% at the 650-m depth. tunnel is perpendicular to the maximum horizontal stress.
 ARTICLE IN PRESS
228 C.D. Martin, R. Christiansson / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 219–228
depth of failure. The onset of spalling is assessed using the rock
mass spalling strength and the Kirsch equations for elastic
stresses for a circular opening. The rock mass spalling strength
was assessed using the results from in situ experiment and
expressed as a fraction of the uniaxial compressive strength. For
the crystalline rocks this factor was found to be 0:57  0:02. This
value was found to be only slightly above the onset of crack
initiation (dilation) measured in laboratory samples. Hence in the
absence of in situ results, the laboratory crack initiation stress
could be taken as a lower bound for the spalling strength.
A calibration of the methodology showed that it was reason-
ably robust and that a mean factor of safety for spalling of 1.25
would reduce the probability for spalling to very low values. The
probability-based methodology was applied to a site currently
being investigated in Sweden for a nuclear waste repository at a
depth between 400 and 700 m in crystalline rock. The repository
is being designed using the KBS3 concept which requires the
construction of 1.8-m-diameter 8-m deep deposition holes from
approximately 5-m-diameter deposition tunnels. The methodol-
ogy showed that the probability for spalling for the deposition
holes was very low between 400 and 550 m. Below 550 m the
probability of spalling increased. The risk for spalling for the
deposition tunnels could be eliminated by orienting the tunnels
parallel to the maximum principal stress.
It is envisaged that the methodology proposed would be
suitable for the preliminary design stages of a repository when
multiple layouts are being evaluated. It should be noted that the
confidence in the results from such a probabilistic approach is a
function of the confidence in the input data. Such analyses cannot
substitute for a well planned site characterisation program to
determine those input parameters with confidence.
Acknowledgements
The authors would like to thank Dr. John Hudson and Prof.
Håkan Stille and two anonymous reviewers for their constructive
review comments.
References
[1] Martin CD, Kaiser PK, Christiansson R. Stress instability and the design of
underground excavations. Int J Rock Mech Min Sci 2003;40(7–8):1027–47.
[2] Barton N, Grimstad E. The Q-system following twenty years of application in
NWT support selection. Felsbau 1994;12(6):428–36.
[3] Morgenstern NR. The 3rd Casagrande lecture: managing risk in geotechnical
engineering. In: Proceedings of the 10th Pan-American conference on soil
mechanics and foundation engineering, Guadalajara, Mexico, vol. 4. Mexico
City: Mexican Society of Soil Mechanics; 1995. p. 102–26.
[4] Canadian Standard Association. Risk analysis requirements and guidelines.
Technical Report CAN/CSA-Q634-91. Quality Management—A National
Standard of Canada, Rexdale, Toronto, Canada; 1991. 42p.
[5] Stille H, Andersson J, Olsson L. Information based design in rock engineering.
SveBeFo Rapport 61, SveBeFo, Swedish Rock Engineering Research, Stockholm,
Sweden; 2003. 147p.
[6] Fairhurst C, Cook NGW. The phenomenon of rock splitting parallel to the
direction of maximum compression in the neighbourhood of a surface. In:
Proceedings of the 1st congress of the international society of rock mechanics,
Lisbon, 1966. p. 687–92.
[7] Andersson C. Äspö pillar stability experiment. In: Franzén T, editor.
Bergmeckanikdag. Stockholm: SveBeFo; 2005. p. 69–78.
[8] Andersson CJ. Äspö Hard Rock Laboratory. Äspö pillar stability experiment.
Final Report, Rock mass response to coupled mechanical thermal loading. SKB
TR 07-01, Institution Swedish Nuclear Fuel and Waste Management Co.,
Stockholm; 2007.
[9] Martin CD. Seventeenth Canadian geotechnical colloquium: the effect of
cohesion loss and stress path on brittle rock strength. Can Geotech J
1997;34(5):698–725.
[10] Read RS, Martino JB, Dzik EJ, Chandler NA. Excavation stability study—-
analysis and interpretation of results. OPG 06819-REP-01200-0028-R00,
Ontarion Hydro, Nuclear Waste Management Division, Toronto; 1997.
[11] Andersson C, Martin CD, Christiansson R. SKB’s Äspö Pillar Stability
Experiment, Sweden. In: Yale DP, Willson SM, Abou-Sayed AS, editors. CD-
ROM Proceedings of the 6th North American rock mechanics symposium-
NARMS04, Houston, 2004. p. 1–8, NARMS 04-503.
[12] Brace WF, Paulding B, Scholz C. Dilatancy in the fracture of crystalline rocks. J
Geophys Res 1966;71:3939–53.
[13] Scholz CH. Microfracturing and the inelastic deformation of rock in
compression. J Geophys Res 1968;73(4):1417–32.
[14] Hallbauer DK, Wagner H, Cook NGW. Some observations concerning the
microscopic and mechanical behaviour of quartzite specimens in stiff, triaxial
compression tests. Int J Rock Mech Min Sci Geomech Abstr 1973;10:713–26.
[15] Lajtai EZ. Microscopic fracture processes in a granite. Rock Mech Rock Eng
1998;31(4):237–50.
[16] Martin CD, Chandler NA. The progressive fracture of Lac du Bonnet granite. Int
J Rock Mech Min Sci Geomech Abstr 1994;31(6):643–59.
[17] Martin CD, Kaiser PK, McCreath DR. Hoek–Brown parameters for predicting
the depth of brittle failure around tunnels. Can Geotech J 1999;36(1):136–51.
[18] Martin CD, Christainsson R, Söderhäll J. Rock stability considerations for siting
and constructing a KBS3 repository: based on experiences from Äspö HRL,
AECL’s URL, tunnelling and mining. Technical Report TR-01-38, Swedish
Nuclear Fuel and Waste Management Co., Stockholm, Stockholm; 2001.
[19] Read R. 20 years of excavation response studies at AECL’s Underground
Research Laboratory. Int J Rock Mech Min Sci 2004;41(8):1251–75.
[20] Martin CD, Read RS. AECL’s Mine-by Experiment: a test tunnel in brittle rock.
In: Aubertin M, Hassani F, Mitri H, editors. Proceedings of the 2nd North
American rock mechanics symposium, Montreal, vol. 1. Rotterdam: A.A.
Balkema; 1996. p. 13–24.
[21] Andersson C, Martin CD. Stress variability and the design of the Äspö Pillar
stability experiment. In: Sugawara K, Obara Y, Sato A, editors. Proceedings of
the 3rd international symposium on rock stress, Kumamoto. Rotterdam: A.A.
Balkema; 2003. p. 321–6.
[22] Hajiabdolmajid V, Kaiser PK, Martin CD. Modelling brittle rock failure. Int J
Rock Mech Min Sci 2002;39(6):731–42.
[23] Rosenblueth E. Point estimates for probability moments. Proc Natl Acad Sci
USA 1975;27(10):3814–21.
[24] Harr ME. Reliability-based design in civil engineering. New York: McGraw-
Hill Inc.; 1987.
[25] Olsson O. Underground storage for nuclear waste in Sweden. In: für
Geotechnik eV (DGGT) DG, editor. Proceedings of the EUROCK 2000
symposium, Aachen. Essen: Verlag Glückauf GmbH; 2000. p. 117–24.
[26] SKB. Preliminary site description, Forsmark area-version 1.2. SKB Report R-
05-18, Swedish Nuclear Fuel and Waste Management Co., Stockholm; 2005.
[27] Sjöberg J, Klasson H. Stress measurements in deep boreholes using the Borre
(SSPB) probe. Int J Rock Mech Min Sci 2003;40(8):1205–23.
[28] Sjöberg J. Overcoring rock stress measurements in borehole KFM01B.
Forsmark site investigation. Technical Report, SKB P-04-83, Swedish Nuclear
Fuel and Waste Management Co.; 2004.