An upper bound solution for tunnel face stability analysis considering the

free span
⁎
Salvador Senent, Congke Yi, Rafael Jimenez
Technical University of Madrid, Spain

A B S T R A C T

Keywords: One of the main factors that affect the tunnel face stability is the free or unsupported span (i.e., the unlined
Limit-analysis length adjacent to the tunnel face). In this work, a 2D failure mechanism is developed to compute, in the context
Face stability of Limit Analysis, the collapse pressure for tunnel faces in Mohr-Coulomb materials, considering the free span.
Forepole umbrella The new mechanism comprises three non-deformable blocks (two translational and one rotational) in such a way
Free-span
that the interface between one of the translational blocks and the rotational one is constructed by an iterative
process that fulfills the associated flow rule imposed by Limit Analysis. A numerical model is employed to
validate the proposed mechanism, demonstrating that it provides good predictions of the critical pressure and of
the failure geometry, especially for tunnels in materials with higher friction angles (i.e., φ ≥ 35° ). Results confirm
the expected increase of the collapse pressure with the free span; however, the increase is noticeably higher
when no pressure is applied on the free span than when the support pressure is applied on the tunnel face and on
the free span, especially for cases with lower cohesion (i.e., c ≤ 15 kPa ). In addition, the effect of a forepole
umbrella is incorporated into the Limit Analysis mechanism, and a new methodology is proposed to estimate the
collapse pressure when a forepole umbrella is installed at the face. Results of the analytical solution in this
situation are compared with those of the numerical model, with results confirming that the proposed mechanism
is able to capture the stability improvement (i.e., the reduction of the collapse pressure) produced by the um-
brella, as well as the shape changes of the failure mechanism; in particular, it can predict a local failure of the
face below heavy umbrellas.

1. Introduction (1981), who carried out tests in cohesive materials, two different in-
stability mechanisms can occur depending on the free span length. For
Face stability causes great concern during tunneling construction lower L/ D ratios, a mechanism that mainly affects the ground ahead of
since serious consequences and great losses may occur due to a face the tunnel face develops; for larger L/ D ratios, the failure develops
failure. The free or unsupported span (i.e., the unlined length adjacent above the free span producing an almost 2D failure. (Similar results
to the tunnel face) is a relevant factor in tunnel construction, as it in- were obtained in frictional materials by, for example, Chambon and
fluences the deformability of the excavation (e.g., Hayati et al., 2013) Corte, 1994). Lee and Schubert (2008) studied the stability of the
and consequently its impact on near structures (e.g., van Tol, 2006); but tunnel face considering the free span and other parameters such as the
the free span is also a crucial factor on tunnel face stability, as shown by overburden and the lining stiffness. Their laboratory test results also
numerical and experimental studies (e.g., Vermeer et al., 2002; Casarin reveal an increase of the critical or collapse pressure –i.e., the minimum
and Mair, 1981; Chambon and Corte, 1994) and by real-life experiences pressure that needs to be applied at the tunnel face to avoid its in-
(HSE, 1996). However, most researchers focusing on the stability of the stability– with the free span.
tunnel face have considered a completed unlined tunnel (i.e., an infinite These observations agree with the results of numerical simulations.
free span; see e.g., Davis et al., 1980), or a completely lined tunnel (i.e., Different three-dimensional models using the Finite Element Method
with a null free span; see e.g., Leca and Dormieux, 1990). (e.g., Vermeer et al., 2002), the Finite Difference Method (e.g., Costa
Several experimental studies have employed the ratio between the et al., 2007) or the Discrete Element Method (e.g., Lee and Schubert,
length of the free span and the tunnel diameter (L/ D ) to assess the effect 2008) have been employed to (i) analyze the failure mechanism and (ii)
of the free span on tunnel face stability. According to Casarin and Mair to determine the critical pressure of tunnel faces considering the free

⁎
Corresponding author at: ETSI Caminos, C. y P., c/ Profesor Aranguren s/n, 28040 Madrid, Spain.
E-mail address: [email protected] (R. Jimenez).
 Nomenclature Si area of Block i (i = 1, 2 and 3)
ω2 angular velocity around O2 of the Block 2
Symbols V1 velocity of Block 1
V3 velocity of Block 3
H excavation height V13 relative velocity between Block 1 and Block 3
C overburden V2, i ̇ velocity of Block 2 in Point Pi
L unsupported length V23, i ̇ relative velocity between Block 2 and Block 3 in Point Pi
c soil cohesion Wγg work rate of the soil weight
φ friction angle WTg work rate of the support force
γ unit weight W Dg work rate of the internal energy dissipation
σt support pressure d diameter of the structural element of the forepole um-
σc collapse pressure brella
α1 angle in Point A of the interface AF with the horizontal di interior diameter of the structural element of the forepole
α2 angle of the velocity of Block 3 with the horizontal umbrella
β1 angle of the interface AD with the horizontal As cross-sectional area of the structural element of the fore-
RA , βA polar coordinates of Point A pole umbrella
RB , βB polar coordinates of Point B α angle of the cross-section of the structural element with
RF , βF polar coordinates of Point F the failure surface
RE , βE polar coordinates of Point E η angle of the forepole umbrella with the horizontal
RK , βK polar coordinates of the barycenter of Block 2 μ non-dimensional parameter
ZD, YD Cartesian coordinates of Point D δ thickness of the continuous deformation zone
ZG, YG Cartesian coordinates of Point G δBS δ suggested by Buhan an Salençon (1993) (Eq. (27))
ZO2, YO2 Cartesian coordinates of center O2 δNM δ approximated from the numerical model
Ti,(i = 1... N ) points in the interface AF δNM δ estimated employing Eq. (28)
→
Pi,(i = 1... M ) points in the interface BF VF velocity of the neutral axil of the structural element at the
z i , yi Cartesian coordinate of Point Pi velocity jump surface
s interval of adjacent points in the interface AF fy yield strength of the structural element
→
ui ̇ unit vector of the interface AF in Point Pi N0 ultimate axial force of the structural element
lCD length of the interface CD V0 ultimate shear force of the structural element
lAD length of the interface AD M0 ultimate bending moment of the structural element
lDG length of the interface DG WFg work rate of the forepole umbrella
lFG length of the interface FG

span. Vermeer et al. (2002) studied the stability of tunnel faces in co- arch umbrella is one of the most common support systems. The um-
hesionless materials. They showed that the stability of the face is almost brella helps to support the weight of the material above the free span,
independent of the free span for lower L/ D ratios (i.e., L/ D < 0.3); hence avoiding the development of a progressive failure that could
however, they pointed out that these results depend on the geometry of reach the surface. Many authors have studied the performance of
the tunnel, as in less circular tunnels the influence of the free span is forepole umbrellas to improve the stability of the tunnel face (e.g., Le
higher. Costa et al. (2007) recognized the same two mechanisms as- and Taylor, 2018). However, there are not many works about forepole
sociated to the free span reported by Casarin and Mair (1981), and they umbrellas in the Limit Analysis literature, as the mechanisms employed
pointed out that the transition between both mechanisms depends on have mainly considered a nail reinforcement –a support that only works
drainage, and that the effect of the free span on the collapse pressure is by axial force– (e.g., Pan and Dias, 2017). Pinyol and Alonso (2011,
more relevant under undrained conditions. 2013) studied the effect of a forepole umbrella on the stability of the
Some analytical approaches have been proposed to study the in- tunnel face in undrained soils, employing one of the translational me-
fluence of the free span. Tamez et al. (1997) proposed a Limit Equili- chanisms proposed by Leca and Dormieux (1990), and Qian et al.
brium failure mechanism, similar to the classical wedge mechanism (2019) combined the strength reduction method and an advance rota-
proposed by Horn (1961), but considering a prismatic block above the tional mechanism to compute safety factors of non-circular tunnel faces;
free span. Employing a similar mechanism, and also in the framework however, these works did not consider the influence of the free span.
of Limit Equilibrium, Anagnostou and Perazzelli (2013) studied the In this work, the mechanism proposed by Senent and Jimenez
effect of the free span, and of the support pressure distribution, on the (2017) is improved to generate, in the framework of Limit Analysis, a
stability of the tunnel face. Solutions to consider the free span in the tunnel face failure mechanism that considers the influence of the free
framework of Limit Analysis are rare: Senent and Jimenez (2017) span; the main advantages of the new mechanism are that it is less
presented a two-dimensional failure mechanism combining rotational complex, while providing results that are more similar to numerical
and translational non-deformable blocks, but they did not discuss the simulations and to experimental tests. Moreover, the formulation of the
details of the mechanism and their solution shows some differences, analytical solution is presented in detail, and its results are compared
both in the collapse pressure and in the failure geometry, with the re- with a numerical model that studies the influence of the free span on
sults of a numerical model. Similarly, Zhang et al. (2018) proposed a tunnel face stability. In addition, the effect of a forepole umbrella is
two-dimensional Limit Analysis failure mechanism, composed of a tri- incorporated into the proposed Limit Analysis mechanism, comparing
angular block above the tunnel crown and a spiral block ahead of the the results of the analytical solution with those of the numerical model,
tunnel face; their results showed that the collapse pressures given by analyzing the stability improvement provided by the forepole umbrella,
their Limit Analysis mechanism are clearly lower than those obtained and proposing a methodology to compute the collapse pressure in this
with a Limit Equilibrium approach that was also proposed by these situation.
authors.
To improve the stability of the tunnel face, a forepole umbrella or
2. An upper bound solution in the framework of Limit Analysis 2–. Since the location of O2 is defined by parameters βE y RE , its
Cartesian coordinates are:
2.1. Model description
H
YO2 = RE ·cos(βE ) −
2 (1)
A tunnel excavation of height H under a cover depth C in a Mohr-
Coulomb material, defined by its cohesion (c ), friction angle (φ ) and
unit weight (γ ), is considered. Note, therefore, that the analysis is two- ZO2 = −RE ·sin(βE ) (2)
dimensional, corresponding to a longitudinal section of the excavation;
The proposed free span mechanism comprises three blocks. Hence,
but interesting outcomes can be obtained from this type of analyses
it is necessary to define the relative velocities between blocks, in such a
(e.g., Mollon et al., 2011a), and it could be particularly useful for other
way that all terms in the energy balance equations are expressed as a
underground works, such as long-wall mining (e.g., Yan et al., 2018).
function of a unique velocity that could be later removed from the
The length of unsupported excavation (or free span) is L and a uniform
equations. The relative velocity between Blocks 1 and 3 (V13 ) must form
support pressure σt is applied on both the tunnel face and on the free
an angle with their interface (line AD) equal to the friction angle. Since
span. Nonetheless, the case with an unsupported free span is also
the interface orientation is defined by the angle β1, velocities V3 and V13
considered in the discussion.
can be calculated with the hodograph (Fig. 2(a)) employing the fol-
lowing equations:
2.2. Collapse mechanism
π
sin( 2 − β1 − φ)
Fig. 1 shows the proposed Limit Analysis mechanism to study the V3 = ·V1
stability of the tunnel face considering the free span. It is formed by sin(α2 + β1 + φ) (3)
three non-deformable blocks: (i) a translational triangular block above
the free span (Block 1); a rotational block ahead of the tunnel face cosα2
V13 = π ·V3
(Block 2); and a translational block above them that closes the me- sin( 2 − β1 − φ) (4)
chanism (Block 3).
The failure of the material above the free span is analogous to the Similarly, the angular velocity of Block 2 (ω2 ) can be obtained im-
failure mechanism in the trapdoor problem (Terzaghi, 1936), so that posing an associated flow rule in the initial point of the interface be-
Block 1 is similar to the mechanism proposed by Evans (1984) for this tween Blocks 2 and 3 (Point A). Since, at Point A, the interface forms an
experiment. In this work, because of the boundary conditions –i.e., the angle α1 with the horizontal, and the velocity of Block 2 in Point A is
tunnel face–, an asymmetrical triangular block (Block 1), with a verti- perpendicular to O2¯ A, the angular velocity of Block 2 can be expressed
cally descending velocity (V1), is proposed. Its external boundary DC as (Fig. 2(b)):
forms an angle φ with the vertical to fulfill the Limit Analysis hypoth-
esis of associated flow. Another parameter, β1, is introduced to define its
internal boundary with Block 3. This simplifies the mechanism origin-
ally proposed by Senent and Jimenez (2017), in which a symmetrical
triangular block was used, hence requiring one more block on its side.
The proposed asymmetrical block eliminates the need for such addi-
tional block, so that the mechanism becomes simpler, and its results are
more similar to numerical and experimental results (see e.g., Costa
et al., 2007; Lee and Schubert, 2008). Block 2 suffers a rotational
movement, with angular velocity ω2 , around a center O2 whose position
is defined by two variables (βE , RE ). Its geometry is similar to the 2D
rotational mechanism proposed by Mollon et al. (2011a), which has
been shown to outperform other translational mechanisms, both in the
computed collapse pressure and in the failure geometry. In this case, the
lower boundary of Block 2 (curve BF) is a logarithmic spiral of para-
meter φ that starts in the lower corner of the excavation (Point B); its
upper boundary (curve AF) is generated by an iterative process, as
explained later, considering the velocity of Block 3. The position of this
internal boundary between Blocks 2 and 3 is determined by a new
parameter (α1) that defines its inclination in Point A. Finally, the
translational Block 3 is characterized by a velocity (V3 ), towards the
tunnel face, that forms an angle α2 with the horizontal. The angle in the
upper corner of Block 3 (point G) is 2·φ to fulfill the associated flow
requirement in the external boundary of the mechanism.
Considering the previous explanation, the mechanism is defined by
5 parameters (β1, βE , RE , α1, α2 ). Given that the proposed analytical
solution is an upper-bound Limit Analysis solution, and that the face
pressure acts against the movement of the mechanism, the collapse
pressure must be maximized in relation to these variables (Chen, 1975).

2.3. Formulation of the analytical solution

2.3.1. Coordinate systems, velocity field and geometry construction

Two coordinates systems are employed to produce the mechanism:
(i) a Cartesian system ( y , z ) with origin in Point A; and (ii) a polar Fig. 1. Outline of the proposed Limit Analysis mechanism for the stability of the
system (R , β ) with origin in O2 –i.e., in the center of rotation of Block tunnel face considering the free span.
 ZO2
βA = tan−1 (− )
YO2 (7)
The boundaries of Block 2 are not straight lines, and they must be
computed first to obtain Point F (see Fig. 1). The logarithmic spiral
emerging from point B can be expressed in polar coordinates as:
R = RB ·exp((β − βB ) tanφ) (8)
The interface AF is generated point by point starting from Point A
and, discretizing it in segments of length s , following the process shown
in Fig. 2(c). For each point Pi of the interface ( x i , yi in Cartesian co-
→
ordinates), its velocity corresponding to Block 2 is V2, i ̇ and forms an
angle of 90° with O2¯P i ̇ :
→
V2, i ̇ = (yi − YO2, z i − ZO2)·ω2 (9)
Since the relative velocity between Blocks 2 and 3 in Point Pi is
→ → →
V23, i ̇ = V2, i ̇ − V3 , and since this velocity must form in Point Pi an angle
with the interface equal to φ , the unit vector of the next segment (of
→
length s ) of the interface is obtained rotating an angle φ the vector V23, i :̇
→
→ cosφ sinφ ⎤ V23, i ̇
ui ̇ = ⎡ · →
⎢−
⎣ sinφ cosφ ⎥
⎦ | V23, i|̇ (10)
Hence, the next point of the interface (Pi + 1) can be computed as:
→ → →
̇ 2 − ui ·̇ s
P i +̇ 1 O2 = P i O (11)
Finally, Point F (RF , βF in polar coordinates) can be computed as the
intersection of the two previous curves (BF and AF).
The remaining boundaries of the mechanism are straight lines, and
they are easy to obtain. Starting from the coordinates of Point C
(− L, 0 ), Point D can be obtained as the intersection of two straight
lines, the first emerging from Point C, with an inclination of (90 − φ ),
and the second from Point A, with an inclination of β1. Then, the co-
ordinates of Point D are:
−L
ZD =
1 + tanβ1·tanφ (12)

YD = −ZD·tan(β1) (13)
Similarly, the coordinates of Point G can be obtained as the inter-
section of two straight lines emerging (i) from Point D, with and in-
clination of (α2 − φ ), and (ii) from Point F, with an inclination of
(π −φ − α 2 ). Consequently, the coordinates of Point G are:
cot(α2 + φ)·YF + cot(α2 − φ)·YD + ZF −ZD
YG =
cot(α2 + φ) + cot(α2 − φ) (14)

ZG = ZD + cot(α2 − φ)·(YG − YD ) (15)

Once all the internal and external boundaries of the mechanism
have been obtained, it is possible to compute the areas of each block
and the length of each boundary, and to formulate the Limit Analysis
energy balance needed to compute the collapse pressure.

2.3.2. Work equation

The external forces applied on the failure mechanism comprise the
soil weight, and the support force applied on the tunnel face and on the
free span. We assume that the mechanism never outcrops at the surface,
so that the ground surcharge is not considered in the formulation.
Fig. 2. Hodographs: a) interface between Blocks 1 and 3; b) in Point A between
Blocks 2 and 3; c) interface between Blocks 2 and 3. The rate of work of the soil weight is:

Wγg = ∭V →γ ·→v ·dV = γ·(S1·V1 + S2·sin(βK )·RK ·ω2 + S3·sinα2·V3) (16)

sin(α1 + α2 − φ) V3
ω2 = ·
sin(α1 − φ + βA ) RA (5) where Si(i = 1,2,3) is the area of Block i and βK and RK are the polar co-
ordinates of the barycenter of Block 2.
where RA and βA are the polar coordinates of point A: The rate of work of the support pressure is:

RA = ZO22 + YO22 (6) WTg = ∬S →σt ·→v ·dS = −σt·(L·V1 + RE ·ω2·H·cosβE ) (17)
.

Finally, the rate of the internal energy dissipated in the proposed α2 ).

mechanism is due to the plasticity work that occurs along the internal
and external velocity discontinuities: 3. Numerical validation

W Dg In this section, the main results of the proposed mechanism –the

N−1
→ value of the collapse pressure and the failure geometry– are compared
(
= −c·cosφ · lCD·V1 + lAD·V13 + ∑ |Ti T ̇ 1|·R Ti· ω2
̇ i+
to the results of a numerical simulation conducted with the two-di-
1
M−1 → mensional finite element Limit Analysis code OptumG2 (OptumCE,
+ ∑1 |Pi P ̇ 1|· V23, i + lDG·V3 + lFG·V3
̇ i+ ) (18) 2019). (This code implements the finite element formulation of the
limit theorems of classical plasticity, as described in Sloan, 1988, 1989;
where lIJ represent the length between point I and J in the boundary of Lyamin and Sloan, 2002a, 2002b).
the mechanism (see Fig. 1), and Ti,(i = 1... N ) and Pi,(i = 1... M ) are the points in
which curves BF and AF are discretized. 3.1. Test cases
By equating the work rate of the external forces to the rate of in-
ternal dissipated energy, the collapse pressure can be computed from: An excavation of 10 m height with a free span L varying between 0
to 3 m is employed for the validation of the proposed mechanism.
σc
(L = 0 corresponds to a completely lined tunnel, and it is used as
⎡ control case). Table 1 lists the 36 test cases used in the numerical va-
= ⎢γ·(S1·V1 + S2 ·sin(βK )·RK·ω2 + S3 ·sinα2·V3) − c·cosφ ·
⎣ lidation. Two different friction angles (30° and 35° ) and three different
N−1 cohesions (5, 15 and 25 kPa ) are used. All the cases have the same unit
(lCD·V1 + lAD·V13 + ∑ |Ti→ →
M−1
T ̇ 1|·R T · ω2 + ∑
̇ i+ i |Pi P ̇ 1|· V23, i
̇ i+ weight (γ = 20 kN/m3) and the same overburden (C = 5D ), which has
1
1
been chosen to avoid that the failure mechanisms reach the surface.
⎤
)
+ lDG·V3 + lFG·V3 ⎥/(L·V1 + RE·ω2 ·H·cosβE)
3.2. Numerical model
⎦ (19)

As previously pointed out, this expression must be maximized in Fig. 3 illustrates the model used to validate the analytical solution
relation to the five variables that define the mechanism (β1, RE , βE , α1, numerically. The excavation is 10 m high and the cover is 50 m. The

Table 1
Test cases for the numerical validation of the proposed Limit Analysis mechanism.
Case Parameters Collapse Pressure [kPa] δNM [m]

φ [°] c [kPa] Free span L [m] Proposed mechanism σc, LAM Numerical model(1) σc, NM Difference(1) σc, NM − σc, LAM

1 30 5 0 33.2 35.7–37.0 2.5–3.8 1.70

2 0.5 33.6 36.7–38.0 3.1–4.4 1.74
3 1 34.2 37.8–39.2 3.6–5.0 1.83
4 1.5 35.0 39.2–40.6 4.2–5.6 1.86
5 2 36.0 40.6–42.2 4.6–6.2 1.87
6 3 38.7 44.2–46.0 5.5–7.3 1.97
7 15 0 15.8 18.4–19.7 2.6–3.9 1.70
8 0.5 16.5 19.4–20.6 2.9–4.1 1.77
9 1 16.8 20.5–21.9 3.7–5.1 1.82
10 1.5 17.6 21.8–23.3 4.2–5.7 1.90
11 2 18.7 23.3–24.9 4.6–6.2 1.83
12 3 21.3 26.9–28.7 5.6–7.4 2.00
13 25 0 0.0 1.0–2.4 1.0–2.4 1.73
14 0.5 0.0 2.1–3.4 2.1–3.4 1.81
15 1 0.0 3.2–4.6 3.2–4.6 1.84
16 1.5 0.3 4.5–6.0 4.2–5.7 1.87
17 2 1.3 6.0–7.6 4.7–6.3 1.86
18 3 4.0 9.6–11.4 5.6–7.4 1.98
19 35 5 0 24.3 24.9–25.9 0.6–1.6 1.16
20 0.5 24.5 25.4–26.3 0.9–1.8 1.23
21 1 24.8 25.9–26.9 1.1–2.1 1.25
22 1.5 25.2 26.5–27.6 1.3–2.4 1.36
23 2 25.8 27.3–28.5 1.5–2.7 1.36
24 3 27.4 29.3–30.6 1.9–3.2 1.42
25 15 0 10.0 10.6–11.6 0.6–1.6 1.17
26 0.5 10.2 11.1–12.0 0.9–1.8 1.25
27 1 10.5 11.6–12.6 1.1–2.1 1.25
28 1.5 10.9 12.2–13.3 1.3–2.4 1.38
29 2 11.5 13.0–14.2 1.5–2.7 1.40
30 3 13.1 15.0–16.3 1.9–3.2 1.44
31 25 0 0.0 0.0–0.0 0.0–0.0 –
32 0.5 0.0 0.0–0.0 0.0–0.0 –
33 1 0.0 0.0–0.0 0.0–0.0 –
34 1.5 0.0 0.0–0.0 0.0–0.0 –
35 2 0.0 0.0–0.0 0.0–0.0 –
36 3 0.0 0.7–2.0 0.7–2.0 1.38

(1) The left value corresponds to an upper bound analysis (Element type = Upper in OptumG2) and the right value corresponds to a lower bound analysis (Element
type = Lower in OptumG2).
 Fig. 3. Geometry of the numerical model built in OptumG2.

dimensions of the model are 70 m × 80 m, and the length of the model excavation boundary have been fixed except in the free span. (“Limit
ahead of the tunnel face is 50 m. The boundary conditions are given by Analysis” has been selected as Analysis Type in OptumG2. Similarly, the
fixed displacements at the boundaries of the model, i.e., at its lateral Element Type has been configured as “Upper” and “Lower”; since the
perimeter and at its base, as shown in Fig. 3. Similarly, the tunnel face pressure acts against the movement of the mechanism, this makes
support has not been included and displacements at the tunnel OptumG2 to compute, respectively, lower and upper bounds of the

Fig. 4. Comparison of collapse face pressures computed with the proposed mechanism and with the numerical model.
collapse pressure).

3.3. Results

Table 1 and Fig. 4 list the collapse pressures obtained with the
proposed mechanism and with the numerical model. As it can be ob-
served, the results are very similar, in all cases with differences lower
than 7.4 kPa. Note also that the differences are higher for the lower
friction angle (i.e., for φ = 30° ) and that they are almost independent of
the cohesion. For the higher friction angle (φ = 35° ), the differences are
lower than 3.5 kPa comparing to the results of the lower bound ana-
lysis, and lower than 2.0 comparing to the results of the upper bound
analysis.
Similarly, Fig. 5 compares the collapse geometries obtained with the
analytical solution and with the numerical model. (Failure mechanisms
in the OptumG2 model have been estimated considering the distribu- Fig. 6. Collapse pressure vs free span length for different strength properties.
tion of shear dissipation.) As it can be seen, the failure geometry ob-
tained with the proposed mechanism captures the shape of the failure materials with higher friction angles (i.e., φ ≥ 35° ). Therefore, such
given by the numerical model ahead of the tunnel face and above the mechanism is employed next to further explore the influence of the free
free span. Moreover, the failure surfaces in the numerical model define, span on the stability of the face.
approximately, the three blocks that compose the analytical solution;
thereby, the new mechanism is an improvement over the four blocks
mechanism proposed in Senent and Jimenez (2017). Despite the above, 4. Effect of the free span on tunnel face stability
the mechanism in the transition zone (Block 3) seems to be excessively
sharp compared with the results of the numerical simulation. In gen- As shown in Fig. 6, the free span affects the collapse pressure of the
eral, predictions of the failure mechanism geometry improve for cases tunnel face. As expected, the collapse pressure increases with the free
with higher soil strength. span. Moreover, a larger free span can cause the instability of a tunnel
These results suggest that the proposed Limit Analysis mechanism face that would be otherwise stable for shorter free spans; this is illu-
can be used to predict the collapse pressure and the failure geometry of strated by the case with c = 25 kPa and φ = 30° , in which the face is not
tunnel faces considering their free span, especially for tunnels in self-stable for L larger than about 1 m. (When the collapse pressure is

Fig. 5. Comparison of failure geometries obtained with the proposed mechanism and with the numerical model for different strength parameters (φ , c ) and lengths of
the free span (L ).
 null, the face is self-stable and it is not necessary to support it). How- triangular with a constant upper angle (2φ) and, for larger free spans, its
ever, the effect of the free span is less relevant than the effect of the soil base is bigger); and, second, the velocity vector of Block 3 becomes
strength: a reduction in the friction angle from 35° to 30° produces, for slightly more horizontal for larger free spans –i.e., angle α2 of the
the case of c = 5 kPa , an increase of the collapse pressure of around translational velocity of Block 3 (V3 ) reduces–.
40%; whereas its increase due to a larger free span, from 0 to 3 m, is less As previously mentioned, Zhang et al. (2018) studied the effect of
than 13%. These results are coherent with Vermeer et al. (2002), who the free span on the stability of the tunnel face using a Limit Analysis
obtained that there is almost no effect of the free span for frictional mechanism. Fig. 10 compares the collapse pressures computed (i) with
materials and L/ D ratios lower than 0.3. Conversely, Anagnostou and the mechanism proposed by Zhang et al. (2018), (ii) with our me-
Perazzelli (2013) show a higher effect of the free span, even for lower chanism, and (iii) with a numerical simulation in OptumG2 (Upper
L/ D ratios. One reason for such different results could be that, in this Bound analysis) as a function of the free span. Results are computed for
work and in Vermeer et al. (2002) the support pressure is applied both two values of the friction angle: φ = 20° (Fig. 10(a); see Fig. 14 in Zhang
on the face and on the free span; whereas Anagnostou and Perazzelli et al., 2018); and φ = 40° (Fig. 10(b)). (Zhang et al. (2018) considered
(2013) apply support only on the face. Therefore, the proposed me- different values of pressures being applied on the tunnel face and on the
chanism has also been employed to compute the collapse pressure free span: for the results in Fig. 10(a), the pressure on the free span is
without considering any support applied on the free span. (Note that constant and equal to 50 kPa; whereas in Fig. 10(b) the pressure on
the formulation presented in Section 2.3 continues to be valid, after both the free span and the tunnel face are equal). Results show that
removing the first term between parentheses in Eq. (17)). Fig. 7 shows trends computed with both models are similar (i.e., there is an increase
the results of this analysis for the same Test Cases used in Section 3 (see of the collapse pressure with the free span), but, our mechanism pro-
Table 1). As shown in Fig. 7, the increase of the collapse pressure with L vides higher critical pressures than those computed with the me-
is noticeably higher for unsupported free spans, especially for cases chanism proposed by Zhang et al. (2018). Consequently, since both are
with lower strength properties or, more precisely, with lower cohesions. upper-bound limit analysis solutions, our mechanism improves this
Fig. 7 shows some cases for which the collapse pressure cannot be previous mechanism that also considers the free span.
calculated with the proposed mechanism (i.e., with unsupported span, Fig. 10 also shows that, for a friction angle equal to 20° (i.e., sig-
for c = 5 kPa (φ = 30° and 35°), and L > 1 m ). The reason is that the nificantly lower than 35° ), the results of our mechanism are noticeably
free span, when no pressure is applied, is unstable, no matter how much
pressure is applied on the tunnel face. Consequently, a maximum length
of the free span can be defined (LMAX ), above which the free span is
unstable if no pressure is applied on it. Anagnostou and Perazzelli
(2013), based on silo theory, proposed Eq. (20) to calculate LMAX de-
pending on cohesion (c), soil weight (γ ) and excavation width (B); Eq.
(20) can be reduced to Eq. (21) if an infinite excavation width is as-
sumed.
c 1
LMAX =
γ ⎡1 − c
⎤
⎣2 γB ⎦ (20)

2c
LMAX =
γ (21)
Fig. 8 plots the LMAX results computed with Eq. (21) for a soil weight
of 20 kN/m3 and compares them with the LMAX obtained with the
proposed Limit Analysis mechanism. As can be seen, the new analytical
solution divided by two the cohesion needed to keep the free span
stable. This result, as expressed in Eq. (22), can be also derived from the
Limit Analysis upper bound solution proposed by Evans (1984) for a
trapdoor problem of width L (Eq. (23)), if a null pressure is supposed
(σt = 0 ). The difference between both methodologies (Eqs. (21) and
(22)) is probably due to the associated flow rule hypothesis assumed in
Limit Analysis.
4c
́ =
LMAX
γ (22)

1
σt = cotφ ·⎛ ·L·γ − c ⎞
⎝4 ⎠ (23)
Fig. 5 also shows the effect of the free span on the collapse geo-
metry. (In addition, it shows that lower friction angles typically tend to
produce larger mechanisms, and that there is an almost negligible effect
of cohesion on the failure geometry.) For different free span lengths, the
main variation occurs in the zone above the tunnel, as the geometry of
the mechanism ahead of the tunnel face remains almost equal. To il-
lustrate the influence of the free span better, Fig. 9 directly compares
the failure mechanisms obtained with the proposed mechanism for the
Test Cases used in Section 3 (see Table 1). Obviously, the failure always
covers the free span completely, but two other effects can be appre-
ciated: first, the failure mechanism extends further upwards as the free Fig. 7. Collapse pressure vs free span length for different values of cohesion
considering a supported and an unsupported free span: a) φ = 30° ; b) φ = 35° .
span length increases (this can be expected, because Block 3 is
 lower than those computed with the numerical model; in contrast, both
approaches provide very similar collapse pressures for the case with
higher friction angle (40°), hence suggesting that the proposed analy-
tical solution can be used to accurately predict the collapse pressure of
tunnel faces with a free span in materials with higher friction angles
(i.e., φ ≥ 35° ).

5. Effect of a forepole umbrella on the stability of the tunnel face

Forepole umbrellas are a common method to improve the stability

of tunnel faces with an unsupported span. Next, we incorporate the
influence of such support into the proposed Limit Analysis mechanism,
comparing its results with the results of a numerical simulation, and
studying the influence of a forepole umbrella on the collapse pressure
and on the failure geometry.

Fig. 8. Maximum unsupported span vs cohesion.

Fig. 9. Variation of the failure geometry depending on the free span (L = 1, 2, 3): a) φ = 30° , c = 5 kPa ; b) φ = 30° , c = 15 kPa ; c) φ = 35° , c = 5 kPa ; d) φ = 35° ,
c = 15 kPa .
 work by axial force only, and that can fail by tension or by pull-out from
the soil (Michalowski, 2005), forepole umbrellas typically resist forces
perpendicular to its axis, and they therefore work by shear and by
bending –i.e., like a beam–. A similar problem can be found in slope
stability analyses considering anti-slide piles (e.g., Rao et al., 2017);
however, in that case, the vertical piles receive horizontal forces that
depend on displacements (e.g., Winkler model), whereas in the tunnel
face stability problem the support element is horizontal and receives
the weight of the ground over it. Pinyol and Alonso (2011), for un-
drained soils, stated that the micropiles forming the umbrella are
characterized by a limiting (yielding) bending moment, and that the
soil reacts against the micropile with a force per meter of micropile
given by 9cu d (Broms, 1964), where cu is the undrained strength of the
soil and d the diameter of the micropile. In Pinyol and Alonso (2013)
the micropile is idealized as a beam subjected to an imposed dis-
placement, considering that it fails when its tensile strength is reached,
and considering no contribution of the bending moment to the stabi-
lizing work.
In this work, the methodology proposed by Buhan and Salençon
(1993) for slopes is used to study the effect of a forepole umbrella on
the face stability of tunnels excavated in Mohr-Coulomb materials. Two
failures modes are considered, assuming perfect bonding between the
soil and the structural element:

(i) The structural element would fail at the velocity jump surface de-
fined by the Limit Analysis mechanism (Fig. 11(a)), and the max-
imum resistance energy produced by it can be expressed as (Buhan
and Salençon, 1993):
→
W Fg = |VF |·[N20 ·sin2 (φ − α ) + V20·cos2 (φ − α )]1/2 (24)
→
where VF is the velocity of the neutral axis of the structural element at
the velocity jump surface (see Fig. 11(a)); N0 and V0 are the ultimate
axial and shear forces of the umbrella; and α is the angle between the
failure surface and the cross-section of the structural element.

(ii) The structural element would fail after being affected by a con-
tinuous deformation zone with a certain thickness (δ ). Two plastic
Fig. 10. Collapse pressures computed, as a function of the free span, with (i) the
hinges are produced in the reinforcement (Fig. 11(b)) and its
proposed mechanism, (ii) the Limit Analysis methodology presented by Zhang
maximum resistance energy can be expressed as (Buhan and
et al. (2018), and (iii) an Upper Bound Analysis in OptumG2. (D = 10 m;
c = 5 kPa ; γ = 18 kN/m3 ): a) φ = 20° (the pressure applied on the free span is Salençon, 1993):
constant and equal to 50 kPa); b) φ = 40° (the pressure applied on the free span → M
and on the tunnel face are equal). W Fg = |VF |·(N0 ·sin(φ − α ) + 2 0 ·cos(φ − α ))
δ (25)

where M0 is the ultimate bending moment of the structural element.

5.1. Incorporation of a forepole umbrella into the Limit Analysis tunnel face
The collapse pressure of the tunnel face considering the effects of
stability mechanism
the forepole umbrella can be computed incorporating Eq. (24) or (25)
into the work equation (Eq. (19)). As explained later, the forepole
Introducing the effects of a forepole umbrella into a Limit Analysis
umbrella modifies the mechanism in such a way that the velocity of
upper bound solution is not straightforward, since the failure mode of
Block 3 becomes vertical (α2 = 90 ). Therefore, the interface between
the umbrella must be defined in advance, so that its contribution into
Blocks 1 and 3 disappears and the resistance energy due to the forepole
the work equation can be computed. Unlike nail reinforcements that
umbrella only affects interfaces CG and FG (see Fig. 1). Moreover,

Fig. 11. Failures modes considered for the forepole umbrella (modified from Buhan and Salençon, 1993).
φ − α = η (the angle of the umbrella with the horizontal, see Fig. 11) where d is the micropile diameter and μ is a non-dimensional para-
and, considering the small inclination with which umbrellas are usually meter. A typical value for μ is 1/10 (Buhan and Salençon, 1993), so,
installed (typically between 0 and 15°), the contribution of the axial since the forepole diameters range from 25 to 140 mm, δ varies be-
resistance is neglected in Eqs. (24) and (25). tween 0.1 and 0.6 m. The second one (Mechanism II - δNM in Fig. 13) is
obtained after approximating the thickness of the failure zone ahead of
5.2. Effect of the forepole umbrella on the collapse pressure and on the the tunnel face in the numerical model (see Fig. 14) (values of 1.42,
failure geometry of the tunnel face 1.36 and 1.25 m are considered for Cases 24, 23 and 27, respectively;
see Table 1). A wider δ thickness produces a smaller rotation at the
To carry on this analysis, three specific cases from Section 3 with hinges that develop within the structural element (see Fig. 11(b)),
different free span lengths are chosen, in particular, Case 24 (φ = 35° ; hence mobilizing less resistance energy, and, consequently, requiring
c = 5 kPa ; L = 3 m ), Case 23 (φ = 35° ; c = 5 kPa ; L = 2 m ) and Case higher support pressures to maintain the stability of the tunnel face.
27 (φ = 35° ; c = 15 kPa ; L = 1 m ); their support pressures without When a continuous deformation band is considered, the reduction is
reinforcement are equal to 27.4, 25.8 and 10.5 kPa respectively. more gradual than when a ‘sharp’ velocity jump surface is assumed,
Umbrellas vary greatly in stiffness, costs and installation times (Oke although the obtained results are still lower than those computed with
et al., 2014). Typically, umbrellas are divided into (i) light umbrellas, the numerical model. In both cases, the minimum values, for heavier
which consist of steel bars of relatively small diameter (25–50 mm) and umbrellas, corresponds to the minimum values obtained with the nu-
a length shorter than the tunnel height; and (ii) heavy umbrellas, merical model.
formed by micropiles –i.e., steel pipes (usually of diameter 70–150 mm Results in Fig. 13 discussed so far suggest that the failure modes
and wall thickness 6–15 mm) filled with grout and with a length of displayed in Fig. 11 are not completely accurate and may predict unsafe
around 20–25 m–. In this analysis, different reinforcements are con- results. Differences may be caused by the behavior supposed in the
sidered, from a light umbrella composed of 25 mm diameter bars analytical solution, that, considering the results of the numerical model,
spaced 1 m, to a heavy one composed by micropiles of 139.7 mm ex- seems imprecise. In the numerical model, the umbrella works as a fixed
ternal diameter, 14.2 mm thickness and spaced 0.3 m. Table 2 sum- ended beam, with a length larger than the width of the mechanism, and
marizes the geometry, weight, and ultimate loads of the umbrellas developing plastic zones at the fixed ends and center of the structural
considered. (Diameters and thickness have been selected from a com- element. But the proposed analytical model does not exactly reproduce
mercial brochure, trying to cover different amounts of support. Ad- this behavior: the first failure model (by shear) mobilizes to much en-
ditionally, ultimate (yielding) loads (N0 , M0 , V0 ) have been computed ergy, hence providing low estimations even for the lightest umbrella;
according to the Spanish Recommendations for Micropile Design and the second failure mode (by bending) develops four plastic hinges
(Ministerio de Fomento, 1990) disregarding any reduction coefficients (two at each of the two external boundaries of the mechanism crossed
or safety factors and considering a yield strength of the steel of by it). (As previously explained, introducing the forepole umbrella
355 MPa; Eq. (26)). In all cases, a 20 m long umbrella is supposed. modifies the mechanism in such a way that the interface between
(Note that, although this is an excessive length for light umbrellas, this Blocks 1 and 3 disappears, so that the structural element only crosses
allows us to compare results for analogous situations, in which the re- two velocity jumps).
inforcement completely passes through any possible failure block). The Notwithstanding the above, the value of δ can be adjusted to re-
umbrella is assumed to have been installed in the previous excavation produce the results of the numerical simulation. Fig. 13 shows the
cycle –i.e., the initial point of the umbrella is located at a distance of 2L collapse pressures computed with the proposed mechanism, con-
from the tunnel face– and with an inclination angle of 5° (see Fig. 12). sidering values for δ equal to 3.2 times the thickness of the failure zone
observed in the numerical model (δ = 3.2·δNM ) (i.e., values of 4.54, 4.35
1 d3 − di3 2·As and 4.00 m for Cases 24, 23 and 27 respectively; see Mechanism II -
N0 = As ·f y · ; M0 = ·f y ; V0 = ·f
1, 10 6 π· 3 y (26) 3.2·(δNM ) in Fig. 13). It can be observed that the agreement with the
numerical results is excellent. Therefore, it seems possible to estimate
For each case, the umbrellas defined in Table 2 are introduced into
the numerical model described in Section 3.2, and the corresponding
collapse pressures are computed. (For simplicity, in this analysis only Table 2
lower bound values of the collapse pressure –corresponding to upper Properties considered to study the effect of the forepole umbrella on the sta-
bility of the tunnel face.
bound analyses– have been computed). To model the umbrella in Op-
tumG2, a “plate” structural element is used, which “in plane strain is Case Umbrella properties Ultimate loads
equivalent to a standard Euler-Bernoulli beam” (OptumCE, 2019). Fig. 13
Diameter Thickness Center Weight Axial Bending Shear
shows the collapse pressures obtained, as a function of the ultimate
[mm] [mm] separation [kg/m] force moment force
bending moment of the umbrella. As observed, collapse pressures gra- [m] [KN/ [KNm/ [KN/
dually reduce as the umbrella becomes stronger, converging towards a m] m] m]
minimum value that corresponds to a failure mechanism that develops
below the umbrella (see Fig. 15(c)). For comparison, Fig. 13 also shows 1 – – – – – – –
2 25 – 1.0 4.0 223 1 90
the results computed with the proposed Limit Analysis mechanism, and 3 25 – 0.5 8.0 446 3 180
incorporating the effect of the umbrella according to the two failure 4 76.1 6.3 1.0 10.9 446 11 180
modes described previously. For the first failure mode (Fig. 11(a); i.e., a 5 76.1 6.3 0.5 21.8 892 22 360
‘sharp’ failure by axial and shear forces at the velocity jump surface; see 6 82.5 7.1 0.5 26.4 1086 28 438
7 88.9 8.0 0.5 31.8 1312 38 530
Mechanism I in Fig. 13), a fast decrease of the collapse pressure towards
8 95.0 8.8 0.5 37.8 1538 46 622
such minimum value occurs in all cases, even for the lightest umbrella. 9 101.6 9.5 0.5 43.2 1774 58 718
For the second failure mode (Fig. 11(b); i.e., failure by bending mo- 10 108.0 9.5 0.5 46.2 1898 66 768
ment), results depend on the thickness (δ ) of the continuous deforma- 11 114.3 10.0 0.5 51.4 2114 78 856
tion zone. Fig. 13 plots the results obtained for several hypotheses. The 12 121.0 10.0 0.5 54.8 2250 88 910
13 127.0 11.0 0.5 63.2 2588 106 1046
first one (Mechanism II - δBS in Fig. 13) comes from the following ex- 14 133.0 12.5 0.5 74.8 3054 130 1234
pression suggested by Buhan and Salençon (1993): 15 133.0 12.5 0.4 93.5 3818 163 1543
16 133.0 12.5 0.3 124.7 5090 217 2057
4 17 139.7 14.2 0.3 146.7 6023 267 2437
μ= d/ δ
3π (27)
Fig. 12. Outline of the analyzed case to study the effect of the forepole umbrella
on the stability of the tunnel face.

the effect of the forepole umbrella employing the second failure mode
from Buhan and Salençon (1993) (Fig. 11(b); failure by bending mo-
ment) and a thickness of the continuous deformation zone equals to
3.2·δNM . (Note that this methodology is not a rigorous upper bound
solution, so higher values that the actual critical pressure can be ob-
tained).
But note that the methodology just described requires an estimate of
δNM , which is a result from the numerical model. And the methodology
would be useless if it requires conducting numerical analyses that
provide the collapse pressure result that one is aiming to obtain.
Therefore, a simple method to estimate δNM easily is required to make
the methodology useful in practice. To that end, we employed the δNM
values from the 36 validation Tests Cases of Section 3 (Table 1), but
considering three different tunnel diameter values (8, 10 and 12 m), to
fit an equation that estimates δNM as a function of the soil strength
parameters (c and φ ) and of the geometry of the tunnel (D and L ). Based
on our fitting trials, the following equation is proposed.

(L[m])0.03·(D[m])0.39
δNM [m] = − 1.92, L ≥ 0.5
(c[kPa])0.01·(tanφ)0.85 (28)

Fig. 14(a) compares the values of δ extracted from the numerical

model (δNM ) and the values estimated from Eq. (28) (δEq.28 ). The max-
Fig. 13. Collapse pressure vs ultimate bending moment of the forepole um-
imum difference is 0.14 m; considering that the sensitivity of the col-
brella for the three cases analyzed: a) Case 24 (φ = 35° ; c = 5 kPa ; L = 3 m );
lapse pressure to specific values of δ is relatively minor, these predic- b) Case 23 (φ = 35° ; c = 5 kPa ; L = 2 m ); c) Case 27 (φ = 35° ; c = 15 kPa ;
tions are useful. To support this statement, Fig. 14(b) plots the values of L = 1 m ).
the collapse pressure computed for the three cases analyzed (i.e., cases
24, 23 and 27) with the methodology proposed employing the δEq.28
and all the movement is located ahead of the tunnel face, developing a
values estimated with Eq. (28). As shown, the adjustment between the
typical rotational failure.
analytical and the numerical results continues to be excellent. Fig. 14(a)
Similar results are obtained with the proposed Limit Analysis me-
discriminates between different diameters and friction angles. As it can
chanism. When the ‘light’ umbrella is introduced into the mechanism,
be observed, the value of δ increases with D and reduces with φ .
the velocity of Block 3 becomes vertical –i.e., α2 ≈ 90°– so that the ve-
Comparing the displacements of the numerical model and the
locity of Block 3 is equal to the velocity of Block 1 and the interface
failure geometries of the proposed approach provides some interesting
between them disappears. (Because of this, a simpler mechanism de-
observations. As an example, Fig. 15 shows the displacement fields in
pending on only three variables (βE , RE , α1,) can be used, as employed to
the numerical model for Case 24 for three situations: (i) without re-
obtain the results of Figs. 13 and 14.) For the case with the ‘heavy’
inforcement (Fig. 15(a)); (ii) reinforced with a light umbrella
umbrella, the velocity of Block 2 is much larger than the velocities of
(Fig. 15(b)); and (iii) reinforced with a heavy umbrella (Fig. 15(c)).
Blocks 1 and 3. Consequently, the overall energy balance is mainly
(The results of the proposed Limit Analysis mechanism, employing the
between Block 2 and the support pressure, denoting that the ground
failure mode by bending (Fig. 11(b)) and with a width given by 3.2·δNM ,
fails below the umbrella. (The minimum collapse pressures shown in
are superimposed for comparison.) In the case without reinforcement,
Fig. 13 could also be obtained with a mechanism only composed of one
the three blocks described in Section 2 can be clearly identified. When
rotational block ahead of the tunnel face).
the ‘light’ umbrella is introduced (Case 2 in Table 2), displacements
above the tunnel become vertical and only one block can be dis-
tinguished in this zone. With the ‘heavy’ umbrella (Case 17 in Table 2),
the displacements above the reinforcement are practically negligible,
Fig. 14. Analysis of Eq. (28): a) comparison between δ values estimated from
the numerical model (δNM ) and those computed with Eq. (28) (δEq.28 ); b) com-
parison of collapse pressures computed employing δNM and δEq.28 for the three
cases analyzed.

6. Conclusions

In this work, we proposed and validated an improved 2D Limit

Analysis failure mechanism to study the face stability of tunnels con-
structed in Mohr-Coulomb materials. The main contribution of the new
mechanism is that it considers the influence of the free span (or the
length of the tunnel without support), and the formulation has been
Fig. 15. Comparison of failure geometries obtained with the proposed me-
further extended so that it can consider the effect of a forepole um-
chanism and with the numerical model in Case 24 (φ = 35° ; c = 5 kPa ;
brella. The mechanism is formed by three non-deformable blocks (two L = 3 m ) for different amounts of supports: a) without umbrella; b) light
translational and one rotational) in such a way that the interface be- umbrella; c) heavy umbrella.
tween one of the translational blocks and the rotational block is ob-
tained by an iterative process that assures that the associated flow rule
the influence of the free span is also shown to significantly depend on
imposed by Limit Analysis is fulfilled.
the assumption about how the support pressure is applied, as the be-
Results of the proposed mechanism are compared with results of a
havior can change significantly when (a) the support pressure is applied
numerical model built in the Finite Element Limit Analysis code
only on the tunnel face but not on the free span, as opposed to (b) when
OptumG2, showing that the proposed analytical solution can be em-
the support pressure is applied both on the tunnel face and on the free
ployed to reasonably predict (i) the value of the collapse pressure, and
span.
(ii) the collapse geometry, for tunnels in materials with higher friction
The failure geometry varies with the length of the free span, so that
angles (i.e., φ ≥ 35° ).
its length mainly affects the failure geometry above the tunnel crown,
Results confirm the expected effect of the free span on the collapse
as higher and more vertical mechanisms are produced for larger free
pressure, as larger free spans entail higher collapse pressures. However,
spans. However, the influence of the free span on the geometry of the
rotational block that develops ahead of the tunnel face is almost neg- Engineering V. CRC Press, pp. 228–235.
ligible, and results have shown that changes of this block are mainly Davis, E.H., Gunn, M.J., Mair, R.J., Seneviratine, H.N., 1980. The stability of shallow
tunnels and underground openings in cohesive material. Geotechnique 30 (4),
due to the well-known variation of shape of face rotational blocks that 397–416. https://doi.org/10.1680/geot.1980.30.4.397.
occur associated to changes of the ground friction angle (see e.g., Evans, C.H., 1984. An examination of arching in granular soils. Master Thesis,
Massachusetts Institute of Technology, Cambridge, USA.
Mollon et al., 2011b).
Hayati, A.N., Ahmadi, M.M., Hajjar, M., Kashighandi, A., 2013. Unsupported advance
Finally, the stability of the tunnel face with a free span is studied length in tunnels constructed using New Austrian Tunnelling Method and ground
considering the reinforcement provided by a forepole umbrella. The surface settlement. Int. J. Numer. Anal. Methods Geomech. 37 (14), 2170–2185.
https://doi.org/10.1002/nag.2128.
contribution of the umbrella is introduced into the mechanism con- Health and Safety Executive, 1996. Safety of new Austrian tunnelling method (NATM)
sidering two failure modes proposed by Buhan and Salençon (1993). tunnels. A review of sprayed concrete tunnels with particular reference to London
Results show that a direct implementation of such failure modes tends Clay, (HSE) Books, Sudbury.
Horn, N., 1961. Horizontaler erddruck auf senkrechte abschlussflächen von tunnelröhren.
to overpredict the effect of the umbrella on the support pressure. Landeskonferenz der ungarischen tiefbauindustrie. Deutsche Überarbeitung durch
Consequently, a new methodology has been proposed to estimate the STUVA, Düsseldorf, 7–16.
Le, B.T., Taylor, R.N., 2018. Ground response to tunnelling incorporating a soil re-
value of the collapse pressure more accurately, and without the need to inforcement system. Can. Geotech. J. https://doi.org/10.1139/cgj-2018-0075.
conduct numerical analyses. Results confirm that the proposed me- Lee, Y.Z., Schubert, W., 2008. Determination of the round length for tunnel excavation in
chanism is able to capture the stability improvement (i.e., the reduction weak rock. Tunn. Undegr. Sp. Tech. 23 (3), 221–231. https://doi.org/10.1016/j.tust.
2007.04.001.
of the collapse pressure) produced by the umbrella, as well as the Leca, E., Dormieux, L., 1990. Upper and lower bound solutions for the face stability of
variation of the failure mechanism. Therefore, the proposed limit ana- shallow circular tunnels in frictional material. Géotechnique 40 (4), 581–606.
https://doi.org/10.1680/geot.1990.40.4.581.
lysis mechanism can reproduce the critical failure geometry associated
Lyamin, A.V., Sloan, S.W., 2002a. Lower bound limit analysis using non–linear pro-
to different amounts of support; in particular, it can predict a local gramming. Int. J. Numer. Meth. Eng. 55, 573–611. https://doi.org/10.1002/nme.
failure below the support for heavy umbrellas. 511.
Lyamin, A.V., Sloan, S.W., 2002b. Upper bound limit analysis using linear finite elements
and non–linear programming. Int. J. Numer. Anal. Methods Geomech. 26, 181–216.
CRediT authorship contribution statement https://doi.org/10.1002/nag.198.
Michalowski, R.L., 2005. Limit analysis in geotechnical engineering. Prediction and si-
mulation methods in geomechanics. State of the art report on selected topics by
Salvador Senent: Conceptualization, Methodology, Software, TC-34.
Validation, Investigation, Formal analysis, Writing - original draft, Ministry of Development, 1990. Guía para el proyecto y la ejecución de micropilotes en
obras de carretera. Centro de Publicaciones, Secretaría General Técnica, Ministerio de
Funding acquisition. Congke Yi: Validation, Investigation, Formal
Fomento, Madrid.
analysis, Writing - original draft. Rafael Jimenez: Conceptualization, Mollon, G., Phoon, K.K., Dias, D., Soubra, A.H., 2011a. Validation of a new 2D failure
Methodology, Investigation, Writing - review & editing, Supervision, mechanism for the stability analysis of a pressurized tunnel face in a spatially varying
sand. J. Eng. Mech. 137, 8–21. https://doi.org/10.1061/(ASCE)EM.1943-7889.
Project administration, Funding acquisition. 0000196.
Mollon, G., Dias, D., Soubra, A.H., 2011b. Rotational failure mechanisms for the face
stability analysis of tunnels driven by a pressurized shield. Int. J. Numer. Anal. Meth.
Declaration of Competing Interest
Geomech. 35, 1363–1388. https://doi.org/10.1002/nag.962.
Oke, J., Vlachopoulos, N., Marinos, V., 2014. Umbrella arch nomenclature and selection
methodology for temporary support systems for the design and construction of tun-
The authors declare that they have no known competing financial
nels. Geotech. Geol. Eng. 32 (1), 97–130. https://doi.org/10.1007/s10706-013-
interests or personal relationships that could have appeared to influ- 9697-4.
ence the work reported in this paper. Optum Computational Engineering, 2019. OptumG2, Finite Element Program for
Geotechnical Analysis. Version 2019, www.optumce.com.
Pan, Q., Dias, D., 2017. Safety factor assessment of a tunnel face reinforced by horizontal
Acknowledgements dowels. Eng. Struct. 142, 56–66. https://doi.org/10.1016/j.engstruct.2017.03.056.
Pinyol, N.M., Alonso, E., 2011. Design of micropiles for tunnel face reinforcement: un-
drained upper bound solution. J. Geotech. Geoenviron. 138 (1), 89–99. https://doi.
This research was funded, in part, by the Spanish Ministry of org/10.1061/(ASCE)GT.1943-5606.0000562.
Economy, Industry and Competitiveness under Projects BIA 2015- Pinyol, N.M., Alonso, E., 2013. An undrained upper bound solution for the face stability of
tunnels reinforced by micropiles. In: Proceedings of the 18th International
69152-R and TUÑEL. The second author holds a PhD Scholarship pro-
Conference on Soil Mechanics and Geotechnical Engineering, pp. 3395–3398.
vided by the China Scholarship Council. The support of both institu- Qian, Z., Zou, J., Pan, Q., Dias, D., 2019. Safety factor calculations of a tunnel face re-
tions is gratefully acknowledged inforced with umbrella pipes: A comparison analysis. Eng. Struct. 199, 109639.
https://doi.org/10.1016/j.engstruct.2019.109639.
Rao, P., Zhao, L., Chen, Q., Li, L., 2017. Limit analysis approach for accessing stability of
Appendix A. Supplementary material three-dimensional (3-D) slopes reinforced with piles. Mar Georesour Geotec 35 (7),
978–985. https://doi.org/10.1080/1064119X.2016.1273982.
Senent, S., Jimenez, R., 2017. Study of the tunnel face stability considering the free span
The MATLAB source code for computation of the critical pressure of by means of numerical models and analytical solutions. In: EURO: TUN 2017,
tunnel faces considering the free span using the proposed mechanism Innsbruck University, Austria.
Sloan, S.W., 1988. Lower bound limit analysis using finite elements and linear pro-
has been made available as supplementary material to this article.
gramming. Int. J. Numer. Anal. Methods Geomech. 12, 61–77. https://doi.org/10.
Supplementary data to this article can be found online at https://doi. 1002/nag.1610120105.
org/10.1016/j.tust.2020.103515. Sloan, S.W., 1989. Upper bound limit analysis using finite elements and linear pro-
gramming. Int. J. Numer. Anal. Methods Geomech. 13, 263–282. https://doi.org/10.
1002/nag.1610130304.
References Tamez, E., Holguin, E., Rangel, J.L., 1997. Diseño geotecnico de túneles. TGC Geotecnia,
Mexico.
Terzaghi, K., 1936. Stress distribution in dry and in saturated sand above a yielding trap-
Anagnostou, G., Perazzelli, P., 2013. The stability of a tunnel face with a free span and a door. In: Proc., 1st International Conference on Soil Mechanics and Foundation
non-uniform support. Geotechnik 36 (1), 40–50. https://doi.org/10.1002/gete. Engineering, Cambridge, Massachusetts, pp. 307–311.
201200014. van Tol, A.F., 2006. The effects of tunnelling on existing structures. In: Bakker et al. (Eds.
Broms, B.B., 1964. Lateral resistance of piles in cohesive soils. J. Soil Mech. Found. Div. ), Geotechnical Aspects of Underground Construction in Soft Ground. Taylor and
90 (2), 27–64. Francis. London. pp. 31-41.
Buhan, P., Salençon, J., 1993. A comprehensive stability analysis of soil nailed structures. Vermeer, P.A., Ruse, N., Marcher, T., 2002. Tunnel heading stability in drained ground.
Eur. J. Mech. A-Solid. 12 (3), 325–345. Felsbau 20 (6), 8–18.
Casarin, C., Mair, R.J., 1981. The assessment of tunnel stability in clay by model tests. In: Yan, S., Wang, R., Bai, J., Wu, W., Elmo, D., 2018. An alternative approach to determine
Soft Ground Tunneling, A. A. Balkema, Rotterdam, pp. 33–44. cycle length of roadway excavation in coal mines. J. Ambient. Intell. Human.
Chambon, P., Corte, J.F., 1994. Shallow tunnels in cohesionless soil: stability of tunnel Comput. 1–8. https://doi.org/10.1007/s12652-018-0987-9.
face. J. Geotech. Eng. 120 (7), 1148–1165. https://doi.org/10.1061/(ASCE)0733- Zhang, Z., Shi, X., Wang, B., Li, H., 2018. Stability of NATM tunnel faces in soft sur-
9410(1994)120:7(1148). rounding rocks. Comput. Geotech. 96, 90–102. https://doi.org/10.1016/j.compgeo.
Chen, W.F., 1975. Limit Analysis and Soil Plasticity. Elsevier, Amsterdam. 2017.10.009.
Costa, A., Cardoso, A.S., Sousa, J.A., 2007. The influence of unlined excavation length in
face stability. In: Applications of Computational Mechanics in Geotechnical