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

net/publication/284015427

Analysis of tunnel support design to withstand spalling induced by blasting

Article  in  Tunnelling and Underground Space Technology · October 2015

DOI: 10.1016/j.tust.2015.10.006

CITATIONS READS

24 198

2 authors:

Amichai Mitelman Davide Elmo

University of British Columbia - Vancouver University of British Columbia - Vancouver
8 PUBLICATIONS   72 CITATIONS    105 PUBLICATIONS   1,788 CITATIONS   

SEE PROFILE SEE PROFILE

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

Cave-to-Mill: a Mine-to-Mill approach for block cave mines View project

All content following this page was uploaded by Amichai Mitelman on 16 June 2021.

The user has requested enhancement of the downloaded file.

 Tunnelling and Underground Space Technology xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

journal homepage: www.elsevier.com/locate/tust

Analysis of tunnel support design to withstand spalling induced

by blasting
Amichai Mitelman ⇑, Davide Elmo
NBK Institute of Mining Engineering, University of British Columbia, Vancouver, Canada

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents modeling of rock spalling induced by a dynamic pulse using the finite-discrete ele-
Received 3 March 2015 ment method, for the purpose of tunnel support design to withstand blasting. 1D and 2D model results
Received in revised form 13 May 2015 are compared to analytical spalling equations and field test findings. It was found that only the 2D models
Accepted 7 October 2015
are suitable for support design. A distinction between heavy spalling and light rockfall is made based on
Available online xxxx
an estimation of the ratio of the peak stress of the arriving wave to the rock tensile strength. Accordingly,
different design approaches are suggested: for heavy spalling a low impedance isolating layer between
Keywords:
the tunnel liner and surrounding rock is recommended. For light rockfall, a simplified static FEM analysis
Support design
Spalling
procedure is presented.
Blasting Ó 2015 Elsevier Ltd. All rights reserved.
Numerical analysis

1. Introduction Measurements of the amount of broken rock and the velocity of

the rock ejection into the tunnel were made as part of those tests,
A blast coupled in rock imposes extremely high pressures that so that tunnel support could be designed accordingly (note that
cause complex fracturing mechanisms to occur. The close field those tests were performed for unsupported tunnels). Although
response of the rock has been studied by many researchers: the ERA tests serve as an invaluable database for assessing blast
Grady and Kipp (1980) proposed a continuum based damage damage, using those to draw conclusions for proper support is
model to investigate oil shale fracturing; Donze et al. (1997) inves- not a straightforward process.
tigated the influence of different parameters on the size of the Field tests to investigate tunnel support performance under
crushed zone and length of the radial cracks; Zhu et al. (2008) stud- blast loads have been conducted by various authors. Most com-
ied fragmentation mechanisms around a borehole loaded with monly, peak particle velocity (PPV) is used as a damage threshold.
explosives using the continuum code AUTODYN in the context of Kendorski et al. (1973) tested bolted and lined tunnels in heavily
mining and quarrying. jointed schist using ANFO as an explosive and found that hairline
Design of a tunnel support system to withstand the close field cracks formed in the liners under PPVs of 900 mm/s and that dis-
blasting impact would be impracticable (EM 1110-345-432, placement of the cracks was associated with PPVs of 1200 mm/s.
1962). Therefore, this paper focuses on the far field response of a Wood and Tannant (1994) found that the reinforced shotcrete
tunnel to an explosive blast, where the reflection of the blast wave can maintain its functionality even when the PPV was up to
from the tunnel boundary can cause rock fragments to fly into the 1500–2000 mm/s.
tunnel (spalling). Tunnel support systems that have been designed Different authors have used numerical methods to simulate the
to carry static loading may not be sufficient to withstand these response of tunnel support to blast loads (Wu et al., 2011; Deng
additional loads. Limited guidelines exist for the engineer tasked et al., 2014). Mitelman and Elmo (2014) validated a hybrid FEM/
with designing tunnel support to withstand blast loads or evaluat- DEM (FDEM) approach to explicitly simulate damage to tunnels
ing the performance of an existing support system. induced by blast loads. In this paper, a similar approach is further
The most extensive tests of damage to tunnels from blasting applied to investigate the adequate protective support required to
took place between the years 1948–1952, conducted by Engineer- prevent blast induced damage. In general, numerical modeling is a
ing Research Associates together with the US Army Corps of Engi- rigorous process that requires several inputs and assumptions, and
neers (COE) and the US Bureau of Mines (ERA, 1953). intensive computational capabilities. Hence, there is an incentive
to simplify the problem and analysis methods. This paper discusses
⇑ Corresponding author. the validity and limitations of different approaches of tunnel sup-
E-mail address: [email protected] (A. Mitelman).
port design to withstand blasting.

http://dx.doi.org/10.1016/j.tust.2015.10.006
0886-7798/Ó 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Mitelman, A., Elmo, D. Analysis of tunnel support design to withstand spalling induced by blasting. Tunnel. Underg. Space
Technol. (2015), http://dx.doi.org/10.1016/j.tust.2015.10.006
2 A. Mitelman, D. Elmo / Tunnelling and Underground Space Technology xxx (2015) xxx–xxx

Rockbursts are another form of dynamic loading that may be presented in Fig. 1. According to the notation used in the figure, the
imposed on a tunnel. Rock bursting into the tunnel occurs as a first spall will form when:
result of high stresses in deep mines or tunnels. Kaiser and Cai
(2012) discusses the rockburst phenomenon and lays out general rm
rx ¼ rt ð1Þ
guidelines for dynamic support. Although blast induced damage
and rockbursting are triggered by different mechanisms, the two where rm is the peak stress of the wave, rt is the tensile strength of
share similarities in terms of the manner of how they impact sup- the rock material, and rx is the compressive stress of the incident
port; both are manifested by a series of small impacts, with veloc- wave at the point of the first spall. With the knowledge of the wave
ities ranging from 3 to 10 m/s. It is the authors’ opinion that work velocity c of a material, the above equation can be rearranged to
related to rockburst support and blast load support can shed light find the thickness of the first spall h as:
on each other, as currently both subjects are far from being fully
ct 2 rt
developed. h¼ ð2Þ
2rm

where t2 is the time of the descending part of the wave. Note that
2. 1D spalling
the rise time of the wave does not affect results. The PPV of tensile
waves are opposed to the direction of their propagation and there-
2.1. Analytical equations
fore the velocity of the first spall is the PPV contribution from both
the incident wave and the reflected wave. The PPV is equal to the
A comprehensive discussion on tunnel dynamic support design
product of stress and the materials impedance qc. Using this rela-
based on the findings of the ERA tests can be found in the COE
tionship the velocity of the first spall Vo is given by:
manual (EM 1110-345-432, 1962). The spalling damage to the tun-
nels is explained using a 1D simplification. The theory is backed by

ðrm þ rx Þ 2rm Dt
laboratory experiments on steel bars subjected to explosive pulses. Vo ¼ ¼ 1
ð3Þ
Understanding of the spalling phenomenon (Fig. 1) and proper pre-
qc qc t2
diction of the extent and velocity of spalling is essential for assess-
where Dt is the time interval from the time when the peak stress
ment of the load that will act upon the tunnel liner.
reaches the free surface to the time of the first spall.
Ahmed and Ansell (2012) compared simplified 1D and 2D mod-
Subsequent to the detachment of the first spall a new free face
els of rock and shotcrete liner subjected to blast loads and found
is formed and the remainder of the compressive pulse, now with a
that results are comparable. Zhao et al. (2010) proposed a simpli-
peak stress of rx , will reflect from the newly formed free face. The
fied decoupled method of dynamic liner design based on the 1D
second spall will occur if the tensile stress will once again exceed
idealization to estimate the load on the tunnel liner. The same
the strength of the rock. The number of the spalled layers will be
author idealized a support system of liner and rock bolts to behave
given by the integer number smaller than the ratio rm =rt (EM
as a beam on two springs with a span equal to the bolt spacing.
1110-345-432, 1962).
This beam is further simplified to an equivalent Single Degree of
Freedom (SDOF) system. The load from the 1D analysis is subse-
quently used as an input to compute the displacements and stres- 2.2. 1D Numerical models
ses that develop in the liner.
Analytical equations for the first spall thickness and first spall Numerical models of bars subjected to a dynamic 1D load are
velocity were derived by Zhao et al. (2010) based on the geometry set up in order to compare results of spalling to the analytical
equations. Modeling is undertaken using the code ELFEN, a hybrid
Finite-Discrete Element (FDEM) code that incorporates fracture-
mechanics principles to allow for the realistic simulation of brittle
fracture-driven processes (Hamdi et al., 2014). As the spalling phe-
nomenon is a tensile failure process, the Rankine rotating crack is
used as a failure criterion. This failure criterion is based on the con-
cept of Mode 1 fracturing studied in fracture mechanics. Once the
maximum principal stress reaches the tensile strength limit, ten-
sile softening is initiated and the elastic modulus is degraded in
the direction of the major principal stress invariant (Rockfield,
2005).
Four models are set up as shown as illustrated in Fig. 2. The
material properties are listed in Table 1, where only the tensile
strength rt varies while all other parameters are kept constant.
The model geometry consists of a bar in the dimensions of
0.5  5.0 m. The right end of the bar is constrained in both vertical
and horizontal directions, and the horizontal boundaries of the bar
are constrained in the vertical direction to maintain the one-
dimensionality nature of the problem. The results presented here-
after refer to a face load with a triangular shaped curve with a peak
of 10 MPa and a time period of 0.8 ms applied to the left side of the
bar. No damping is used in the models, as the purpose of these
models is to compare results to the analytical equations, which
do not account for wave attenuation. A preliminary numerical
Fig. 1. Spalling process due to a compressive pulse reflecting from a free face. rm is
investigation conducted by the authors found that the effect of
the peak stress of the wave and rx is the compressive stress of the incident wave at damping on spall thickness and velocity can be neglected for these
the point of the first spall. high frequencies and distances.

Please cite this article in press as: Mitelman, A., Elmo, D. Analysis of tunnel support design to withstand spalling induced by blasting. Tunnel. Underg. Space
Technol. (2015), http://dx.doi.org/10.1016/j.tust.2015.10.006
 A. Mitelman, D. Elmo / Tunnelling and Underground Space Technology xxx (2015) xxx–xxx 3

Fig. 2. Schematic presentation of 1D model.

Table 1
Properties for the 1D bar models in ELFEN.

Model #
1 2 3 4
rt (MPa) 2 4 6 8
rm (MPa) 10
E (GPa) 15.1
q (kg/m3) 2300
c (m/s) 2562
t2 (ms) 0.7
Gf (J/m2) 22

2.3. Analysis of results and discussion for the 1D models

Spalling occurred in all four models as shown in Fig. 3. The

thickness and velocity of the spalled pieces are measured from
ELFEN and compared to thickness and velocity obtained from
Eqs. (2) and (3), and presented in Fig. 4. It can be seen that the
modeled results for spall thickness are in good agreement with
the analytical equations. Results of the first spall velocity are rela-
tively lower in ELFEN, but follow the analytical trend.
The difference in spall velocity may be attributed to mesh
effects. In ELFEN, fracturing cannot be simulated with a structured
mesh. The fracture is not formed all at once as a clear cut through
the material. Separate cracks initiate and coalesce to cause full
detachment. Some portion of the wave continues to travel to the
intact bar before the full detachment of the first spall occurs. In
turn, there is a lower amount of trapped energy in the detached
Fig. 4. Graphs of analytical equations with ELFEN models for (a) spall thickness to
fragment and this causes spall velocities to be lower.
tensile strength and (b) spall velocity to tensile strength.
It is unclear if the analytical equations based on tests on steel
bars are a better representative of the true behavior of a rock sam-
ple. It can be argued that irregularity of the mesh serves as a good size distribution, pre-existing micro-cracks, and strength aniso-
representation of reality as rocks tend to exhibit a great deal of tropy. The authors are unaware of publications discussing labora-
heterogeneity at both the laboratory and field scales. It is assumed tory tests on rock specimens which could help better explaining
that formation of a clear cut through rock will seldom occur, as the the observed difference in spall velocity.
fracturing process is affected by the randomness of the rock grain When comparing both the analytical and the ELFEN 1D models
to results recorded in the ERA tests two significant differences are
found. Firstly, when comparing the first spall velocities to the PPV
measured at the tunnel free face, in the field tests first spall veloc-
ities were significantly higher than the PPV whereas in the 1D ana-
lytical equation and models the first spall velocity is slightly lower
than the PPV at the tunnel free face. Second, in the field tests the
size of subsequent spalled layers decreases whereas in the 1D the-
ory and models the first spall has the maximum thickness.

3. 2D tunnel spalling analysis

The suitability of the 1D simplification is questioned as blast

waves propagate spherically. Moreover, tunnel roofs are typically
arch shaped rendering the 1D approximation even less suitable.
Fig. 3. Spalling results of Models 1–4. To investigate the influence of the problem geometry on spalling,

Please cite this article in press as: Mitelman, A., Elmo, D. Analysis of tunnel support design to withstand spalling induced by blasting. Tunnel. Underg. Space
Technol. (2015), http://dx.doi.org/10.1016/j.tust.2015.10.006
4 A. Mitelman, D. Elmo / Tunnelling and Underground Space Technology xxx (2015) xxx–xxx

2D models are set up with input parameters identical to those of slightly, and that the stress to strength ratio dominates the tunnel
the 1D models listed in Table 1. response.
The model geometry is shown in Fig. 5. As there is some atten- Tracking the fracturing evolution for the models showed that
uation of the wave due to geometrical spreading, the load assigned the first crack initiation occurs at a distance above the center of
to act upon the arc was 16 MPa so that it would arrive at approx- the tunnel equal to the solutions derived from the 1D theoretical
imately 10 MPa at the tunnel roof, so rm is the same as in the 1D Eq. (2). However, as time progresses, different points of the wave
models. A mesh element size of 12 cm was used in the models arrive at the tunnel roof and reflect at different angles, and interact
based on the results of a preliminary mesh sensitivity analysis. with each other and with the newly formed free faces of the cracks.
Results of fracturing of the 2D models are presented in Fig. 6. Fracturing propagates to the sides causing the spalling to be in the
For this case, only in Model #1 fracturing causes full detachment radial direction. Another important phenomenon is that for the
of pieces of material and spalling. According to these results, for weaker rock fracturing propagates downwards from the first point
the given high frequency load, spalling will occur only when the of crack initiation. Thus, the first spalls are smaller, and have a lar-
ratio of the stress of the wave to the strength of the rock rm =rt ger velocity. The velocity of the first spalls is found to be 3.7 m/s,
is as large as 4–5. Additional modeling conducted to examine the and exceeds the PPV of 2.6 m/s recorded at the tunnel boundary
effect of frequency, varying the frequency from 1200 Hz to as shown in Fig. 7. This is similar to results encountered in the
100 Hz found that decreasing frequency increases damage only ERA tests, and contrary to the 1D bar problem.

Fig. 7. Spall vertical velocity above tunnel roof at t = 7 ms from Model 1

Fig. 5. 2D model geometry. (rt ¼ 2 MPa).

Fig. 6. 2D model results at t = 7 ms for varying tensile strengths.

Please cite this article in press as: Mitelman, A., Elmo, D. Analysis of tunnel support design to withstand spalling induced by blasting. Tunnel. Underg. Space
Technol. (2015), http://dx.doi.org/10.1016/j.tust.2015.10.006
 A. Mitelman, D. Elmo / Tunnelling and Underground Space Technology xxx (2015) xxx–xxx 5

Models with a tensile strength greater than 4 MPa show that concrete, the support considered for this paper is a homogenous
spalling does not occur, and fracturing is limited to an area above fiber reinforced concrete liner.
the tunnel crown. However, wave attenuation is greater for points As demonstrated in Section 3, for the case of light damage
progressively distant along the tunnel boundary and therefore the cracking occurs at a distance above the tunnel roof and does not
amplitude of the arriving stress wave is not sufficient to cause propagate downwards, and only small amounts of rock fall with lit-
cracking as it is less than the rock tensile strength. tle to no velocity. For this case, a simpler static analysis can be
In the ERA field tests, light damage zones were characterized by undertaken where the quality of rock above the tunnel is degraded,
detachment of rock occurring only due to pre-existing joints and at and the liner response investigated. For the case of heavier damage,
low velocities of 0–1 m/s. To demonstrate this observed physical the momentum of the series of impacts induced by the spalls
behavior, Model 4 (which is the model with the highest strength) imposes significantly greater stresses on the liner than the dead
was modified to include 3 vertical joints. As shown in Fig. 8, the loads. Hence, the cases of light damage and intensive damage will
presence of such simple fracture geometry is already sufficient to be discussed separately.
cause the detachment of rock blocks from the tunnel roof. Numerical investigations conducted by Deng et al. (2014) show
In general terms, results of the 2D models are in better agree- that the initial stress state induced by gravity on blast induced
ment with observations from the ERA field tests than the simplified damage to tunnels in depths of up to 200 m is insignificant.
1D models. In particular, the latter, despite their appealing simplic- Accordingly, for this work the effects of initial stress state were
ity and relatively short run times cannot account for geometrical not directly accounted for in the models.
2D effects on tunnel spalling; the 1D simplification would likely
provide misleading results in terms of spall size, velocity, and 4.1. Modeling intensive blast induced damage
occurrence.
Using a two-staged decoupled approach where the magnitude
of spalling is initially assessed in a separate stage and the corre-
4. 2D supported tunnel spalling analysis
sponding forces subsequently imposed on the liner structure, could
lead to highly erroneous conclusions. Once the wave reaches the
In a previous paper, Mitelman and Elmo (2014) validated a
rock-liner interface transmission and reflection of the of the inci-
numerical FDEM approach to study tunnels subjected to blasting.
dent wave stress rI depends on the acoustic impedance q1 c1 and
The DEM capabilities in ELFEN can be further utilized to evaluate
q2 c2 of the two materials. The equations for the transmitted and
performance of different support systems by modeling their
reflected stress are:
impact with the spalls. The penalty method is used to enforce
the impenetrability constraint and to calculate the normal and tan- 2rI ðq2 c2 Þ
rT ¼ ð4Þ
gential stresses between interacting discrete bodies (Rockfield, q2 c2 þ q1 c1
2005). In order to estimate the degree of support required to pre-
vent spalling damage, the 2D models presented in the previous rI ðq2 c2
q1 c1 Þ
section can be modified to include a support system. The degree rR ¼ ð5Þ
q2 c2 þ q1 c1
of spalling is a direct function of the magnitude of PPV close to
the tunnel. For assessment of an actual case, field tests or empirical Concrete generally has similar impedance values to rock. Conse-
formulae can be used to estimate the magnitude of PPV. quently, the wave traveling from the rock to the concrete will be
Rock bolts and/or shotcrete liners are widely used as support transmitted almost fully, and the reflected wave will be of negligi-
systems for ensuring the stability of underground excavations. In ble magnitude. Thus, fracturing could propagate into the concrete
order to prevent spalling into the tunnel, a support system consist- liner and cause failure. Counter intuitively, applying a high
ing of bolts alone could not be effective; even when considering a strength concrete would result in an increased transmission into
closely spaced support system, fracturing and spalling could occur the concrete which could exacerbate the liner response.
between bolts. A combination of shotcrete and bolts is arguably a Possibly, a densely spaced steel mesh in the concrete liner could
more efficient means of support. Yet proper modeling of the bolts reduce or even prevent spalling of the concrete. Another apparent
would have to consider not only the bolts themselves, but also solution is to separate the liner from the rock by means of an inter-
account for a complicated 3D dynamic interaction between the mediate isolating layer made of a material with low impedance,
bolts, the grout, the rock, and the liner. such as Styrofoam. The isolating layer would reflect most of the
The addition of wired mesh or fiber reinforcement to the tunnel energy and allow the liner to resist potential impacts while it is
liner increases the liner’s bending capacity and imparts ductility to intact. The drawback of this solution is that the liner would only
an otherwise brittle material. In order to avoid the complexities serve as a dynamic support, and an initial and separate lining
involved with the simulation of bolts and/or wire mesh reinforced would have to be installed to accommodate for the static loading.

Fig. 8. (a) Modified Model 4 with the addition of joints and (b) the resultant spalling of the tunnel roof.

Please cite this article in press as: Mitelman, A., Elmo, D. Analysis of tunnel support design to withstand spalling induced by blasting. Tunnel. Underg. Space
Technol. (2015), http://dx.doi.org/10.1016/j.tust.2015.10.006
6 A. Mitelman, D. Elmo / Tunnelling and Underground Space Technology xxx (2015) xxx–xxx

Table 2
Properties of the fiber reinforced concrete.

Properties of the fiber reinforced concrete

3
Density (kg/m ) 2500
E (GPa) 35
Shear modulus (GPa) 13.4
Poisson ratio 0.2
Tensile strength (MPa) 6.5
Gf (J/m2) 50
Cohesion (MPa) 5
Friction angle 52

The limiting factor for the thickness of the isolation layer should be
dictated by the tunnel’s anticipated static displacement.
To compare the performance of a tunnel liner with and without
Fig. 10. Proposed simplification of spalling impacts on liner to an equivalent single
an isolation layer, two models are considered in this paper. The block problem.
models are identical with the exception of the rock-liner boundary
condition. For the first model, the liner is assumed to be bonded to
the rock, and the adhesive strength for the bond between the liner
and rock is assumed to be 1 MPa (Ahmed and Ansell, 2012). For the
second model, a gap of 3 cm is left between the rock and the liner.
The properties of the fiber reinforced concrete are listed in
Table 2. The failure criterion for the concrete liner is a combined
Mohr–Coulomb and Rankine model to account for both tensile
and shear failure. The properties of the rock are the same as listed
in Table 1 for Model 4. The ratio of the peak stress of the blast wave
to the tensile strength is assigned a high value of 5 to allow inten-
sive spalling to occur. The model geometry is identical to the 2D
analyses presented in Section 3, with only the addition of the liner.
The liner is shaped as a semicircle and constrained at its bottom.
No damping is assigned to the liner as in impact problems damping
has a negligible influence (Chopra, 2011). The liner thickness is
gradually increased by increments of 5 cm until failure does not
occur.

4.2. Analysis of results and discussion

For the model with the liner bonded to the rock, under low
thickness values of 20–25 cm the fracturing due to the wave reflec-
tion is limited to the rock, and the liner fails as a result of spalling
impact. As the thickness is increased fracturing propagates from Fig. 11. Phase2 Model for light damage assessment.
the rock downward to the liner, and failure is combined of both
inner spalling of the concrete and rock impact.
Simulated results of fracturing are shown in Fig. 9. The model- detailed investigation may lead to selection of more efficient sup-
ing results demonstrate the ineffectiveness of the bonded fiber port systems. A 3D analysis could account for the lateral resistance
reinforced liner to resist intensive blast damage. When the rock of the tunnel liner, and would be a better means of modeling addi-
is stronger than the liner, the liner is not only ineffective, but even tional support systems such as bolts, mesh reinforced concrete or
poses a greater spalling hazard than when compared to no support. steel sets embedded in concrete. However, a full 3D simulation
However, even when isolation is applied, a considerable thick- of spalling behavior with explicit fracturing would be computa-
ness of 65 cm is required to prevent damage. Possibly, a more tionally expensive.

Fig. 9. Simulated results of liners with 65 cm thickness for (a) model with gap between liner and rock (b) model with adhesive bond between liner and rock.

Please cite this article in press as: Mitelman, A., Elmo, D. Analysis of tunnel support design to withstand spalling induced by blasting. Tunnel. Underg. Space
Technol. (2015), http://dx.doi.org/10.1016/j.tust.2015.10.006
 A. Mitelman, D. Elmo / Tunnelling and Underground Space Technology xxx (2015) xxx–xxx 7

Table 3 where k is a reduction factor, n is the number of spalls and v eq the

Input parameters for static analysis. equivalent velocity. The correction factor is applied because the
Unit weight (MN/m3) GSI Young’s modulus (GPa) energy of the equivalent system is less than the energy of the orig-
Rock 0.027 70 14 inal system. The approach is explained considering:
Degraded rock 0.027 10 0.6 1. The first layer of spalls that carries the greatest velocity will
Liner 0.025 – 35 impact the liner first. Thus, the subsequent layers will impact
on the first spalls when they are traveling at some speed and
contribute less energy to the system.
2. Spalling from walls with horizontal components cause vibration
When the liner is isolated from the rock a decoupled approach modes different than spalls from the roof with vertical compo-
where the liner is modeled as a standalone structure is acceptable. nents. The superposition of the modes may cancel each other out.
A simplification of the problem from a series of small impacts to an 3. Some of the energy is dissipated by inner collisions between the
impact of an equivalent single block (Fig. 10) could be an efficient rock pieces.
means of reducing the complexity of the problem. The equivalent
velocity can be assessed by: However, finding the ranges for suggested values of the reduc-
X
n tion factor k would require additional and extensive investigations
mtotal v 2eq ¼ k mi v 2i ð6Þ that are beyond the scope of the current paper.
i¼1

Fig. 12. Capacity diagrams for a 15 and 25 cm liner.

Fig. 13. Flowchart for design of support for tunnel under blasting.

Please cite this article in press as: Mitelman, A., Elmo, D. Analysis of tunnel support design to withstand spalling induced by blasting. Tunnel. Underg. Space
Technol. (2015), http://dx.doi.org/10.1016/j.tust.2015.10.006
8 A. Mitelman, D. Elmo / Tunnelling and Underground Space Technology xxx (2015) xxx–xxx

4.3. Modeling light blast induced damage the two modeling approaches and that 2D modeling is better at
simulating the damage observed in physical tests, in terms of
In the ERA tests light damage to the tunnel was found to occur velocity, size, and occurrence of spalling.
under PPVs of 1–2 m/s. In these tests the area of damaged rock rel- A distinction was made between light and heavy blast loads,
ative to the tunnel area was for tunnels of 10 m diameter was where only under heavy blast loading fracturing propagates down-
approximately 1% and the broken pieces of rock were observed wards from the initial crack formed above the tunnel. The implica-
to have a velocity of 0–1 m/s. It is therefore proposed that for these tion for heavy loads is that isolating the liner from the rock is
cases a simple static analysis would be a sufficient means of sup- necessary. Simulated results showed that slender liners bonded
port design. Low velocity spalls that may eject due to pre- to the rock failed due to their lack of capacity to withstand the
existing joints can be accounted for by some increase of the static dynamic impacts of the spalls, whereas for thick bonded liners
load. Compared to heavy spalling, the stresses that will develop in the spalling process progressed into the liner itself rendering such
the liner under small PPVs are smaller and therefore the initial a support as ineffective and even hazardous.
state of stresses in the liner has a considerable effect on failure. Modeling of a fiber reinforced concrete liner showed that very
As an example, an analysis is carried out using the FEM code thick liners are required to withstand heavy spalling. Alterative
Phase2 (RocScience, 2011), as shown in Fig. 11. A 10 m diameter support systems should be considered using 3D models. A prelim-
tunnel is supported by liner elements. An area above the tunnel inary discussion was presented, proposing a simplified method of
roof which corresponds to 2% of the tunnel area is assigned extre- an equivalent block impacting the support structure in order to
mely weak strength properties to account for the 1% of rock falls avoid computationally intensive modeling.
measured in the ERA tests and additional movements and fractures For light loads which are defined by a low stress to tensile
that may occur from the blast load. On its lateral boundaries, the strength ratio, a simplified analysis method was proposed using
degraded material is bound by joints, to enable increased vertical static FEM analysis.
displacements. A load of 10 kN is assigned to account for additional
dynamic loads from the blast vibration. For a given mass, impact References
loading can result in stresses and deformations 10 times and even
more, than the static load imposed by gravity (Malmgren and Ahmed, L., Ansell, A., 2012. Structural dynamic and stress wave models for analysis
of shotcrete on rock exposed to blasting. Eng. Struct. 35 (1), 11–17.
Nordlund, 2006). The rock and liner properties are listed in Table 3, Chopra, A., 2011. Dynamics of structures, fourth ed. Prentice-Hall International
where the liner is assumed to be reinforced with a 30  30 cm steel Series in Civil Engineering and Engineering Mechanics.
wire mesh with a diameter of 6 mm. Deng, X.F., Zhu, J.B., Chen, S.G., Zhao, Z.Y., Zhou, Y.X., Zhao, J., 2014. Numerical study
on tunnel damage subject to blast-induced shock wave in jointed rock masses.
The problem is staged as follows: Tunnel. Underground Space Technol. 43, 88–100.
Donze, F.V., Bouchez, J., Magnier, S.A., 1997. Modeling fractures in rock blasting. Int.
1. Initial stage to allow stress distribution of in-situ stresses. J. Rock Mech. Min. Sci. 34, 1153–1163.
EM 1110-345-432 US Corps of Engineers Manual. 1962. Engineering and Design of
2. Softening of the inclusion to account for deformations that Underground Installations in Rock-Tunnels and Linings. Armed Services
occur prior to support installation, according to the core Technical Information Agency, Arlington, Virginia.
replacement technique proposed by Hoek et al. (2008). Engineering Research Associates, 1953. Underground Explosion test Program Final
Report, vol. 2. Sandstone Rock, Sacramento District.
3. Installation of liner.
Grady, D.E., Kipp, M.E., 1980. Continuum modeling of explosive fracture in oil shale.
4. Degradation of the rock material above the tunnel, activation of Int. J. Rock Mech. Min. Sci. 17, 147–157.
joints at its boundaries, and activation of vertical load. Hamdi, P., Stead, D., Elmo, D., 2014. Damage characterization during laboratory
strength testing: a 3D-finite-discrete element approach. Comput. Geotech. 60
(7), 33–46, 2014.
Support capacity diagrams of bending moment vs. thrust are Hoek, E., Carranza-Torres, C., Diederichs, M.S., Corkum, B., 2008. Integration of
presented in Fig. 12. It can be seen that for the 15 cm liner at geotechnical and structural design in tunnelling – 2008 Kersten Lecture. In:
one point the FoS was less than one, and 2 points less than 1.2. Proceedings University of Minnesota 56th Annual Geotechnical Engineering
Conference. Minneapolis, 29 February 2008, pp. 1–53.
Due to an increase of liner thickness to 25 cm all points of the liner Kaiser, P.K., Cai, M., 2012. Design of rock support system under rockburst condition.
possess a FoS of greater than 1.2. For both cases, shear forces in the J. Rock Mech. Geotech. Eng., 215–227
liner were all above a FoS of 1.2. Phase2 analyses can be set up and Kendorski, F.S., Jude, C.V., Duncan, W.M., 1973. Effect of blasting on shotcrete lined
tunnels. Min. Eng. 25 (12), 38–41.
run at very short times, enabling optimization and/or evaluation of Malmgren, L., Nordlund, E., 2006. Behavior of shotcrete supported rock wedges
an array of varying conditions. subjected to blast-induced vibrations. Int. J. Rock Mech. Min. Sci. 43, 593–615.
Mitelman, A., Elmo, D., 2014. Modelling of blast-induced damage in tunnels using a
hybrid finite-discrete numerical approach. J. Rock Mech. Geotech. Eng. 6, 565–
5. Conclusions: toward the development of a tunnel blast 573.
support design Rockfield, 2005. Rockfield Software Ltd., Technium, Kings Road, Prince of Wales
Dock, Swansea, SA1 8PH, UK.
Rocscience Inc., 2011, Phase2 Version 8.0 – Finite Element Analysis for Excavations
By using a toolbox approach to modeling liner response to rock and Slopes, Toronto, Ontario, Canada, <www.rocscience.com>.
mass damage due to blasting, the authors propose a new approach Wood, D.F., Tannant, D.D., 1994. Blast damage to steel fiber reinforced shotcrete. In:
to support designs to account for blast induced damage. The mod- Fiber Reinforced Concrete – Modern Developments. UBC Press, pp. 241–250.
Wu, Z., Chen, G., Zen, K., Kasama, K., Wang, D., 2011. Effect of blasting on the
eling procedure is summarized in Fig. 13. adjacent underground tunnels. Appl. Mech. Mater. 90–93 (2011), 1870–1878.
This paper has initially shown that numerical modeling using a Zhao, P.J., Lok, T., Yin, Z., Zhou, Z., 2010. Simplified design of rock cavern concrete
hybrid FDEM code to model spalling behavior in 1D produces lining to resist shock loading. J. Central South Univ. Technol. 17 (5), 1087–1094.
Zhu, Z., Xie, H., Mohanty, B., 2008. Numerical investigation of blasting-induced
results in agreement with analytical solutions. However, compar- damage in cylindrical rocks. Int. J. Rock Mech. Min. Sci. 45 (2), 111–121.
ison of spalling behavior in 1D and 2D models with respect to dam-
age to tunnels shows that there are significant differences between

Please cite this article in press as: Mitelman, A., Elmo, D. Analysis of tunnel support design to withstand spalling induced by blasting. Tunnel. Underg. Space
Technol. (2015), http://dx.doi.org/10.1016/j.tust.2015.10.006
View publication stats