Diederichs Kaiser 1999b

Tensile strength and abutment relaxation as failure control
mechanisms in underground excavations

M.S. Diederichs *, P.K. Kaiser
Geomechanics Research Centre, F217 Laurentian University, Sudbury, Ont., Canada P3E 2C6
Accepted 8 November 1998
Abstract
Classical assessment of instability potential in underground excavations are normally based on yield and rupture criteria for
stress driven failure and on limit equilibrium analysis of structurally controlled failure. While it is true that ultimate failure and
falls of ground can be an eventual consequence of stress fracturing and unfavourable structure within the rock mass, the timing
of such failure is often controlled by the presence of residual tensile capacity, in the form of rock bridges separating joint
segments and fractures and by the mechanisms of clamping and relaxation. Using crack and rock-bridge analogues in
conjunction with an updated voussoir beam model, this paper explores the inuence of residual tensile strength and boundary
parallel relaxation on the failure process. The impact on support design is also examined. In underground hard rock mines with
complex geometries and interacting openings, relaxation is identied as a key controlling factor in groundfall occurrence.
Empirical stability assessment techniques for underground tunnels and for mining stopes are updated to account for relaxation.
# 1999 Elsevier Science Ltd. All rights reserved.
1. Introduction
Stability assessment of underground excavations is
classically divided into two procedural domains. These
two domains are based on a distinction between struc-
turally controlled and stress driven modes of instabil-
ity. It is the premise of this paper, however, that in
non-squeezing and non-bursting ground, structure and
stress serve merely as ground conditioning mechan-
isms. Gravity loading is ultimately responsible for
large scale groundfalls or for signicant loading of
support. When the rock mass jointing is non-persist-
ent, the rock bridges contribute to the stability of an
excavation through the rock mass residual tensile
strength or load bearing capacity. The failure to con-
sider the rock mass self-supporting capacity and the
eects of abutment connement or abutment relax-
ation can lead to erroneous predictions of failure or of
support load.
Gravity induced groundfalls are common occur-
rences in underground excavations of all depths.
Numerous techniques can be applied to assess the po-
tential for such groundfalls provided that an appropri-
ate failure mode is assumed. Typical failure modes
which can be analyzed include wedge fallout, slab or
plug failure, gravity driven caving and beam failure.
Models taking these failure modes into account are
generally usually utilized for associated with stability
assessments in low stress or near surface excavations.
The underlying assumption of most of these models is
that through-going joints or discontinuities are fully
persistent and that stability is controlled by structural
geometry and by friction (with or without dilation).
The inherent tensile or cohesive strength of moderately
jointed rock masses is often assumed to be negligible
in these models.
Alternatively, for excavations at depth or for shal-
lower excavations in weaker rock masses, linear elastic
stress analysis can be used to determine the location
and extent of problematic stress concentrations around
the openings. Failure criteria such as MohrCoulomb
and HoekBrown [1, 2] based on combined cohesive
and frictional strength components can be applied suc-
International Journal of Rock Mechanics and Mining Sciences 36 (1999) 6996
0148-9062/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S0148- 9062( 98) 00179- X
PERGAMON
* Corresponding author. 105 William St. W., Waterloo, Ont.,
Canada N2L IJ8. Tel.: +1-519-578-5327; e-mail: mdiederi@nickel.
laurentian.ca.
cessfully if the rock mass yields with signicant plastic
deformation. These criteria, however, meet with only
limited success [3, 4] for failure prediction around exca-
vations in hard rock environments due to their insensi-
tivity to low connement behaviour. In brittle ground,
it has been shown [59] that under low connement
(such as that which exists adjacent to an opening), tan-
gential compressive stress above a dened threshold in-
itiates and propagates boundary parallel cracks or
fractures.
Such rock mass damage can be responsible for
observed seismicity [9] and stress redistribution [10]
and was found to correlate well with a constant critical
deviator criteria for damage initiation, both in intact
and in moderately jointed rock [9, 11]. This initial
cracking or fracturing is generally parallel to the exca-
vation surface and therefore parallel to the major com-
pressive stress. This crack damage is strain dependant
and may be exacerbated by preferential deection and
dilation into an opening and by existing planes of
weakness such as foliation or meta-bedding resulting
eventually in moderately persistent planar laminations
in otherwise massive or moderately jointed rock
(Fig. 1).
Nevertheless, a prediction of rock mass stress in
excess of yield limits, based on elastic models, does not
inevitably correspond to actual catastrophic failure of
the underground opening. This is fortunate given the
depths encountered in modern mining, where much of
the ground above mining openings has reached a
damage or yield stress threshold at some point in the
mining sequence and yet, with the exception of highly
stressed ground experiencing rockbursting, shearing or
stress buckling, generally remains in place in the short
term. This is due to the rock mass' residual tensile
load bearing capacity normal to the excavation bound-
ary and due to arching to the abutments.
Paradoxically, ultimate failure of this damaged ground
is often induced by changes in mine geometry which
reduce [1216] rather than increase the stresses across
the back or walls as might have been the case in Fig. 1.
It is the premise of this paper that the key to this
discrepancy between modeled or predicted failure and
the actual observed occurrence of failure in under-
ground excavations in hard rock is due to the domi-
nance of the rock mass's tensile load bearing capacity.
However, if combined with abutment relaxation, the
inuence of gravity can exceed the tensile load bearing
capacity of naturally jointed or stress fractured rock
masses leading to the type of failure illustrated in
Fig. 1. This paper will not attempt to explain the gen-
esis of natural fractures and associated rock bridges,
Fig. 1. Gravity collapse of moderately jointed to massive ground after stress induced fracturing.
M.S. Diederichs, P.K. Kaiser / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 6996 70
nor will it examine the mechanics of formation for
boundary parallel stress cracks. The foregoing discus-
sion assumes that such damage is omni-present in
underground excavations and explains the possible
impact on ultimate stability.
In this paper we will discuss the nature of residual
tensile capacity in jointed or fractured rock masses
comparing its capacity to that of economical engin-
eered support systems. The eect of abutment relax-
ation on rock wedge and blocky slab stability is
demonstrated. Dierent mining scenarios are described
which may lead to destabilizing relaxation. The vous-
soir arch analogue is used to further demonstrate the
importance of boundary normal tensile capacity and
abutment relaxation. The voussoir analogue is then
calibrated to correspond with empirical stability guide-
lines for rock mass stability. The eect of relaxation is
further demonstrated in this context.
2. Boundary-normal tensile strength of a damaged rock
mass
While tensile strength can often be justiably neg-
lected in cases of weak and highly fractured rock, the
residual strength of rock bridges in moderately jointed
hard rock masses must be considered. Extension joints
are rarely fully persistent at depth [1719]. Hence, it
would be realistic to expect the initial presence of
intact rock bridges. Foliation planes and bedding
planes are weaker then the surrounding rock but still
possess some limited tensile strength in many cases
until full parting occurs due to excessive deformations.
This can be seen in the case of progressive hangingwall
delamination reported by [20], for example, and par-
tially explains why ground can remain unsupported in
spite of contrary limit equilibrium calculations based
on the stability of a fully bounded wedge, block or
slab. In spite of this initial and static stability, how-
ever, support may still be required to protect against
groundfalls induced by dynamic disturbance, time and
stress dependent tensile strength corrosion [21], and
mining induced deformations. Support in such con-
ditions can also prevent or inhibit the propagation of
fractures and the degradation of rock mass tensile ca-
pacity.
In jointed or in massive rock masses at depth, under
the low normal connement conditions encountered
adjacent to an excavation, surface parallel cracks can
develop as a result of compressive stress. This micro-
cracking parallel to the maximum principal stress has
limited eect on the ultimate compressive strength
since it is an inherently stable fracture process [22].
The same cracks, however, subjected to tensile loading
normal to the plane of the crack (as occurs above a
horizontal excavation roof), propagate easily when a
critical tensile strain is exceeded and have a profound
impact on the strength, particularly if the compressive
stress parallel to the boundary is maintained. Under
such compressive stress, boundary interaction also
plays a role in the enhanced propagation of near-exca-
vation cracks [23]. It can be assumed however, that in
many cases the initial jointing or parallel fracturing is
not complete and that intermittent rock bridges remain
to be exploited by gravity.
The ratio of tensile (s
T
) to uniaxial strength (UCS)
of intact rock is signicantly lower than for other ma-
terials such as steel. Typical tensile values range from
one tenth to one twentieth of the UCS for hard
rocks [24]. These relatively low values, combined with
the conservative assumption of fully persistent jointing
lead to the common neglect of the rock mass tensile
strength as a practically signicant factor in excavation
stability. Grith theory [25] has become a classic
means of explaining the relatively low tensile strength
of heterogeneous and imperfect solids such as rock,
describing how small imperfections such as cracks
serve to concentrate tensile stresses locally resulting in
a reduction in macroscopic sample strength. The as-
sociated stress intensity relationships for internal and
external cracks in plates [2628] can be extended for
isolated cracks and rock bridges in three dimensions
(Fig. 2).
For a given mode I stress intensity factor at crack
extension K
IC
, the tensile strength for a partially
cracked solid can be computed for circular non-inter-
acting cracks (Fig. 2a and d) of radius c, with crack
normals oriented at angle g to the direction of tensile
loading [28]:
s
T
= K
IC
p
2

pc
1
cos
2
g
(1)
Using Kemeny and Cook's solution [28] for the so-
called external crack, the tensile strength of a cylinder
of rock with a total cross sectional area, A, containing
and a circular rock bridge of radius a (surrounded by
a planar, annular crack) is given by
s
T
= K
IC
2a

pa
1
cos
2
g
(2)
If N/V is the number of regularly distributed cracks or
rock bridges in a unit cubic volume (V=1 m
3
), then
the total coplanar cross sectional area (cracked and
uncracked) associated with the crack or rock bridge, A
is
A =
1
N
2a3
(3)
If (A
c
*
)A is the area of the crack and (A
a
*
)A is the area
of the rock bridge (where A
c
*
and A
a
*
are the ratios of
cracked and intact area, respectively, to the total cross
sectional area):
cI
A
*
c
pN
2a3
s
(4)
aI
A
*
a
pN
2a3
s
(5)
where A
a
*
=1 A
c
*
.
It is possible now to express the tensile strength with
respect to the percentage of cracked cross sectional
area for cracks perpendicular to loading (cos
2
g =1):
s
T
=
p
2
K
IC
N
1a3
p(A
*
c
)
q
v
u
u
t for A
*
c
<1 (6)
or inversely, with respect to percentage of intact rock
bridges:
s
T
= 2K
IC
(A
*
a
)
3a2
N
1a3
s
for A
*
a
<1 (7)
These equations for isolated cracks and isolated rock
bridges, analyzed as `external cracks' by Kemeny and
Cook [28], serve as limiting cases for progressive crack
damage as illustrated in Fig. 3 using a unit stress
intensity factor. The actual tensile strength varies line-
arly with K
IC
as per Eqs. (6) and (7).
Fig. 3 clearly illustrates the signicant reduction in
theoretical tensile strength with small amounts of dis-
tributed micro-cracking (A
c
=0 to 0.1). It is acceptable
to presume that even intact rock samples normally
contain a signicant amount of internal damage and
therefore exhibit tensile strengths which are lower than
those for a perfect solid. Fig. 3 also illustrates that the
theoretical tensile strength is apparently insensitive to
the areal extent of cracking for the intermediate range,
A
c
=0.1 to 0.9. For values of A
c
>0.9 or A
a
<0.1,
however, the tensile strength decays rapidly until com-
plete (fully persistent) cracking is achieved. Fig. 3 also
shows that a few large cracks (low N) within the rock
degrade the strength more than numerous smaller
cracks (high N) with the same total crack area. For
comparison with measurements of relative linear joint
trace persistence, P, the regular array of cracks in
Fig. 2 yields the approximate relationship for average
linear persistence, P:
P = c
c
c a
= c
A
*
c
1 2
A
*
c
(1 A
*
c
)
q (8)
where c is a factor which ranges from 1/
for small
(isolated) cracks and from 1 to
for rock bridges

(extensive cracking), depending on the linear persist-
ence measurement direction in the plane of the circular
crack or rock bridge. This factor appears due to the
Fig. 2. Crack geometries: (a) 2D non-interacting crack; (b) 2D rock bridge; (c) 3D assembly of cracks; (d) 3D isolated non-interacting crack; (e)
3D isolated rock bridge.
superposition of a circular crack or bridge within a
square associated area.
In order to fully examine the role of tensile strength
as a support mechanism it is also necessary to examine
the impact of crack damage on rock stiness. This is
achieved using the Grith locus for tension [29]. By
assigning spherical averages to the orientation terms in
an expression for the strain energy due to an individ-
ual crack and incorporating the result into a variation
of Betti's reciprocal theorem, Kemeny and Cook [28]
calculate the ratio of Young's modulus for a solid with
randomly distributed non-interacting cracks:
E
c
E
0
=
1
1 (16a45)w(1 n)
2
(10 3u)a(2 u)
(9)
and for rock bridges in a randomly cracked solid:
E
c
E
0
=
1
1 (p
2
a15)w(1 n)(5 4n)a(m(m1))
(10)
Eq. (9) can be simplied for the case of mono-direc-
tional cracks perpendicular to the direction of tensile
loading:
E
c
E
0
=
1
1 (16a3)w(1 u
2
)
(11)
while Eq. (10) can be simplied for regularly spaced
parallel rock bridges within an axisymmetric (cylindri-
cal) volume:
E
c
E
0
=
1
1 (p
2
w
/
(1 n)
2
)a(m(m1))
(12)
where w= N(c)
3
/V is the crack density for isolated
cracks, w
/
= N(c
/
)
3
/V is the crack density associated
with isolated rock bridges, N is the number of cracks
or rock bridges, V=1 is the unit volume in the study
and a is the average radius of rock bridges surrounded
by a cracked annulus of width c
/
such that:
c
/
=
1
A
*
a
q
pN
2a3
(13)
m = aac
/
=
A
*
a
q
1
A
*
a
q (14)
The value of c
/
represents the average edge separation
for an array of equivalent cylindrical (and not cubic)
rock bridge volumes. The denition is dierent than
that obtained for c = f(a) from Eqs. (4) and (5) and
diers from the three dimensional formulation of
Kemeny and Cook [28]. These modied terms are
necessary to obtain a three-dimensional solution for
rock bridges which is behaviourally compatible with
the original two-dimensional formulation [28].
Eqs. (11) and (12) combined with Eqs. (6) and (7),
dene the Grith locus and the stress/strain relation-
ships for isolated cracks (Fig. 4) and isolated rock
bridges (Fig. 5). In Fig. 4, the cracks nucleate at points
arranged in a regular cubic array with the initial separ-
ation dened by Eq. (3). These cracks then propagate
in a plane perpendicular to loading. In Fig. 5 the
cracks have coalesced leaving a regular array of intact
and progressively shrinking rock bridges with the same
separation as the original cracks. The total number of
cracks, N, remains constant. The genesis of natural
and induced joints or fractures is undoubtedly sensitive
to initial boundary condition and response and may
not necessarily progress to full rupture in a single
stage, terminating at some point in this sequence. At
issue here is the renewed response of non-persistent
Fig. 3. Tensile strength versus normalized cracked cross-sectional area.
cracks and rock bridges upon exposure to excavation
induced gravity loading.
In Fig. 4, it is interesting to note that while a drastic
loss in crack propagation stress does occur in the pre-
sence of very small cracks, practically detectable re-
duction of the reloading modulus does not occur until
the cracks occupy more than 10% of the cracked area
and does not become signicant until the cracked area
exceeds 20%. This latter stage is the point at which
the crack diameter exceeds the minimum separation
between the edges of neighbouring cracks (Inset in
Fig. 4). In other words, this is the point at which the
ratio of inter-crack edge separation to crack length
falls below 1:1. Similarly, a cracked area of 10% corre-
sponds to an edge separation of approximately 2 crack
lengths. This is consistent with the theoretical work of
Pollard and Segall [30] who show that the strain per-
turbation due to a crack decays considerably at an
edge separation of one crack length and becomes insig-
nicant for an edge separation of 2 crack lengths.
At a level of crack density beyond the limit of 10
to 20% cracked area, it is also dicult to reconcile
the Grith locus for isolated cracks with the locus
for isolated rock bridges as end members of a transi-
tional cracking process. It is even more dicult then,
to attach any physical meaning to the residual tail of
the isolated crack locus (Fig. 4) which can be calcu-
lated for high crack densities corresponding as shown
to cracked areas in excess of 100%, a nonsensical
result.
Fig. 5 illustrates the theoretical stressstrain re-
lationships and crack propagation strengths for a
highly cracked solid. In these two plots, the cracks
have coalesced and are converging on remnant (circu-
lar) rock bridges, a more realistic end process for in
situ fracturing. There is a stage between approximately
15% intact area and 1% intact area (corresponding to
85 and 99% cracked area, respectively) through which
the rupture occurs at a near-constant level of overall
extension strain. For a single crack (N=1), a unit
fracture toughness and with E=10 GPa and u =0.2
for the intact rock, this corresponds to a range of re-
sidual tensile strength between 0.3 and 0.03 MPa at a
strain limit of approximately 0.004%. For N=1000
(indicating a highly distributed cracking process), this
threshold strain is approximately 0.012%, spanning re-
sidual strengths of 1 to 0.1 MPa for the same range of
cracked areas. Okubi and Fukui [31] performed direct
tensile tests on numerous rock types using an extre-
mely high resolution servo-control and reported re-
sidual tensile strengths within these ranges of
magnitude. Of particular interest for support design,
given an assumption for crack or rock bridge distri-
bution (N), is the constant nominal strain limit which
can be expected of residual rock bridges. The threshold
is small indicating that if failure is to occur due to
gravity loading, only relatively sti support systems
such as resin grouted rebar on a tight pattern can pre-
vent the loss of rock bridges and residual strength.
The minimum residual strengths 30 to 100 kPa, pre-
Fig. 4. Example of Grith locus and vertical stiness relationships for a rock sample containing isolated horizontal cracks.
dicted in Fig. 5 are also of interest, however, since
they are similar in magnitude to the holding capacities
of conventional support systems used in mining.
Pattern support in the form of regularly spaced
mechanical rockbolts, grouted rebar, split sets or
cablebolts serve primarily to resist gravity and to
`hold' the rock mass adjacent to an excavation. These
support congurations also contribute to the inherent
integrity of the rock mass, resisting dilation and
increasing frictional resistance of rough fractures and
discontinuities. In stress-damaged rock, sti or actively
pre-loaded reinforcement can also serve the role of
preventing the failure of rock bridges and thus the
complete loss of rock mass tensile strength. In their
primary role, however, these support systems can pro-
vide distributed tensile load capacities ranging from 20
to 300 kPa, respectively, for sparse and for economi-
cally limiting support patterns as shown in Table 1.
The demand equivalence in terms of vertical height of
supported rock (assuming no other stabilizing mechan-
ism) is given for comparison. Table 1 also illustrates
the very small relative cross sectional areas of intact
rock (rock bridges) required to achieve the same tensile
load capacity using the example values from Figs. 3
and 5.
The main dierence between the supporting mechan-
isms of articial tendons and internal rock bridges is
the magnitude of the rupture strains in the direction of
tensile loading. While steel tendons such as rockbolts
and cables typically yield at strains of approximately 1
to 2% for average grades of steel [33], reasonably
intact rock specimens (Fig. 4) begin to rupture at ten-
sile strains of less than 0.1% and small rock bridges
(A
c
*
<10%) such as those modeled in Fig. 5 rupture
at strain levels below 0.01%. The convex shape of the
locus in Fig. 5 also indicates that any tensile rupture
Fig. 5. Example of Grith locus and vertical stiness relationships for rock sample containing isolated horizontal rock bridges.
Table 1
Support patterns [32] and equivalent rock bridge area (N=125)
Support type Support
pattern
(mm)
Equivalent
pressure
(kPa) [30]
Maximum
supported
thickness (m)
Capacity
equivalent rock
bridge area
(% cross section)
Rockbolts 22 20 0.7 0.1
Rebar 1.31.3 60 2.0 0.4
Single strand cablebolts 22 65 2.2 0.4
Double strand cablebolts 22 130 4.3 1.2
Double strand cableholts 1.31.3 300 10 4.0
1% rockbridge area is equivalent to a 1010 cm rockbridge per 1 m
2
total area.
involving rock bridges will be sudden, unstable or cat-
astrophic and can be considered brittle constitutive
behaviour for the purpose of comparison to most arti-
cial support systems. Only very sti reinforcement
can be considered eective in the preservation of the
rock mass tensile strength.
If tensile rupture of rock bridges under gravity load-
ing is considered a small-strain brittle phenomenon,
then it is appropriate to consider residual tensile
strength as a supporting mechanism for gravity loaded
blocks as illustrated, for example, in Fig. 6. Consider,
for example, a design factor of safety of 2 against
gravity loading for the 45 degree wedge. This dictates
a maximum span of 4 m if full persistence (intact
area =0%) is assumed and a 22 m pattern of single
strand cablebolts is installed. If it could be determined
that rock bridges exist across the bounding planes of
the wedge and that the relative intact area across these
planes is merely 1%, then a maximum span of 11 m
would be acceptable without the need for support.
This result demonstrates that the stability of an exca-
vation or the factor of safety can be profoundly inu-
enced by the presence of small rock bridges within the
jointed rock mass and thus draws into question the
conventional assumption of full joint persistence in
limit equilibrium analyses. It also points to the import-
ance of blast control for the preservation of such rock
bridges in moderately jointed rock masses.
3. Tensile strength corrosion and time dependency
A caution is warranted with respect to long-term
rock bridge strength. Time dependent failure of exca-
vations is an issue of particular concern in mining and
in particular, with respect to support costs for tempor-
ary openings. Spans which can remain safely unsup-
ported upon excavation can fail suddenly and
catastrophically after a long period of stability with lit-
tle displacement warning. Subcritical crack growth [34]
and humidity and chemically assisted stress
corrosion [21, 35, 36] are recognized as controlling fac-
tors for the time-dependent rupture of cracked solids.
According to Atkinson and Meredith [34] water
induced stress corrosion and subcritical crack growth
can occur in soda lime glass at stress intensities (K
0
) of
approximately 0.3 times the critical intensity factor
(K
IC
). This ratio could be as low as 0.2 for minerals
such as quartz with a corresponding reduction in ten-
sile crack propagation stress or strain for a given crack
geometry. Such crack growth under constant strain
can lead, over time, to a critical cracked area (Fig. 5)
resulting in unstable propagation and ultimate rupture.
In particular, the humidity uctuations or chemically
laden air in underground mines exacerbate crack
instability. Rock bridges, in loose and permeable
ground, which are acting as a signicant supporting
component and are strained are particularly suscep-
tible. Stress corrosion of rock bridges leading to un-
stable crack propagation may also be a source of
delayed microseismicity around underground exca-
vations.
4. Abutment relaxation as a destabilizing mechanism
In order for a fractured or jointed rock mass to sup-
port itself against gravity driven failure, it must have
the ability to transfer load to the abutments through a
frictional or arching mechanism. Both mechanisms rely
Fig. 6. Example of residual tensile strength of rock bridges (N=125/m
3
) acting as eective wedge support (gravity loaded 458 prismatic wedge).
on the existence of stable abutments and on boundary
parallel connement which can result from the in situ
stresses or from deection-induced compression.
Abutment deformation or displacement leading to tan-
gential stress relaxation in the adjoining roof or hang-
ingwall can be an important catalyst for failure in
jointed or fractured ground [37]. In the context of this
discussion `relative relaxation' is synonymous with
connement reduction while `absolute relaxation' or
simply `relaxation' is used to describe a state of com-
plete connement loss, parallel to the excavation
boundary. Relaxation displacement results in the open-
ing of joints and fractures and is equivalent in elastic
models to the generation of boundary parallel tensile
stresses. Of particular interest are situations in which
an entire wall is engulfed in such a `tensile' or relaxed
zone. This scenario is common in mining where high
in situ stress ratios and complex and extreme exca-
vation geometries are present.
It is important to note rst, however, that while the
analyses here are based on simple analogues for relax-
ation in situ, the processes leading to abutment and
boundary parallel relaxation are complex and depend
on rock properties and geological setting, mining
method and sequence, backll procedures, etc. The
objective of this discussion is to convince the reader of
the importance and ubiquity of relaxation in complex
mining environments and to qualitatively and semi-
empirically accommodate the concept into design.
For purposes of illustration, reconsider the simple
example of a gravity driven (non-sliding) wedge above
a horizontal excavation. It can be shown [38] for mod-
erate spans that a two dimensional symmetrical wedge
with an semi-apex angle signicantly less than the in-
herent friction angle (including dilation) is unlikely to
be liberated if the lateral compressive stresses across
the back are signicant greater than zero as illustrated
in the example of Fig. 7. In three dimensions, a kine-
matically feasible pyramid shaped wedge is inherently
stable, for a practical excavation span and under mod-
est connement, if its sides and edges fall entirely
within an inverted vertical cone with an angular radius
equal to the friction angle of the joints. This explains
why, in conned hard rock environments with rough
and unweathered joint surfaces, failures of wedges stee-
per than 458, a typical hard rock joint friction angle,
are rare [39]. On the other hand, Fig. 7 also shows
that if subsequent and surrounding mining activities
contribute to a signicant reduction in connement,
the factor of safety drops rapidly.
In low stress situations near surface or where signi-
cant relaxation has occurred due to back deection or
changes in mine geometry, steep wedges (and of course
shallow wedges as well) which would normally have
been clamped in place can be liberated and destabilized
causing serious and often catastrophic failures (Fig. 8).
At depth in the Canadian shield, for example, where
the far eld horizontal stress is 1.5 to 2 times the verti-
cal stress, isolated mining drifts will have high conn-
ing stresses across the back as shown by the example
boundary element analysis in Fig. 9(a). Such a drift
will tend to exhibit fewer groundfalls than excavations
adjacent to mining blocks (even after discounting
those failures directly induced by stope blasting). In
the latter case the stress ow is disrupted by large sub-
vertical stopes creating back relaxation (indicated by
horizontal elastic tension parallel to the back) and
groundfall potential in the access drifts (Fig. 9b).
Stopes which are aligned normal to the major principal
compressive stress can, in extreme cases, generate
Fig. 7. Example of clamped wedge stability.
extensive tensile stresses in elastic models which reect,
in reality, tangential hangingwall relaxation (Fig. 9c).
In tabular stopes in hard rock mines, inclined hang-
ingwalls, composed of blocky or laminated ground,
which are inherently stable under design conditions,
can be brought to the point of failure both in reality
(Figs. 10 and 11) and in simulation (Fig. 12) by abut-
ment softening or inelastic displacement resulting in
stress relaxation within the hangingwall. Fig. 13 shows
the results of stress change monitoring in the hanging-
wall, in the vicinity of the cablebolt array shown in
Fig. 10, as the stope was mined past the array. All
directions show a signicant measured reduction in
stress in the hangingwall due to stress shadowing and
also due to relative abutment relaxation. The model in
Fig. 12 is a discrete element model with elastic deform-
able blocks representing the hangingwall and sur-
rounding rock mass. The stope in Fig. 12 was highly
stable until the lower hangingwall abutment was sof-
tened (using viscous constitutive behaviour [12]) to
account for the undercutting due to the crosscut
shown in Fig. 10. This undercut and the induced relax-
ation in the hangingwall resulted in failure both in the
model and in reality [12].
Gravity driven failure of stress fractured ground can
also be triggered by relaxations caused by mining of
adjacent panels. Stress driven or structurally controlled
failures can induce unraveling and large scale caving
of adjacent fractured ground [13]. The creation of
intersections in jointed or fractured rock masses
reduces the arching ability or the rock mass in the
back by extending outwards the zone of deformation.
This is equivalent to relaxing the abutments of a tun-
nel span. Barton et al. [40] suggest that the destabiliz-
ing impact of intersection creation is equivalent to a
minimum 50% reduction in eective rock mass qual-
ity, a factor conrmed by Diederichs and
Hutchinson [39]. Other examples of relaxation mechan-
isms are illustrated in Fig. 13.
The inuence of low tangential connement across
excavation spans is considered explicitly in the Q
classication system [40]. For an excavation near sur-
face or with near-zero stress conditions, the impact on
stability, relative to moderate depth and connement,
is equivalent to one order-of-magnitude reduction in
rock mass quality, Q. In other words, there is a corre-
sponding reduction of up to 60% in allowable unsup-
ported span, due to relaxation, for a given rock mass.
As discussed in the following sections, relaxation can
be included in the modied stability graph analysis [41]
for stope stability. Relaxation causing near zero stress
conditions tangential to excavation spans can be
shown to reduce the self-supporting capacity of an ex-
cavation in structured or fractured ground.
Furthermore, there is conclusive evidence [42, 43]
that relative stress relaxation tangential to excavation
surfaces can drastically impact on the performance of
frictional support systems such as plain strand cable-
bolts. The relaxation measured in Fig. 11, which led to
the failure in Fig. 10, was also linked to measured re-
duction in cablebolt interface bond strength. The
increase in gravity demand coupled by the reduction in
Fig. 8. Liberation and failure of steep wedges due to (a) low stress near surface and (b) localized deection-induced tangential relaxation.
support capacity explains the often catastrophic nature
of relaxation induced groundfalls in mining.
Erosion of residual tensile strength with time and
with dynamic disturbance as well as relaxation due to
mine geometry changes or rock movements can be im-
portant mechanisms causing the delayed failure and
collapse of excavations. In order to further illustrate
the impact of residual tensile strength on excavation
stability, to demonstrate the destabilizing impact of
abutment relaxation and to provide a simplied tool
to assess the impact of these factors for at back or
hangingwall design, the voussoir beam analogue for
Fig. 9. (a) High induced compressive (horizontal) stresses aligned parallel to back of isolated mining drift (s
H
/s
V
=2); (b) induced tension or
relaxation parallel to drift back adjacent to large subvertical stope and (c) induced hangingwall relaxation (expressed as equivalent elastic tension
parallel to face) due to long stope axis perpendicular to major principle stress. Stresses are in MPa.
Fig. 10. Failure of jointed hangingwall due to relaxation of lower abutment above crosscut (after Ref. [39]).
stability of laminated ground will be utilized in Section
5. Again it should be noted that the simplicity of the
following analogue is not intended to explicitly rep-
resent the normally complex mechanisms of relaxation.
It is a tool to obtain an understanding of the impact
of tensile strength and relaxation and to modify
empirical recommendations accordingly.
5. The voussoir mechanism of self support
The voussoir beam forms in laminated or blocky
ground when the tensile strength is reduced to zero in
the radial or normal direction (Fig. 14) due to
through-going fractures perpendicular to the beam.
The symmetrical distribution of compression and ten-
Fig. 11. Stress changes monitored in the hangingwall during the excavation in Fig. 10 (after Ref. [12]).
Fig. 12. UDEC simulation of failure in Fig. 10. Softening of lower abutment in model was required to induce failure (after Refs. [12, 20]).
sion through a cross-section of the elastic beam is
replaced by a compressive arch (Fig. 15a) which varies
in thickness but is typically between 0.5 and 0.75 times
the beam thickness, T, at midspan for a highly stable
beam [44]. The compression arch thickness reduces to
below 0.5T as the critical span is approached as orig-
inally suggested by Evans [45]. Failure of the beam is
by snap-through or by crushing at the upperside of the
beam at midspan where the compressive stress is the
highest.
The problem is statically indeterminate but can be
solved in an approximate fashion using simplifying
assumptions [45, 46] or accurately through a iterative
procedure introduced by Brady and Brown [38] and
modied by Diederichs and Kaiser [44]. Only a brief
summary is presented here. The basis of the solution is
to balance the moment generated at the abutment by
the self-weight of the half-span with the opposing
moment generated by the oset reaction force, F, at
the midspan. (Fig. 15b). Two key independent
unknowns are the thickness of the compressive arch,
NT, and the moment arm between the reaction resul-
tants at the abutment and at midspan, Z. An iterative
solution which minimizes the peak compressive stress,
f
max
, at the abutments and midspan results in the equi-
librium solution for N and Z. The procedure presented
in Diederichs and Kaiser [44] is followed here with the
inclusion of support pressure or rock mass tensile
strength and the incorporation of abutment relaxation.
The reaction stresses at the abutment and at the
midspan section are assumed to be triangular such
that the reaction force, F, acts at the one-third point
Fig. 13. Relaxation due to (a) and (b) unfavourable stress ratio; (c) and (d) changes in mining geometry (excavation step 2 creates stress shadow
around excavation 1); (e) abutment yield; (f) intersection (roof relaxes in the direction of tunnel branches); (g) undercut; (h) concave geometries.
of the distribution (Fig. 15b). The reaction distri-
butions at the midspan and at the abutments are
assumed to be identical such that the initial moment
arm is given by
Z
0
= T
1
2
3
N
(15)
The reaction arch along which F acts is parabolic such
that the length of the arch can be approximated by:
L = S
8
3S
Z
2
0
(16)
Elastic shortening of the arch DL, due to compression
leads to a downward deection at midspan and a new,
reduced moment arm Z:
Z =
3S
8
8
3S
Z
2
0
DL
s
(17)
The deection, D, at midspan is given by (Z Z
0
) and
a negative value for the term under the square root
sign in Eq. (17) indicates that the critical beam deec-
tion for the assumed thickness, NT, has been exceeded.
Prior to calculating the shortening of the arch due
to deection and compression, it is possible to intro-
duce an symmetrical displacement d
A
, acting in oppo-
site directions at each abutment. This displacement
yields a reduced initial moment arm, Z
0
*
:
Z
*
0
=
3S
8
8
3S
Z
2
0
2d
A
s
(18)
for substitution into Eq. (17).
For equilibrium, consider only the right hand half
of the beam and equate moments about the abutment
reaction point (right side of the arch) generated by the
self-weight of the half beam and the reaction at the
midspan. The peak compressive strength a half span
for a horizontal beam is given by:
f
max
=
gS
2
4NZ
(19)
The shortening of the arch, dL, is calculated by assum-
ing a parabolic distribution [43] of stress along the
length of the reaction arch:
DL =
L
E
f
max
2
9
N
3
(20)
Eqs. (19) and (20) are for the span of a tunnel which is
Fig. 14. Voussoir beam formed by subvertical joints and surface par-
allel laminations.
Fig. 15. (a) Nomenclature for voussoir beam calculation; (b) Solution parameters for the voussoir arch.
much longer than it is wide. For a square span, these
equations can be replaced with the following:
f
max
=
gS
2
6N
Z
0
Z
(21)
DL =
L
E
f
max
2
9
N
3
(1 n) (22)
For evenly distributed support pressure p distributed
evenly over the length of the beam, a solution can be
obtained by substituting an equivalent unit weight, g
*
:
g
*
= g
p
T
(23)
into Eq. (19). If g
*
=0, then the beam is fully sup-
ported and no voussoir deection can occur. For tri-
angular distributions of support pressure varying from
0 at the abutment to p at the midspan, use p
*
=2/3p
in Eq. (23).
An iterative solution is needed to nd the equili-
brium value for Z corresponding to NT, the thickness
of the compression arch. Discrete interval values of N
are taken over the range of 0 < N<1. The parametric
pair (N, Z(N)) which minimizes f
max
give the equili-
brium solution for the stable beam. The limit of stab-
ility is determined when no solution is possible for
Eq. (18) for any possible value of N and the beam fails
by `snap through'. For the two-dimensional case this
point corresponds to a deection at the midspan equiv-
alent to approximately 25% of the thickness.
Diederichs and Kaiser [44] propose a more conserva-
tive stability threshold based on the onset of snap-
through instability (deviation from a linear deection
thickness relationship). This limit corresponds to a
deection of 10% of the thickness and is dened by
the parametric set which yields a minimum of 35%
invalid values of N in the range 0 to 1 (i.e. no real sol-
ution for Eq. (17)). The range of unstable displace-
ments (10 to 25%) of the thickness corresponds to the
results of numerical experimentation (H15%) obtained
by Mottahed and Ran [47].
The peak stress in the beam, f
max
, is compared with
the unconned compressive strength, UCS. In ad-
dition, failure can occur through snap thru, through
crushing or through vertical shear at the abutments if
the reaction pressure is insucient to generate ade-
quate friction. Summary design charts, based on the
rst two failure modes, for horizontal beams and
square slabs are given in Fig. 16.
6. Boundary normal tensile strength and the voussoir
beam
From Fig. 16 it is immediately apparent that thicker
laminations are more stable than thinner laminations.
This is because the compression arch is able to rise
higher within a thicker beam increasing the resisting
moment and reducing the necessary deections
required to reach equilibrium. A condition for the
transmission of compressive stresses within the arch is
that there is continuity through the beam thickness.
This may not be satised in partially separated lami-
nations with small and sparse rock bridges. A pro-
portion of the beam's weight, however, can be
transferred to the next beam above through tensile
stress in the rock bridges. This is analogous to a sup-
port pressure acting from above the beam. If the dis-
tributed capacity of the rock bridges is equal to the
self weight of the rock beam it can be considered fully
supported and fully connected to the beam above
forming a beam of twice the thickness. This analogue
can be used to generate a ground reaction curve as
shown in Figs. 17 and 18. The array of linear curves
represent the elastic response for varying lamination
thicknesses as indicated on the left-hand axis. The
required support pressure or boundary normal resist-
ing stress is indicated on the right-hand axis. The com-
posite curve for a laminated rock mass is shown as the
dark curve, obtained as the upper limit for required
tensile support capacity.
Fig. 17 illustrates the minimum support require-
ments or alternatively the minimum internal boundary
normal tensile strength required to limit the beam
deections and to achieve stability. The response
curves for typical support patterns are shown for com-
parison. The minimum lamination thickness is
assumed to be 0.1 m and the composite curve rep-
resents the results of stacking or of partial delamina-
tion (rock bridges). In this case both the support
pressure and the tensile resistance are assumed to be
constantly distributed across the span. For tensile
strength this is likely to be a valid assumption due to
the small strain nature of rock bridge rupture. The ca-
pacity requirements (support or tensile demand)
increased with reduced rock mass modulus as shown
in Fig. 18. These demand values are very low for the
span in question, falling in the range of residual tensile
strength presented in Figs. 3 and 5 for small rock
bridges.
The limiting tensile strength (required for stability)
as a function of span is summarized in Fig. 19. In
Fig. 19, the upper and lower bounds for tensile
strength are dened by initial yield (buckling
limit =35%, deection=10% of thickness) and absol-
ute collapse (buckling limit =100%). Fig. 20 illustrates
the impact of modulus.
Again, it is important to note the extremely small
values of tensile strength required for stability of large
spans. This is not surprising in light of the large num-
ber of large spans found in natural caverns, overhangs
and arches. The strength required to resist gravity
loading is commonly found even in jointed rock due to
the lack of complete persistence and the presence of
rock bridges. If the self supporting and arching ability
of moderate quality rock masses is considered, as in
the voussoir analogue, the tensile strength require-
ments for stability are further reduced.
Support is still required, however, in many appli-
cations in moderately jointed rock, particularly those
with longer or undetermined stand-up time require-
ments. In mining it is commonly and correctly held as
imprudent to leave fractured ground unsupported (if
human trac is a concern) in spite of the fact that re-
sidual tensile strength may create a self-supporting
opening over the short term. The dynamic nature of
mining combined with the eects of moisture on long-
term tensile strength can lead to delayed failure of
unsupported ground. The presence of rock bridges,
however, provides signicant capacity reserve at small
deformations, before tensile strength degradation is
allowed, thus reducing rst pass or short term support
requirements in some cases. Very sti support such as
grouted rebar can serve to reduce this degradation,
leading to economic benets through the use of shorter
tendon support elements designed to preserve tensile
Fig. 16. Voussoir design charts for horizontal tunnel and square span (g =30 kN/m
3
); Maximum stable span for a given thickness is the mini-
mum value determined from Erm or UCS limit.
strength and allowing voussoir-like arching, carrying
load to the abutments.
7. Abutment relaxation and the voussoir beam
Another dominant factor in the reduction of stand-
up time for unsupported (and inadequately supported)
excavations is relaxation, parallel to the excavation
face, which is inuentially, if not mechanistically,
equivalent to an outward displacement of the abut-
ments. Fig. 21 illustrates the physical implementation
of relaxation for the voussoir beam analogue. The
quoted displacements in the following gures apply to
a single abutment. An equal and opposite displacement
is assumed at the other abutment. Before beam deec-
tion in this schematic and in the classic Voussoir ana-
logue, the beam is assumed to be stress free,
corresponding to a zero local in situ stress. The relax-
ation displacement shown is based on this datum and
Fig. 17. Composite ground reaction curve for laminated rock mass based on voussoir calculations: (span=20 m; E=10 GPa, g =30 kN/m
3
;
UCS=25 MPa); response curves for typical support patterns are shown for comparison; internal tensile strength refers to boundary normal
strength.
Fig. 18. Ground reaction curve: relationship between rock mass modulus and tensile support demand (span=20 m; buckling limit,
B.L. =100%).
does not include additional displacement incurred as
initial compressive in situ stress and strain reduce to
zero.
In other words, the displacement pictured here is
equivalent to the presence of tangential tension in the
back of an excavation in an elastic continuum such as
that simulated in Fig. 9. In Fig. 9b, the average ten-
sion adjacent and parallel to the back of the 10 m
wide access drift was approximately 5 MPa. Using an
assumed rock mass modulus of 10 GPa, for example,
the equivalent symmetrical abutment displacement, d
A
,
as per Fig. 21 is equated as 2.5 mm. In Fig. 9c, the
average tangential tension (10 MPa) across the 50 m
hangingwall yields an equivalent d
A
, at both the topsill
and the bottomsill, of approximately 25 mm. These
examples represent only a rst approximation and will
likely underestimate the true analogous relaxation dis-
placement since stress shedding is not taken into
account in the elastic model. Admittedly it is very di-
cult to accurately estimate d
A
for many situations.
This should not, however, be a deterrent to the con-
sideration of relaxation in design.
The impact of this abutment deformation or relax-
ation is to increase the beam deection required to
generate sucient compressive stress and resisting
moment for equilibrium. This additional deection
brings the beam closer to failure. This eect is illus-
trated in the ground reaction diagram in Fig. 22. The
individual curves are generated in a similar fashion to
the composite curve in Fig. 17, assuming that thin
laminations would stack and create a composite beam
Fig. 19. Limiting demand for boundary normal tensile strength in
laminated ground (modulus =10 GPa; collapse B.L. =100%; yield
B.L. =35%).
Fig. 20. Eect of modulus on minimum tensile strength demand (B.L. =100%).
Fig. 21. Schematic abutment relaxation and destabilization of a
voussoir beam.
with a thickness determined by the available tensile
strength (rock bridges) or articial support. For com-
parison, an abutment relaxation of 10 mm over a half-
span of 10 m is equivalent to a tensile strain (parallel
to the excavation face) of 0.1% or 10 MPa of analo-
gous tensile stress using a modulus of 10 GPa. This is
realistic for deep mining excavations.
This abutment relaxation can occur as a result of
changing geometry around the excavation in question,
by unfavourable stress ratios or by increased deec-
tions due to the creation of intersections or undercuts
(Fig. 13). The latter is a source of large and unfavour-
able abutment deformations and is responsible for nu-
merous failures observed by the authors and by
others [12]. Figs. 23 and 24 show the corresponding
increase in support or tensile strength demand due to
abutment relaxation. The term yield in these gures
corresponds to midspan deections equivalent to 10%
and a buckling limit, B.L., of 35% as discussed earlier.
The term collapse (B.L. =100%) describes the least
conservative solution. Figs. 23 and 24 show that very
little tensile strength (normal to laminations) can com-
pensate for abutment deformation, but that the maxi-
mum span to ensure the marginal stability of
Fig. 22. Ground reaction curves for laminated ground illustrating the impact of abutment relaxation (modulus =10 GPa, B.L. =35%;
span=20 m).
Fig. 23. Impact of abutment relaxation on the minimum demand for boundary normal tensile strength (B.L. =100%).
unsupported openings (stopes) decreases signicantly
in the presence of relaxation if additional strength or
support is not available.
8. Critical span reduction by relaxation
For unsupported beams, there exists a critical relax-
ation at which failure (snap-thru) occurs. For a rock
mass with a modulus of 10 GPa, Fig. 25 shows the
critical relaxation (as a function of span and thickness)
for yield and collapse of a horizontally laminated roof
(without rock bridges). Critical relaxation varies, of
course, with rock mass modulus as illustrated in
Fig. 26a). In order to compare the absolute relaxation
displacements in Figs. 25 and 26 to modeled strains
and equivalent tensile stresses in elastic stress analysis
models, Fig. 26b) provides a conversion chart.
Remember that the displacements indicated are
`measured' from the state of zero stress and do not
include recovery of the initial compressive strain pre-
sent in situ.
9. Relaxation and the modied stability graph
To explore the eect of relaxation on empirical
stope design methods, it is desirable to apply the vous-
soir analogue for relaxation to more general exca-
vation design. This can be achieved by calibrating
voussoir analysis to an existing empirical design limit.
The stability graph method was rst proposed by
Mathews et al. [41] and later modied and calibrated
for Canadian mining conditions by Potvin [48].
Additional data has been added [16, 49] and the
method is now a popular tool in Canadian open stope
design. The characterization of the rock mass by
means of a stability number N
/
is based initially on the
Q-system [40]. The stress reduction factor is dropped
(set to 1) as is the joint water factor (since deep mines
Fig. 24. Same simulation as in Fig. 23 showing limits for yield (B.L. =35%) and collapse (B.L. =100%).
Fig. 25. Critical relaxation limits for unsupported voussoir beam (Erm=10 GPa).
are typically dry). Three new factors A, B and C are
dened to account for induced stress, relative joint
orientation and gravity loading respectively such that:
N
/
=
RQD
Jn

Jr
Ja
A B C = Q
/
A B C (24)
where A, B and C are dened in Fig. 27.
Excavations of dierent geometries are equated
through the use of a shape factor or hydraulic radius,
HR, which describes the geometry and size of the
stope face in question. For a rectangular stope face of
dimensions a b:
HR =
Area
Perimeter
=
ab
2a 2b
(25)
Using this relationship the hydraulic radius, HR, of a
tunnel roof (with length much longer than the span) is
given as one half of the span. For a square excavation,
the hydraulic radius is equivalent to one quarter of the
span. The hydraulic radius expression, therefore,
denes that a horizontal tunnel is equally as stable as
a square excavation surface of twice the linear span
(measured orthogonal to the face edge). This assump-
tion creates some diculties which will be discussed.
A calibration database of 189 case histories invol-
ving unsupported stopes has been assembled by
Potvin [48] and Nickson [49]. The database and the
resulting no-support limits are shown in Fig. 28.
In order to link the voussoir analogue to the modi-
ed stability graph, it is necessary to relate the vous-
soir parameters to rock mass quality and the modied
stability number N
/
(Fig. 29). In the absence of com-
prehensive case data, an iterative procedure was used,
bounded by logical upper and lower parametric limits.
Fig. 26. (a) Eect of modulus on critical relaxation and (b) equivalent relaxation stress.
Fig. 27. Stability parameters for the modied stability graph (after Ref. [39]).
The values of A, B and C are taken to correspond to
the voussoir assumptions such that A=1 (low stress),
B=0.3 (surface parallel jointing and C=2 (horizontal
roof). this yields a relationship between N and Q
/
:
Q
/
= Na0X6 (26)
Next a simplied relationship between rock mass mod-
ulus and Q
/
in a moderately relaxed setting proposed
by Diederichs and Kaiser [44] and a similar relation-
ship for unconned compressive strength (UCS) are
assumed:
E
rockmass
(GPa) = 5
Q
/
p
I6X5
(27)
UCS (MPa) = 20
Q
/
p
I25X8
(28)
These relationships are used here, not as engineering
recommendations, but rather as reasonable baseline
assumptions in order to create a generalized model to
Fig. 28. Unsupported stope data from Refs. [48, 49] and no-support limits for modied stability graph (after Ref. [39]).
Fig. 29. Variation with respect to N
/
of voussoir parameters used in simulation (for model calibration purposes only).
assess the eects of relaxation. In addition, while intact
UCS is independent of Q, it is related to A. A relation-
ship (Eq. (28)) is included in this calibration since in
hard rock masses, increased RQD and lower Q is in
general associated with weaker rock such as shists,
rhyolites, talcs and meta-shales, for example. The eec-
tive UCS values resulting from this relationship rep-
resent approximately one half of the nominal lab
strength [9]. Using these relationships, for E and UCS,
in analyses of voussoir stability for dierent hydraulic
radii and lamination thicknesses, it is possible to derive
the following relationship for N
/
as a function of criti-
cal lamination thickness for unsupported openings (in
metres):
N
/
= 150 (Thickness)
3
(29)
While this calibration procedure is somewhat subjec-
tive and is conned in its application to the assump-
tions described, it yields reasonable parametric ranges
for thicknesses between 0.1 to 1 m. `Yield' calculations
were used for the voussoir simulation (buckling
limit =35%), corresponding to a long-term no-support
limit for span. The calibration resulting in the para-
metric set in Fig. 29 is based on the upper no-support
limit in Potvin's [48] stability graph. This limit corre-
sponds to the onset of instability or yield. Fig. 29 is
not for general application and is intended for model
calibration purposes only. The resultant calibration is
shown in Fig. 30 for both the square and the long
(tunnel) span.
The hydraulic radii for critical square openings are
uniformly 0.77 time that of critical tunnel spans for
equivalent N values. This results from a relative over-
prediction of square span stability inherent in the de-
nition of hydraulic radius. The denition of HR infers
that a square span is as stable as a long tunnel of half
(0.5 times) the linear span. Voussoir predicts that a
square opening of span, S, is as stable as a long tunnel
of span 0.65S. The hydraulic radii, therefore, of critical
square spans (as predicted by voussoir simulation) are
0.77 times the HR of critical tunnel spans (for a uni-
form set of voussoir parameters). This discrepancy cor-
responds closely to a correction (approximately 0.72)
proposed by Milne [14] using a radius factor based a
harmonic average distance from the centre of the span
to the perimeter. It can be shown that the hydraulic
radius for a circular face is equal to that for the corre-
sponding circumscribed square. Failure to account for
the additional mass and deformation due to the cor-
ners of the square is partly responsible for this error.
For the purposes of this calibration, the square span is
used to determine HR. HR's derived from tunnel
spans (2D voussoir calculations) should be multiplied
by 0.77.
Now that the voussoir simulation is calibrated to
the stability graph it is possible to simulate the impact
of abutment relaxation on the no-support limit. The
stress parameter A does not consider relaxation, unlike
the stress reduction factor, SRF, which it replaced.
This factor monotonically increases from a low of 0.1
at high stress to a maximum of 1 at low or zero
induced stress. A is set to 1 in this simulation (no
stress) so that the impact of relaxation can be exam-
ined explicitly. In the calibrated voussoir calculations,
outward abutment deection corresponds to relax-
ation. Fig. 31 shows the new no-support limits for
varying degrees of abutment relaxation. It is apparent
Fig. 30. Voussoir calibration for square span. Linear tunnel spans
require geometric correction.
Fig. 31. Revised no-support limits for N
/
resulting from abutment
relaxation (negative relaxation values in mm correspond to stope
wall compression).
that very little relaxation displacement is required to
signicantly increase the potential for instability.
Fig. 32 shows the same results plotted against the
log of the hydraulic radius. Using the modulus re-
lationship in Fig. 29 and the relaxation-stress equival-
ence calculated in Fig. 27(b), it is possible to express
the no-support limit alternatively as a function of
induced average tensile stress tangential to the span (as
derived from elastic models). For the voussoir ana-
logue, relaxation is dened with respect to the position
of the abutments under zero lateral stress (or strain).
Therefore the relaxation displacement used in the
voussoir analysis corresponds directly to equivalent
tensile stress (parallel to the span) in an elastic model.
The eect of relaxation in terms of the equivalent ten-
sile stress is shown in Fig. 33. For competent rock
masses, modeled elastic tensile stress magnitudes, of
appreciable spatial extent, in excess of 30 MPa are
rare. If inelastic compressive strain has occurred adja-
cent to the excavation, as in the case of joint slip or
abutment crushing, full relaxation (tension) may not
be necessary to cause failure. In previously stressed
Fig. 32. No-support limits resulting from abutment relaxation plotted with respect to log HR: function shown is for Potvin's [48] upper no-sup-
port limit.
Fig. 33. Equivalent elastic tensile stresses corresponding to the relaxation displacements in Fig. 32.
ground around excavations, any signicant reduction
in lateral compression in the surrounding regional rock
mass may result in relaxation induced groundfall.
Field evidence conrms the impact of relaxation as
illustrated in Fig. 34. A subset of the data from
Greer [50] and reanalyzed by Bawden [51] is plotted in
the lower half of Fig. 34, corresponding to horizontal
stope backs. Bawden [51] reports that these backs are
under high compression (factor A=0.1 to 0.4). The
data seems to correspond well to the original stability
limits proposed by Potvin [48]. Data for the inclined
hangingwalls are plotted in the upper portion of
Fig. 34. Bawden [51] reports that many of these faces
have tangential tensile stresses according to elastic
models. In spite of this relaxation eect, the stress fac-
tor A must be set to its maximum (=1) in accordance
with the stability graph procedure. Note that Mathews
et al. [41] in their original discussion of this method
state that it is not valid for cases of tangential relax-
ation. In the data of Fig. 34, the method overpredicts
the stability of these hangingwalls. Adjusted stability
limits corresponding to average elastic tangential ten-
sion values of 4, 8 and 16 MPa are superimposed on
Fig. 34. While the magnitude of tension in these cases
is not documented, the value shown is not unreason-
able for typical hangingwall geometries in elastic ana-
lyses as indicated in Fig. 9(c). If the stability limit is
shifted up as indicated by these stress-dened curves,
the unstable stopes can be appropriately captured.
This is equivalent to a corresponding reduction in N
/
for the aected data points.
Using the apparent shift in N
/
, illustrated in Figs. 33
and 34, as a result of tangential tensile stress or relax-
ation, it is possible to derive an approximate function
for a revised stress factor A which reects the inuence
of relaxation. The modeled tensile stresses are rst nor-
malized with respect to a range for nominal UCS of
intact lab samples which is typical of hard rock (e.g.
200 MPa at N
/
=10). The relationship illustrated in
Fig. 35 can be obtained from an examination of
Figs. 33 and 34. An average shift of N (applied
through the A factor), for a practical range of HR, is
determined for dierent values of tension in order to
provide for a simple and general correction to A. The
general equation for factor A adjustment:
A = 0X9e
11(s
T
aUCS)
for s
T
`0 (30)
for tangential boundary stress values (s
T
) in tension, is
obtained as an exponential best t. For practical pur-
poses a linear t with extrapolation limits may also be
adequate given the uncertainties in relating equivalent
relaxation displacement with elastic tensile stress.
It is important to realize that other failure mode
such as wedge fallout are likely to be even more sensi-
tive to relaxation than the voussoir beam. The destabi-
lizing eect of relaxation, therefore, may be
signicantly greater than that illustrated in Fig. 35 and
Eq. (30). It is important to note that this relationship,
Fig. 34. Correlation of relaxation adjustment for upper no-support limit (data from Refs. [50, 51]); Stope backs were under moderate to high
compression while elastic models predicted tension in hangingwalls.
derived through the voussoir analogue is valid only for
low to moderate relaxation or tension. For higher
values of tension, it has been suggested that A drops
to zero and that the method, as a whole, becomes
invalid [41]. The best t coecient 0.9 in Eq. (30) cre-
ates an intercept which indicates an initially vertical
slope to the relationship. In fact it is likely that the
left-hand decline (Fig. 35) in the A factor begins at
compressive stresses below several MPa [16] particu-
larly if wedges are present. It should be emphasized,
again, that Eq. (30) represents an upper bound for the
A factor and that the impact of relaxation can be
much greater than that indicated.
The results presented here can also be applied to the
no-support limit of the Q-system [40] as initially dis-
cussed in Ref. [37]. Using the calibration presented
here and plotting against span and Q, instead of HR
and N
/
, and assuming dry conditions and an initial
SRF of 2, Fig. 33 can be replotted with respect to the
Q-system as shown in Fig. 36. The equivalent tunnel
spans are calculated by dividing the stable HR values
from the square (voussoir) spans by 0.77 (long span
correction according to voussoir analysis). Note that
the base-line no-support limit for tunnels (long spans)
corresponds to the limit for ESR=3, while that for
the square span corresponds to ESR=5. Barton [52]
Fig. 35. Revised denition of stress factor, A, used to determine the modied stability number, N
/
. For consistency with compressive
function [48], relaxation is reected as the normalized maximum tension parallel to the face near midspan.
Fig. 36. Tangential elastic tensile stresses corresponding to abutment relaxation and the shift in the Q no-support limit; Q limits are calculated
from N
/
limits in Fig. 33; Barton's [52] ESR limits are plotted for comparison.
dened this range of stability suitable for temporary
mine openings. Hutchinson and Diederichs [39] rede-
ne the limits ESR=3 to correspond to open stope
backs and limited access drifts while ESR=5 is suit-
able for non-entry mining stopes and stope walls.
These denitions are consistent with the results shown
in Fig. 36. Note again, however, the eect of relax-
ation and the additional design conservatism which
must be implemented to compensate for this relax-
ation.
10. Conclusions
The presence of geometrically unfavourable jointing
or rock mass damage due to elastic predictions of
overstress does not, in itself, dictate that failure will
take place. Neglecting the possibility of dynamic pro-
cesses such as compression-induced buckling, degra-
dation of residual tensile strength or abutment
relaxation is often required before failure can take
place.
Rock masses are inherently discontinuous due to
natural jointing or induced fracturing. It is often erro-
neous to assume that this fracturing is fully persistent.
In massive to moderately jointed rock residual tensile
load bearing strength arising from incomplete fractur-
ing or from rock bridges separating non-persistent
joining is a key factor in the control of ultimate grav-
ity driven failure of jointed or stress damaged ground.
Very little rock bridge cross-sectional area (less than
1% in most cases) is required in hard rocks to replace
most articial support systems. The time dependency
of this residual tensile strength due to stress corrosion
and atmospherically induced crack growth controls
stand-up time and mandates the use of support sys-
tems in most underground excavations. Sti support
such as grouted rebar can suppress dilation strains and
preserve some of this internal tensile strength, contri-
buting to a reduction in both short term and long
term support requirements. There is great economic
advantage to selecting, as part of a multi-component
support system, sti elements which can preserve the
rock mass internal tensile capacity. Careful blast
damage control is also an obvious advantage. Early
installation of shotcrete or spray-on linings in a dry
rock mass at depth can reduce the impact of atmos-
pheric eects and time dependent stress corrosion on
stressed rock bridges.
Boundary parallel relaxation is another dominating
factor in delayed mining induced failure of spans in
underground excavations, signicantly shifting conven-
tional no-support limits so that smaller spans or ad-
ditional support are required. Dangerous abutment
relaxation can occur even at depth, driven by un-
favourable stress ratios, complex mine geometries,
abutment damage, intersection development and
undercutting. Abutment relaxation increases support
demands by reducing the natural ability of the rock
mass to transfer loads to the abutments through arch-
ing. In addition, boundary normal stress relaxation
has been shown to reduce the capacity of frictional
support systems, exacerbating ground control pro-
blems.
A voussoir beam analogue was used to illustrate the
importance of internal boundary normal tensile
strength and of abutment relaxation in controlling the
stability of spans in laminated rock masses. The results
of the voussoir simulation were used to modify empiri-
cal stope design limits, accounting for abutment relax-
ation. A few millimetres of hangingwall or back
abutment relaxation can signicantly shift the no-sup-
port limit, inducing failure in previously stable spans.
It is therefore important to sequence development and
stope extraction properly to minimize this relaxation
and to minimize the size of secondary stopes. The cre-
ation of high relaxation geometries, such as hanging-
wall undercutting, must be avoided.
Acknowledgements
This work was funded by the Natural Science and
Engineering Research Council of Canada.
References
[1] Hoek E, Brown ET. Underground excavations in rock. London:
Inst. of Min. and Metall., 1980. 527 pp.
[2] Hoek E, Brown ET. The HoekBrown failure criterion: a 1988
update. Proc. 15th Canadian Rock Mech. Symp. University of
Toronto: Department of Civil Engineering, 1988. p. 318.
[3] Pelli F, Kaiser PK, Morgenstern NR. An interpretation of
ground movements recorded during construction of the
DonkinMorien tunnel. Can. Geotech. J. 1991;28(2):23954.
[4] Kaiser PK. Observational modelling approach for design of
underground excavations. The Application of Numerical
Modelling in Geotechnical Engineering Symposium. ISRM S.
African National Group, 1994. p. 18.
[5] Brace WF, Paulding BW, Scholz C. Dilatancy in the fracture of
crystalline rocks. J. Geophys. Res. 1966;71(16):393953.
[6] Fairhurst C, Cook NGW. The phenomenon of rock splitting
parallel to the direction of maximum compression in the neigh-
bourhood of a surface. Proc. of the 1st Congr. of the Int. Soc.
of Rock Mechanics. 1966. p. 68792.
[7] Tapponier P, Brace WF. Development of stress induced micro-
cracks in Westerly granite. Int. J. Rock Mech. Min. Sci.
Geomech. Abstr. 1976;13:10312.
[8] Fonseka GM, Murrell SAF, Barnes P. Scanning electron micro-
scope and acoustic emission studies of crack development in
rocks. Int. J. Rock Mech. Min. Sci. Geomech Abstr.
1985;22(5):27389.
[9] Martin CD. The strength of massive Lac du Bonnet Granite
around underground openings. Ph.D. Thesis, University of
Manitoba, 1993.
[10] Landriault D, Oliver P. The destress slot concept for bulk
mining at depth. Rock support in mining and underground con-
struction. Rotterdam: Balkema, 1992. p. 2117.
[11] Castro L, McCreath D, Oliver P. Rock mass damage initiation
around the Sudbury Neutrino Observatory Cavern. Rock
Mechanics: Proc. Of the 2nd North American Rock Mechanics
Symposium, vol. 2. Rotterdam: Balkema, 1996. p. 158995.
[12] Kaiser PK, Maloney S. The role of stress change in under-
ground construction. Eurock '92. Rotterdam: Balkema, 1992. p.
396401.
[13] Kaiser PK, Diederichs MS, Yazici S. Cablebolt performance
during mining induced stress change: three case examples. Rock
support. Rotterdam: A.A. Balkema, 1992. p. 37784.
[14] Milne D. Underground design and deformation based on sur-
face geometry. Ph.D. Thesis, Department of Mining and
Mineral Processing, University of British Columbia, 1996.
[15] Falmagne V. Quantication of rock mass degradation using
microseismic monitoring. Application to mine design. Ph.D.
Thesis, Department of Mining Engineering, Queen's University,
1998.
[16] Suorineni FT. Eects of faults and stress on stope design. Ph.D.
Thesis, Department of Earth Sciences, University of Waterloo,
1998.
[17] Priest SD, Hudson JA. Estimation of discontinuity spacing and
trace length using scanline surveys. Int. J. Rock Mech. Min.
Sci. Geomech. Abstr. 1981;18:18397.
[18] Kulatilake PHSW, Wu TH. Estimation of mean trace length of
discontinuities. Rock Mech. Rock Eng. 1984;17:21532.
[19] Villaescusa E, Brown ET. Maximum likelihood estimation of
joint size from trace length measurements. Rock Mech. Rock
Eng. 1992;25:6787.
[20] Maloney SM, Kaiser PK. Field investigation of hanging wall
support by cable bolt pre-reinforcement at Winston Lake Mine.
Research Report. Geomechanics Research Centre, Laurentian
University, Canada, 1993. 140 pp.
[21] Lajtai EZ, Bielus LP. Stress corrosion cracking of Lac du
Bonnet granite in tension and compression. Rock Mech. Rock
Eng. 1986;19:7187.
[22] Fairhurst C. Fundamental considerations relating to the
strength of rock. Colloquium on Rock Fracture. Department of
Geophysics, Ruhr University, Bochum, 1971. 56 pp.
[23] Dyksin AV, Germanovich LN. Model of rockburst caused by
cracks growing near free surface. Rockbursts and seismicity in
mines. Rotterdam: Balkema, 1993. p. 16974.
[24] Stacey TR, Page CH. Practical handbook for underground rock
mechanics. TransTech, Germany, 1986, 145 pp.
[25] Grith AA., Theory of rupture. 1st Int. Congr. Applied
Mechanics. Delft, 1924. p. 5563.
[26] Irwin GR. Analysis of stresses and strains near the ends of a
crack traversing a plate. J. Appl. Mech. 1957;24:3614.
[27] Sih GC. Handbook of stress intensity factors. Inst. of Fracture
and Solid Mechanics, Lehigh University, Bethlehem, 1973.
[28] Kemeny J, Cook NGW. Eective moduli, non-linear defor-
mation and strength of a cracked solid. Int. J. Rock Mech.
Min. Sci. Geomech. Abstr. 1986;23(2):10718.
[29] Berry JP. Some kinetic considerations of the Grith criterion
for fracture. Part I: equations of motion at constant force. J.
Mech. Phys. Solids 1960;8:194206.
[30] Pollard DD, Segall P. Theoretical displacements and stresses
near fractures in rock: with applications to faults, joints, veins,
dykes and solution surfaces. Fracture mechanics of rock.
London: Academic Press, 1987. p. 277351.
[31] Okubo S, Fukui K. Complete stressstrain curves for various
rock types in uniaxial tension. Int. J. Rock Mech. Min. Sci.
Geomech. Abstr. 1996;33(6):54956.
[32] Stillborg B. Professional users handbook for rock bolting.
TransTech, Germany, 1994, 164 pp.
[33] Illston JM, Dinwoodie JM, Smith AA. Concrete, timber and
metals. Berkshire: Van Nostrand Reinhold, 1979. 663 pp.
[34] Atkinson BK, Meredith PG. The theory of subcritical crack
growth with applications to minerals and rocks. Fracture mech-
anics of rock. London: Academic Press, 1987. p. 11167.
[35] Dunn DE. Molecules crack rocks. Sci. News 1966;90:69.
[36] Wiid BL. The inuence of moisture on the pre-rupture fractur-
ing of two rock types. Proceedings of the 2nd Congress of the
I.S.R.M., vol. 2, issue 34. 1970. p. 23945.
[37] Kaiser PK, Falmagne V, Suorineni FT, Diederichs M, Tannant
DD. Incorporation of rock mass relaxation and degradation
into empirical stope design. 99th Canadian Institute of Mining
AGM. 1997.
[38] Brady BHG, Brown ET. Rock mechanics for underground
mining. Chapman and Hall, 1993. 571 pp.
[39] Hutchinson DJ, Diederichs MS. Cablebolting in underground
mines. Vancouver: Bitech Publishers, 1996. 416 pp.
[40] Barton N, Lien R, Lunde J. Engineering classication of rock
masses for the design of tunnel support. Rock Mech.
1974;May:189236.
[41] Mathews KE, Hoek E, Wyllie DC, Stewart SBV. Prediction of
stable excavations for mining at depth below 1000 metres in
hard rock. CANMET Report DSS Serial No. OSQ80-00081,
DSS File No. 17SQ.23440-0-9020. Ottawa: Department of
Energy, Mines and Resources, 1981. 39 pp.
[42] Kaiser PK, Yazici S, Nose J. Eect of stress change on the
bond strength of fully grouted cables. Int. J. Rock Mech. Min.
Sci. Geomech. Abstr. 1992;29(3):293306.
[43] Maloney S, Fearon R, Nose J, Kaiser PK. Investigations into
the eect of stress change on support capacity. Rock support.
Rotterdam: A.A. Balkema, 1992. p. 36776.
[44] Diederichs MS, Kaiser PK. Stability guidelines for excavations
in laminated ground: the voussoir analogue revisted. Int. J.
Rock Mech. Min. Sci., 1999;36:97118.
[45] Evans WH. The strength of undermined strata. Trans. Inst.
Min. Metall 1941;50:475500.
[46] Beer G, Meek JL. Design curves for roofs and hangingwalls in
bedded rock based on voussoir beam and plate solutions.
Trans. Inst. Min. Metall 1982;91:A18A22.
[47] Mottahed P, Ran J. Design of jointed roof in stratied rock
based on the Voussoir beam mechanism. Can. Inst. Mining
Bull. 1995;88(994):5662.
[48] Potvin Y. Empirical open stope design in Canada. Ph.D. Thesis,
Department of Mining and Mineral Processing, University of
British Columbia, 1988. 343 pp.
[49] Nickson SD. Cable support guidelines for underground hard
rock mine operations. M.A.Sc. Thesis, Department of Mining
and Mineral Processing, University of British Columbia, 1992.
343 pp.
[50] Greer GJ. Empirical modelling of open stope stability in a verti-
cal crater retreat application at INCO's Thompson Mine. 91st
Canadian Institute of Mining AGM, 1989. 12 pp.
[51] Bawden WF. The use of rock mechanics principles in Canadian
hard rock mine design. Comprehensive rock engineering, vol. 5.
Pergamon Press, 1993. p. 24790.
[52] Barton N. Rock mass classication and tunnel reinforcement
selection using the Q-system. In: Kirkaldie, editor. Rock classi-
cation systems for engineering purposes, A.S.T.M. 984.
American Society for Testing and Materials, 1988. p. 5988.

Diederichs Kaiser 1999b

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Diederichs Kaiser 1999b

Hochgeladen von

Copyright:

Verfügbare Formate

Tensile strength and abutment relaxation as failure control

mechanisms in underground excavations

for rock bridges

Das könnte Ihnen auch gefallen