Tunnelling and Underground Space Technology 77 (2018) 13–25

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

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

Analytical and experimental evaluation of the use of fibers as partial T

reinforcement in shotcrete for tunnels in Chile
⁎
Leonardo M. Massone , Francisco Nazar
Department of Civil Engineering, University of Chile, Blanco Encalada 2002, Santiago, Chile

A R T I C LE I N FO A B S T R A C T

Keywords: This work evaluates experimentally the use of steel or polypropylene fibers in shotcrete as a partial replacement
Tunnel for the traditional reinforcement of electrowelded mesh used at the tunnel support for the subway in Santiago
Shotcrete (Chile). For the experimental part, 8 tests were performed on section-scaled (1:2) slender (half-span to depth
Fiber reinforced concrete ratio a/d ∼ 5) specimens under transversal (flexure and shear) and axial loads (0.02f′cAg and 0.07f′cAg) and with
Seismic response
different reinforcements layout; and another 4 tests of similar specimens, but with low half-span to depth ratio
Steel fiber
Polypropylene fiber
(a/d ∼ 1.5). The section includes welded mesh (one face), a central reticulated frame and a plane shotcrete with
a welded mesh (opposite face) or reinforced shotcrete (steel or polypropylene). The results showed a very im-
portant contribution of the reticulated frame in flexion (main internal reinforcement) and a modest contribution
of the welded mesh and fibers. A two-dimensional finite element modeling of the tunnel-ground system using
OpenSees is also performed using calibrated models for the tunnel section based on the experimental part. The
soil is modeled with 9-node quadrilateral elements, the lining is modeled with beam-column elements with fiber
sections, and the interaction between the lining and the soil is modeled by the Winkler approach without tension
in the direction normal to the contact surface and with perfect adherence in the tangential direction. The model
is subjected to the static loads from the excavation, modeled considering the constructive sequence of the tunnel
using the α method, and then to a seismic analysis by means of the shear wave method (distortion). The results
show that the safety factors implicit in the traditional design are high, implying that the structure remains
elastic.

1. Introduction Other support measures are reticulated frames, glass fiber bolts, and
longitudinal umbrellas, among others. The first system consists of
A tunnel is an underground passageway dug through the sur- prefabricating curved segments and joining them together by being
rounding rock or soil, which is used to enable passage for vehicles, seated in the ground. The second consists of shotcreting against the
people, or water. The material surrounding the tunnel depends on the ground and other supporting elements, such as meshes, reticulated
terrain conditions. In rock, for example, there may be no need for any frames, bolts, etc. The last system consists of concreting in a traditional
support to be sustained, and structural stability is entirely dependent on way, with moldings, and could be combined with the second method.
the rock itself. In materials of lower self-supporting capacity, such as Nowadays, thanks to the proliferation of the New Austrian Tunneling
soil, it is necessary to incorporate some additional structure that sup- Method (NATM), the use of shotcrete for the lining has become more
ports the material, and since tunnels usually require their interior with widespread (Kolymbas, 2005). The shotcrete is typically reinforced
open access, the use of linings that act as arches (together with the soil’s with electrowelded meshes, and in recent times discrete fibers have
own self-support capacity) is the preferred solution. been used as replacement of steel mesh, using what is traditionally
The structural design should consider the bending induced on the known as fiber reinforced concrete (FRC). FRC is a material made with
lining. Estimating the bending stresses induced on the tunnel support is hydraulic cement, aggregates of various sizes, incorporating discrete,
complex, because it is difficult to estimate how the loads are distributed discontinuous fibers (Bentur and Mindess, 2006). FRC is attributed
between the lining and the soil, as well as the earthquake-induced great benefits, both structural and non-structural. Examples of the first
loads. The literature recognizes three different systems of commonly are: (i) greater ductility; (ii) better cracking control; (iii) better flexural
used concrete tunnel linings: (i) prefabricated segment linings; (ii) behavior; and (iv) residual tensile strength, among others. Examples of
projected (shotcrete) concrete linings; and (iii) in-situ concrete linings. the second ones are: (i) a better performance in the case of freeze-thaw

⁎
Corresponding author.
E-mail addresses: [email protected] (L.M. Massone), [email protected] (F. Nazar).

https://doi.org/10.1016/j.tust.2018.03.027
Received 4 September 2017; Received in revised form 20 March 2018; Accepted 24 March 2018
Available online 28 March 2018
0886-7798/ © 2018 Elsevier Ltd. All rights reserved.
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

Fig. 1. Sections of the tunnel lining (bottom floor) – (a) Traditional section with double reinforcement of steel welded mesh (primary and secondary lining) and
reticulated frame and (b) Proposed section with steel welded mesh in the secondary lining, steel or polypropylene fibers in the primary lining, and reticulated frame.

cycles; (ii) better impact and abrasion behavior; and (iii) increased mechanical behavior of tunnel lining sections structured with re-
durability resulting from better cracking control, among others. In ad- ticulated steel frames, shotcrete with fibers (steel or synthetic), and
dition, in the case of applications where the fibers completely replace traditional reinforcement meshes subjected to bending, shear and axial
the traditional reinforcement, time saving of reinforcement placing can loading in the direction perpendicular to the axis of the tunnel. The
become considerable. purpose is to compare the effect of the replacement of the traditional
Previous experimental research around FRC (Belletti et al., 2004; reinforcing mesh with fibers, and to observe the contribution of the
Wetzig et al., 2004; Kang et al., 2012) as well as practical usage of FRC reticulated frame.
in tunnel linings (Chiaia et al., 2009; De la Fuente et al., 2012) can be In order to achieve it, material properties were characterized and
found mainly in Europe, where the focus has been placed mainly also specimens were designed and tested to capture flexural and shear
around steel fibers using particular construction methods and tech- response. With respect to the element tests, 12 bending tests were
nology, as well as shapes and quantity of fibers. On the other hand, performed with axial load, with different reinforcement solutions, dif-
polypropylene fibers are currently gaining popularity in the mining ferent half-span to depth ratios (M/Vd, moment-to-shear to reinforce-
industry in Chile and Australia in stiff soil/rock applications, where ment level arm ratio) and also varying the axial load. The specimens
they are chosen because they are less prone to corrosion and they were constructed at 1:2 scale. For the characterization of the material
produce less wear on the machines. Few comparisons can be found in properties, the following additional tests were also performed: (a)
the literature between steel and polypropylene fibers with a particular bending tests on specimens with indentations and without reinforce-
focus on the structural behavior for shotcrete in tunnels. The motivation ment, following the EN 14651 (2007) methodology, (b) compressive
of this work is to evaluate the partial replacement of the traditional tests of cylindrical cores obtained from the test specimens, (c) direct
reinforcement of electrowelded meshes by steel or polypropylene fibers tensile tests on shotcrete cores with and without fibers and (d) uniaxial
for a subway tunnel in Santiago (Chile), using the fibers and con- steel tensile tests of meshes and frames.
struction technology available and typically used in the Chilean prac- According to Nazar (2016), in the actual tunnels under considera-
tice. This work carries out an experimental and analytical evaluation of tion (line 6 of the Santiago Metro), the support (also called the primary
a lining reinforced with electrowelded meshes or with steel or poly- lining) is 15 to 25 cm thick, while the lining (also called secondary
propylene fibers. Scaled (1:2) sections under flexo-compression tests are lining) is of a minimum thickness of 15 cm, being both shotcreted ele-
performed on typical sections of shotcrete linings, constructed with ments. The typical steel meshes used as reinforcement are, depending
concrete with and without fibers, electrowelded meshes, and a re- on the ground type, C295 (295 mm2/m area) or C338 (338 mm2/m)
ticulated embedded frame. On the other hand, the observed behavior is grade AT56-50H (fy = 500 MPa, nominal) electrowelded meshes, with
numerically modeled, and a finite element analysis of the soil-tunnel bar diameters of 7.5 mm and 8.5 mm respectively, spaced at 150 mm. In
system is performed to establish the demands to which the tunnel will addition, three-bar reticulated frames (one 28 mm diameter bar and
be subjected, both to static and seismic loads. two 22 mm diameter bars) of grade A630S steel (fy = 420 MPa, nom-
inal) or grade A42-27H (fy = 270 MPa, nominal) with height variable
2. Design of experiments and construction of specimens with ground type (145 mm for gravel and 180 mm for soil), and with a
variable separation between them, which also depends on the type of
2.1. General description of experiments ground type (1.0 to 1.5 m for gravel, and 0.5 to 1.0 m for fines or ir-
regular areas) are considered. Fig. 1 shows the sections of the tunnel.
The primary objective of the experiments was to evaluate the For the purposes of this work, the tunnels found in gravel were

14
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

Fig. 2. Characteristics of specimens – (a) A (L = 150 cm), (b) B (L = 150 cm), (c) DA (L = 150 cm), (d) CA (L = 80 cm), (e) DP (L = 150 cm) and (f) CP (L = 80 cm).

Table 1
General characteristics of specimens.
Specimen name Dimensions Fibres Conventional reinforcement

Length (cm) Width (cm) Height (cm) Type Base mix dosage Upper mesh Lower mesh Reticulated frame

A1 150 50 15 None None C139 (Φ4.2@100) C139 (Φ4.2@100) None

A2
B1 Φ12 + 2Φ10
B2
3
DA1 DRAMIX 3D 65/35 BG 35 kg/m None
DA2
DP1 EPC BarChip 48 6 kg/m3
DP2
CA1 80 DRAMIX 3D 65/35 BG 35 kg/m3
CA2
CP1 EPC BarChip 48 6 kg/m3
CP2

considered. To maintain the 1:2 scaling, a 15 cm section was considered

Table 2 for the specimen, with a C139 mesh (1.39 cm2/m) and a frame com-
Concrete mixture description. posed of two 10 mm bars and one 12 mm bar, with a frame height of
Material Specification Description Supplier Quantity 80 mm and a separation of 0.5 m. Fig. 2 shows a schematic cross-sec-
tions of the specimens along with their longitudinal dimensions. Also,
Cement Melón Extra High strength grade Melón 425 kg/m3 two lengths of the specimens were considered, 1.3 m and 0.5 m as the
Sand 1 Melón Semi-Industrial Coarse sand Melón 1303 kg/m3
distance between supports, in order to capture the response of slender
Pozo San Bernardo
Sand 2 Melón José Catalán Fine sand Melón 326 kg/m3
and short elements. Table 1 shows the general characteristics of the 12
Aditive 1 MX-1390 Water reducing/ Sika 2.13 kg/m3 reinforced specimens.
plasticising aditive
Aditive 2 Viscocrete 5100 Water reducing/ Sika 3.83 kg/m3 2.2. Materials and construction of specimens
plasticising aditive
Aditive 3 Microsílice Microsilica PSI 29.8 kg/m3
Water – – – 210 lt/m3 The test specimens were built within the tunnel to maintain similar
process as for the tunnel construction, where the concrete was sprayed
with the same characteristics as in the real tunnel construction, except
that the concrete was sprayed onto moldings, instead of projecting it

15
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

directly onto the ground. Concrete mixtures (see Table 2) were identical center of the beam. All 12 specimens simulate sections of tunnels and
for all cases, and fibers were added to the mixture in the plant and then therefore have an axial (compressive) load applied that represent the
the mixture was transported using a concrete mixer truck to the site to actions over the lining due to ground thrust and the geometric shape of
be sprayed. The concrete had a nominal cylindrical strength of 35 MPa, the element. The axial load system consists of two hollow jacks with
and the actual strength was determined via testing on cored cylinders. passing bars of 15 mm in diameter, which are connected to the spe-
The equipment used to spray the concrete was the same used by the cimen and jacks end by means of bolted plates. The jacks are placed
contractors to build the actual tunnel, and consisted in (i) a Putzmeister horizontally and react against a steel beam, which is hold horizontally
Tk40 pump, with 10 S per minute, (ii) an Atlascopco Xas 420 com- with the specimens. On the other side of each specimen an equal steel
pressor at 7 bar pressure, and (iii) Sigunit STM-AF Mining setting ac- profile is placed with a plate, passing the bar through the steel profile
celerator in a quantity of 5% of the weight. parallel to the specimen and anchored.
Additionally to the specimens described in Table 1, six beam tests To measure displacements of the specimens 5 LVDT were used, with
were built to characterize the fracture mechanics of the material under 2 or 3 of them located in the central axis where the transverse load is
bending loading. The beams are 70 cm long, 15 cm wide and high, and a applied. A load cell was used to measure transverse loading during the
notch of 25 mm (height) was placed in the middle on the tensile side, as test, as well as, a tubular load cell (washer type) was used for axial
dictated by EN 14651 (2007). Steel and polypropylene fibers were used loading. DSLR cameras were used to measure displacements and de-
in the same quantities as for the steel reinforced beams tests. In addi- formations through photogrammetry (image correlation). In order to
tion, 3 forms of dimensions 500 mm by 500 mm by 300 mm were perform the digital image correlation analysis (DIC), the program Ncorr
projected with concrete to extract 10 cm diameter by 20 cm high cy- (Blaber et al., 2015) has been used in conjunction with Ncorr_post
linders. (Nežerka et al., 2016). Ncorr is an open source program that performs
Dramix 3D 65/35 BG steel fibers were used in a dosage of 35 kg/m3 digital image correlation analysis (DIC) and has implemented nonlinear
and EPC BarChip 48 polypropylene fibers in a dosage of 6 kg/m3. It is methods of optimization, interpolation using biquintic splines, and is
important to mention that in local practice, due to equipment selection, capable of performing analysis at large deformations.
it is considered that these dosages are the maximum dosages that can be A schematic drawing of the test of one of the slender specimens is
specified without generating major construction problems (i.e., pump shown in Fig. 4. The test of short specimens has a similar testing
clogging). Dramix 3D 65/35 BG fibers are non-textured cold-drawn scheme.
steel fibers of circular cross-section, with single-ended hooks with
0.55 mm in diameter and 35 mm in length (aspect ratio 35/0.55 ∼ 65) . 3. Experimental results
The nominal tensile strength is 1345 MPa, with a nominal elastic
modulus of 200 GPa. On the other hand, BarChip 48 EPC fibers are This chapter details the experimental results, both in terms of
longitudinally straight polypropylene (olefins) fibers, 48 mm long with overall response and photogrammetry. All tests were performed be-
a rectangular section and textured surface. The nominal tensile strength tween approximately 200 and 240 days of concrete maturity, and ma-
is 640 MPa and the nominal elastic modulus is 12 GPa. Photographs of terial properties were characterized at the time of testing.
the fibers are shown in Fig. 3.
3.1. Tensile response of steel

2.3. Test setup Uniaxial tensile tests were performed on the reinforcing steels used
for the construction of the specimens. Fig. 5 shows the tensile curves of
The beams are sustained on simply-supported end joints (separated ϕ10 and ϕ12 bars of grade A630S steel, and of bars ϕ4.2 of grade AT50-
at 1.3 m or 0.5 m) and loaded transversally with an actuator in the 56H steel. The solid lines correspond to representative curves of each
set of tests. The average yield stress of grade A630S steel was 524 MPa
and 498 MPa (diameters of 10 mm and 12 mm, respectively), whereas
for grade AT50-56H steel it was 557 MPa. As shown in Fig. 5, the steel
grade AT56-50H achieves a fracture strain of just about 1.5%, whereas
the steel grade A630S, on the other hand, reaches a fracture strain of
7% or more.

3.2. Shotcrete in compression

Uniaxial compression tests were performed on cylinders (3 per type)

of shotcrete with and without fibers, with dimensions of 100 mm in
diameter and 200 mm in height. The specimens where cored as per
ASTM C42/C42M. The average capacity of plain concrete was
55.3 MPa, whereas for the concrete with steel fibers it was 59.9 MPa. In
the case of concrete with polypropylene fiber, one of the specimens had
an anomaly and was discarded, resulting in an average of 55.7 MPa.
The stress-strain relationships obtained from the experiments are shown
in Fig. 6. It is possible to appreciate that the addition of steel fibers gives
a slightly greater ductility to the material, switching the deformation at
peak strength from 0.25% to 0.4%. For the case of polypropylene fibers,
the results were less conclusive.

3.3. Shotcrete in tension

Direct tension tests were carried out on 100 mm diameter and

Fig. 3. Fiber comparison for the experimental study (Dramix 3D 65/35 BG and 200 mm high shotcrete (with and without fibers) specimens with a
EPC BarChip 48). 10 mm center notch running around the perimeter (implying an

16
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

Fig. 4. Test scheme for a beam with axial load (slender beams).

900 effective diameter of 80 mm in the central area, that is, an effective area
800 AT56-50H reduced by 36%). The specimens are cored as per ASTM C42/C42M.
A630S Each specimen is glued (epoxy) between plates, and these plates are
700
attached to universal testing machine clamps, that pull the core in
600 tension. Two specimens were tested per type, and in 4 of them (both
specimens with steel fibers, one with polypropylene fibers and one
Stress (MPa)

500
without fibers) failure occurred outside the notch (see Fig. 7a), in-
400 dicating that the material is not homogeneous and has a capacity at
AT56-50H 1 AT56-50H 2 least 36% lower in the wider section. The tensile strength based on the
300 AT56-50H 3 AT56-50H 4
A630S-D10-1 A630S-D10-2 reduced section was very variable, with values ranging between
200 A630S-D10-3 A630S-D12-1 1.5 MPa and 2.5 MPa for plain concrete, 1.1 MPa and 1.6 MPa for
A630S-D12-2 A630S-D12-3
concrete with polypropylene fiber, and 0.3 MPa and 0.5 MPa for con-
100
crete with steel fibers. When inspecting the specimens, it was observed
0 that very few fibers crossed the crack, with most of the fibers having an
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
orientation parallel to the crack probably due to the direction of con-
Strain
crete projection (perpendicular to the extraction). The higher capacity
Fig. 5. Tensile steel tests for types A630S and AT56-50H. of the unreinforced specimens suggests that the fibers generated a
failure plane reducing their capacity.
70

60 3.4. Bending tests on prisms

50 Two specimens per fiber type for bending tests were carried out
under controlled procedure EN 14651 (2007) on shotcreted prisms
Stress (MPa)

40 (over molds) of dimensions 150 mm in height, 150 mm in width and

700 mm in length; with spacing between supports of 500 mm. Testing of
30 the test specimens without fibers was not performed. Unlike the direct
tensile test (in this case the fibers would be sprayed placing them
20 parallel to the loading direction), all test specimens have similar ca-
Plain 1 Plain 2 Plain 3
Steel 1 Steel 2 Steel 3 pacities and consistent with the expected strength for the plain con-
10
PlasƟc 1 PlasƟc 2 crete. The impact on fiber ductility is appreciable where steel works
from the first cracking stage, whereas polypropylene requires a larger
0
0 0.002 0.004 0.006 0.008 0.01 0.012
crack size (vertical displacement) to observe its effectiveness. Fig. 8
Strain shows the results.

Fig. 6. Stress versus strain response of concrete cylinders in compression.

17
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

Fig. 7. Tensile test of concrete cylinders – (a) failure in the unreduced section, and (b) failure in reduced section.

20 3.5.1. Specimens A1 and A2

Specimens A1 and A2 have a single central bending crack (without
the presence of minor cracks) starting at approximately 20 kN and
15 27 kN, respectively (Fig. 9). Both reach similar capacities (40 kN). In
Transverse load (kN)

the case of A1, there is a rapid degradation that ends with fracture of
the bottom steel mesh. In the case of A2, the test was stopped before
reaching the fracture of the upper steel mesh (the bottom steel mesh
10
was fractured). Specimen A2 (higher axial load) offers much more
ductility than A1 (lower axial load), but has very similar load carrying
capacity (Fig. 15).
5
Steel 1 Steel 2 3.5.2. Specimens B1 and B2
PlasƟc 1 PlasƟc 2
Specimens B1 and B2 have a slightly higher cracking load than the
0 ones without the reticulated frame (A1 and A2), with loads of 24 kN
0 0.5 1 1.5 2 2.5 3 3.5 and 30 kN, respectively. In this case, however, the damage progresses
Crack mouth opening displacement - CMOD (mm) concentrated in a central flexural crack, but there are minor secondary
Fig. 8. Load versus deflection response of prism. flexural cracks (Fig. 10). Again, it is appreciated that the effect of the
larger axial load is almost negligible in terms of capacity. This time the
ductility is higher for the low axial load case, which is explained by the
3.5. Bending tests of beams with axial load
presence of the reticulated frame, which gives more ductility in both
specimens, with great impact compared to the specimens without the
Twelve flexural tests were performed (Figs. 9–14), with different
frame (Fig. 15). In both specimens (B1 and B2) the bottom steel mesh
bearing lengths (to observe the shear stress impact), as well as re-
was found fractured at the end of the test. The reticulated frame did not
inforcement configuration and axial loads. The axial loads were chosen
fracture.
(60 kN and 200 kN) to be consistent with the tunnel under study, and
such variable was relevant considering that the axial load can have an
3.5.3. Specimens Da1 and Da2
important impact on ductility (Wallace et al., 2008). Table 3 shows the
Specimens DA1 and DA2 presented similar behavior to specimens
specimen description.
B1 and B2, with a central bending crack (this time somewhat diagonal –
All settlement or adjustment of supports observed in the specimens,
Fig. 11) and minor secondary bending cracks, although in the case of
especially the ones with large transversal strength (short beams), was
DA2 there are practically 2 main cracks (Fig. 11b). These cracks start
corrected in the central LVDTs by subtracting the support settlements.
approximately at loads of 21 kN and 24 kN, respectively. The maximum
Another important aspect to note is that initially specimens CP2 and
loads are similar in both specimens. At the end of the test it is possible
CA2 used spacing between supports of 500 mm, however, as a shear
to verify that the steel fibers close to the tensile side failed due to
(diagonal) failure was not observed, the separation was changed to
sliding. No fracture is observed in the steel mesh or frame. Again, it is
400 mm for the specimens CP1 and CA1.
appreciated that the effect of the larger axial load is almost negligible
(Fig. 15).

Fig. 9. Failure mode – (a) specimen A1 and (b) specimen A2.

18
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

Fig. 10. Failure mode of specimens B1 and B2 – (a) concrete cracking and (b) fracture of inferior steel welded mesh.

3.5.4. Specimens Dp1 and Dp2 seen by comparing the response of specimens A (without frame) with
Specimens DP1 and DP2 presented similar behavior to specimens specimens with frame, both with steel mesh (B) and with fibers (DA and
DA1 and DA2, showing 2 main central bending cracks (somewhat di- DP). It is also possible to conclude that the fibers contribute to sectional
agonal) and minor secondary flexural cracks (Fig. 12). These cracks capacity as much as or even more than the welded steel mesh, although
start approximately at loads of 18 kN and 32 kN, respectively. The this contribution is modest. It is observed that the specimens with steel
maximum loads are similar in both specimens. At the end of the test, fibers have a behavior similar to that of the B specimens (frame and
only the bottom steel mesh was fractured. It is possible to verify that the double mesh), while the specimens with polypropylene (DP) fibers have
polypropylene fibers close to the tensile side rupture. Fig. 12b shows sectional capacities somewhat larger than the specimens comparable
the damage concentration in the center, but this time distributed over a with frame (DA and B), showing at the same time minor ductility. The
larger length than the previous tests. higher axial load results in faster strength degradation for all speci-
mens. In the case of short specimens (Fig. 16), higher capacity and
3.5.5. Specimens CA1 and CA2 lower ductility is observed in all specimens, compared to slender spe-
The axial load remained constant, however, in the CA1 experiment, cimens. Small differences are observed between both fiber types, with
the stroke of the axial actuators did not reach to accommodate the slightly earlier initiation of degradation of specimens with poly-
change of length during the experiment, increasing the axial load to propylene fiber. The impact of axial load is less evident than with
110 kN at the end of the test (at an central displacement of 12.0 mm). slender specimens.
Specimen CA1 presented cracking at an estimated load of 85 kN, while For damage distribution purposes, photogrammetry was used by
CA2 at a load of 70 kN. The damage progresses concentrated in a measuring the tensile strains at the most tensioned edge to estimate
central crack in the case of CA2 and in two main central cracks for CA1 curvatures and a plastic hinge length (Dias-da Costa et al., 2014). For
(bending and diagonal tension), without secondary cracks (Fig. 13). that, the integral of the strain (ε) distribution at a particular beam de-
Post-peak degradation is less pronounced than in the case of slender flection δ (considering the small compressed zone or the almost con-
specimens (Fig. 16). stant neutral axis depth in the beam length, this value is consistent with
the curvature distribution) normalized by the maximum tensile strain
(εmax) obtained by photogrammetry, was defined as plastic hinge length
3.5.6. Specimens CP1 and CP2
(lp = ∫ εdx/εmax). This value is again normalized by the beam height
Specimens CP1 and CP2 presented a behavior similar to CA1 and
(lp/Hf) for convenience and familiarity. The results, summarized in
CA2. Specimen CP1 showed cracking at an estimated load of 75 kN,
Table 4, indicate that the damage in slender specimens is much more
while CP2 at a load of 90 kN. The damage progresses concentrated in a
concentrated on specimens without reticulated frame, and within the
central crack (bending and diagonal tension), without secondary cracks
framed specimens, polypropylene (DP) specimens distribute damage
(Fig. 14). Post-peak degradation is less pronounced than in the case of
(i.e., have a higher plastic hinge length lp) much more than the other
slender specimens (Fig. 16).
(comparable) solutions.

3.5.7. Load-displacement responses

The overall load versus displacement (central) response is shown for 4. Analytical studies
all cases in Figs. 15 and 16. In general terms, for the slender specimens
(Fig. 15), it can be seen that the reticulated frame has a very important This chapter presents an analytical study focused on quantifying
influence on both the specimen strength and ductility, which can be forces and demands on the lining of the tunnel under study.

Fig. 11. Failure mode of specimens DA1 and DA2 – (a) Fiber adherence failure, and (b) cracking in specimen DA2.

19
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

Fig. 12. Failure mode of specimens DP1 and DP2 – (a) tensile rupture of the polypropylene fibers, and (b) cracking of specimen DP1.

4.1. Experimental test models The soil is modeled with bi-quadratic quadrilateral elements of 9
nodes and a nonlinear material model PDMY02 (Yang and Elgamal,
All specimens were modeled in OpenSees (Mazzoni et al., 2006), an 2000). The lining is modeled with beam-column elements of fiber
Open Source object-oriented platform developed at UC Berkeley to si- sections with non-linear uniaxial material models, calibrating the steels
mulate the response of structural and geotechnical systems, with an and concrete properties from previous analysis that are validated with
extensive library of research-focused elements and material models. the scaled bending experiments as well as the uniaxial tests. Finally, the
The models used in this study are based on a fiber section (bending- soil-structure interaction was modeled with rigid springs in compres-
axial) model composed by steel and concrete longitudinal fibers (cross- sion and with zero stress in tension for the perpendicular direction, and
section is discretized accounting for concrete and steel components). in the tangential direction with a concrete-soil friction model of perfect
For steel, the uniaxial model by Chang and Mander (1994) was selected adhesion. The nonlinearity associated with the materials (soil, concrete
(Fig. 17c), whereas, for concrete in compression the model by Chang and steels) has been modeled, but not geometric nonlinearity. The final
and Mander (1994) was also considered (Fig. 17a). A fiber model model has approximately 8200 quadrilateral elements (32,000 nodes),
(Taucer et al., 1991) of the EN14651 test was performed using dis- 250 beam-column elements (800 nodes) and 250 contact elements (800
tributed plasticity elements (nonlinear behavior of steel and concrete nodes). The mesh chosen to perform all analyses (Fig. 18) has good
fibers are incorporated in longitudinal elements with 3 degrees per discretization for the number of nodes, not excessive elements and
node – 2D case – that follow the Bernoulli hypothesis) to determine the follows general guidelines from state of the art meshing techniques
appropriate parameters for modeling the tensile response of concrete (Beer, 2012; Potts, 2002), with special attention to element regularity
(Fig. 17b). and avoidance of boundary effects (such as a large domain and ap-
Also, the scaled bending tests (beams with axial load) were modeled propriate boundary conditions).
with fiber sections based on the constitutive laws derived for material
characterization tests. Figs. 15 and 16 show the experimental results
compared with the numerical simulations. As it can be seen, the models 4.2.1. Elements and materials
are able to simulate the general response (stiffness, strength, degrada- For the gravel considered as part of the soil-structure model,
tion, failure of the reinforcement) reasonably well. Considering that PDMY02, a multi-yield material having a failure surface of the Drucker-
tunnel tolerances are not tight and that the actual thickness varies Prager type (Yang and Elgamal, 2000) with a hyperbolic stress-strain
greatly due to the nature of the work, the modelled behavior is good relationship is used, which is capable of stress redistribution. For pur-
enough to capture the general flexural response of the section (capacity, poses of calibration the parameters of the monotonic response (com-
stiffness, failure of the rebar, etc.) with sufficient accuracy for the pression), values are fitted with model simulations of triaxle tests from
purpose of modelling the tunnel lining for all specimens. the literature for the soil under consideration (De la Hoz Alvarez, 2007;
Kort et al., 1979). These values are shown in Table 5, and correspond to
a 52° internal friction angle, a depth dependent (Z) elastic modulus
4.2. Tunnel finite element model from Kort et al. (1979), and a Poisson coefficient of 0.25. The lining was
modeled with non-linear fiber-based beam-column elements (material
The finite element model of the tunnel consists of a two-dimensional and geometrical nonlinearity, although the latter has little impact). The
model of the cross-section of the tunnel-soil system, which will be use of these elements, which do not model the shear response, can be
subjected to an ovalling analysis, that is, a pseudo-static desangulation, justified noting that at the area under interest (vault area), the slen-
such that with a simple model of the transversal seismic response of the derness of lining is controlled by flexural response, and according to the
tunnel-soil system one can establish the demands on the lining of the experimental tests on beams, for the geometry and levels of axial load
tunnel. considered, the model captures the overall response (Figs. 15 and 16).

Fig. 13. Growing of the principal crack of specimen CA1 – (a) δ = 5.9 mm, (b) δ = 12.1 mm and (c) δ = 15.1 mm.

20
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

Fig. 14. Specimens CP1 and CP2 – (a) CP1 at δ = 13.1 mm and (b) CP2 at δ = 9.9 mm.

300
(a)
Table 3
250
Reinforcement and loading condition of specimens.
Specimen Reticulated Tensile Support N (kN) N/Agf′c 200

Load (kN)
frame? reinforcement distance (m) (%)
150
A1 No Φ4.2 @ 100 mesh 1.3 60 2%
100
A2 No Φ4.2 @ 100 mesh 1.3 200 7% Test CA1 Model CA1
B1 Yes Φ4.2 @ 100 mesh 1.3 60 2% Test CP1 Model CP1
50
B2 Yes Φ4.2 @ 100 mesh 1.3 200 7%
DA1 Yes Steel fibres 1.3 60 2% 0
DA2 Yes Steel fibres 1.3 200 7%
0 2 4 6 8 10 12 14
DP1 Yes Polypropylene 1.3 60 2%
fibres Displacement (mm)
DP2 Yes Polypropylene 1.3 200 7%
fibres
CA1 Yes Steel fibres 0.4 60 2%
(b) 300
CA2 Yes Steel fibres 0.5 200 7% 250
CP1 Yes Polypropylene 0.4 60 2%
fibres 200
Load (kN)

CP2 Yes Polypropylene 0.5 200 7%

fibres 150

100
Test CA2 Model CA2
(a) 80 50 Test CP2 Model CP2

0
60
0 2 4 6 8 10 12 14
Load (kN)

Displacement (mm)
40
Fig. 16. Transversal load versus transversal displacement test and model re-
Test A1 Model A1
sponse for short specimens – (a) specimens with low axial load and (b) speci-
Test B1 Model B1
20 mens with high axial load.
Test DA1 Model DA1
Test DP1 Model DP1
0 Table 4
0 10 20 30 40 50 Measured plastic hinge length.
Displacement (mm) Specimen Estimated lp (mm) lp/Hf Measured δ (mm)

(b) 80 A1 125 0.83 10

A2 128 0.85 38
B1 171 1.14 54
60 B2 195 1.3 42
DA1 164 1.09 44
Load (kN)

DA2 161 1.07 47

40 DP1 263 1.75 43
DP2 247 1.64 50
Test A2 Model A2 CA1 154 1.02 15
20 Test B2 Model B2 CA2 129 0.86 14
Test DA2 Model DA2 CP1 113 0.75 14
Test DP2 Model DP2 CP2 162 1.08 14
0
0 10 20 30 40 50
Displacement (mm) In order to impose an oval deformation on the system, auxiliary
elements were used that are not part of the conceptual model, but allow
Fig. 15. Transversal load versus transversal displacement test and model re-
sponse for slender specimens – (a) specimens with low axial load and (b) spe-
a simpler analysis. Rigid bars were placed on the side edges of the floor
cimens with high axial load. grid, connected between them and the adjacent floor by horizontal
(non-vertical) rigid springs, allowing the bars to rotate around their

21
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

(a) (b) (c)

70 3,5 900
Plain
60 3 800
Steel Įber
PlasƟc Įber 700
50 2,5
Stress (MPa)

Stress (MPa)

Stress (MPa)
600
40 2 500 AT50-56H
30 1,5 400 A630S
20 Plain 300
1
Steel Įber 200
10 PlasƟc Įber 0,5 100
0 0 0
0 0,002 0,004 0,006 0,008 0,01 0 0,0005 0,001 0,0015 0,002 0,0025 0 0,02 0,04 0,06 0,08 0,1
Strain Strain Strain

Fig. 17. Material constitutive laws – (a) concrete in compression, (b) concrete in tension, and (c) steel.

respective supports (Fig. 19). Considering that in the analysis a shear Table 5
deformation (desangulation) will be imposed on the model, the lateral Soil-tunnel model parameters.
displacement of degrees of freedom of the left and right elements are Parameter Value Source
constraint (connected) at the top and along the edge. For the definition
of the cross-section, the scheme of Fig. 1a will be used (since small E, Elastic modulus E = 46 000 (Z (m))^(0.55) From Kort et al. (1979)
differences were observed for fibers with the presence of the reticulated (kPa) if Z < 6 m
E = 54 000 (Z (m))^(0.53)
frame). Because the section is constructed in a sequential manner (the
(kPa) if Z > 6 m
support is first placed with the reticulated frame, and then the lining), E > 100 000 (kPa)
the tension state of the section should be estimated considering this Poisson's Ratio 0.25 From Kort et al. (1979)
aspect, assuming that the support is loaded first, then the lining con- Internal friction 52° From Kort et al. (1979) and
angle simulations
nected, and finally the entire section is loaded.

4.2.2. Loads and model sequence

On a tunnel in the ground there are essentially two loads of interest
in the design: (i) the thrust of the soil resulting from gravity; and (ii) the
thrust of the soil caused by an earthquake. The stresses generated by
loads acting on the soil surface are typically negligible for most cases.
The modeled load sequence considers a series of stages: (1) appli-
cation of gravity to the soil not affected by the excavation. These loads
are modeled as body forces on each element of soil without excavating;
(2) generation of the excavation (opening) in the soil model without
support. In the 2D model, this is done by partially flexibilizing the in-
ternal material of the excavation by a factor of α = 0.5 (Alpha Method –
Möller, 2006). This generates redistribution of body forces; (3) place-
ment of the support (primary lining) in the finite element model and
Fig. 19. General tunnel model scheme.
new flexibilization of the internal material by a factor of α = 0.1, which
simulates the advance of the excavation. This new redistribution
translates into support loads; (4) placement of the coating (secondary load as a desangulation to the soil.
lining) in the finite element model, and a new and final flexibilization
of the internal material by a factor of α = 0.01, which simulates the 4.2.3. Gravitational and seismic loads (Oval)
possible relaxation over time of the surrounding material. In this way, In Chile, the value of γ = 0.035% for gravel has been used as de-
this new redistribution allows loading of the secondary lining; and fi- sangulation. These values come from the highway design code (Chilean
nally, (5) application of the seismic model, which in this case is a lateral Ministry of Public Works, 2000), which is based on the proposal by

Fig. 18. Soil-tunnel mesh – (a) overall domain and (b) local tunnel diagram.

22
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

Fig. 20. Desangulation interaction factors – (a) desangulation scheme, and (b) heat map of nodal horizontal displacements.

Kuesel (1969). This proposal assumes, as do most of seismic design

methods for underground structures (Hashash et al., 2001), that the
design-controlling load is the product of vertical shear waves that travel
through the tunnel during the earthquake, and other seismic waves are
considered negligible.
It is important to recognize that the model will be deformed on the side
frame, such that the desangulation of the ground in the center (tunnel
location) is not necessarily the same observed on the side. In addition, due
to the kinematic interaction between the tunnel and the lining, the de-
sangulation within the tunnel is not the same as the desangulation of the
ground in the same location. Consequently, there are three desangulation
magnitudes in the model: (i) γs, desangulation of the ground (soil) on the
side; (ii) γc, desangulation of the soil at the center (tunnel location); and
(iii) γl, desangulation of the lining. The description of the desangulation
magnitudes is shown in Fig. 20a. A heat map with the actual kinematics of
the complete system is shown in Fig. 20b for a value of γc = 0.035%. The
relationship results obtained for the pushover analysis between these de-
sangulation magnitudes is shown in Fig. 21. Thus, the interaction factor R
for γc = 0.035% is 1.76 (Fig. 21b), implying that the desangulation in the
lining would be γl = 0.062%.
Following the positive moment and forces convetion shown in Fig. 22. Convention for positive internal forces and moments.
Fig. 22, Fig. 23 shows the internal forces and moments obtained in the
support and lining for different desangulation values (γl). After ex- of the first lining (support) has no significant final effect on the com-
amining the sectional strains and forces it can be seen that both the plete capacity of the section, because the preload compressive stresses
primary and secondary lining remain in the linear range and therefore are low. It is also worth mentioning that even using lower values for α
the curvature values are obtained linearly from moment-curvature re- the conclusion is maintained, because the surrounding soil is very stiff
lationship (M-ϕ). There is a concentration of moments (and curvature) and the disequilibrium produced by the excavation is taken by the soil
in the lower corner of the tunnel, which is not shown since the analysis arch effect rather than the lining effect.
is focused on the support and lining that cover the upper section of the Fig. 25 shows the interaction diagram M-N that reveals the moment
tunnel (dome). The increased section and geometry of the lower section versus axial capacity for the entire section, with and without the frame
of the tunnel should be studied and modeled locally. This is consistent contribution. The solicitation is obtained from Fig. 24, for a γl = 0.09%
with experience in rectangular tunnels, where corners are typically desangulation, which could be interpreted as an upper limit for the
designed for yielding, while the rest of the structure remains linear. local design. In terms of capacity, the safety factors associated with
Fig. 24 shows the sum of the loads between the lining and the loads results in very conservative design, independent of the axial load,
support. Accordingly, and recalling Fig. 23, the demands on the support when the reticulated frame is considered as structural component.
and the lining are practically the same. The above shows the preloading

(a) (b) (c)

3 1.8 0.65

2.8 0.63
1.7
2.6 0.61
ɶs/ɶc

ɶl/ɶs
ɶl/ɶc

1.6
2.4 0.59
1.5
2.2 0.57

2 1.4 0.55
0 0.05 0.1 0.15 0.2 0 0.02 0.04 0.06 0.08 0.1 0 0.05 0.1 0.15 0.2 0.25
ɶs ɶc ɶs
Fig. 21. Desangulation interaction factors – (a) γs/γc, (b) γl/γc, (c) γl/γs.

23
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

0 12
25 75 125 175 225 ɶс0.012%
-100 ɶс0.024%
8
ɶс0.043%
-200
ɶс0.06%

M (kN m)
4
N (kN)
ɶс0.012% ɶс0.09%
-300
ɶс0.024%
0
-400 ɶс0.043% 25 75 125 175 225
ɶс0.06%
-500 -4
ɶс0.09%
(a) (b)
-600 -8
Theta (°) Theta (°)

100 8
ɶс0.012%
0 6
ɶс0.024%
25 75 125 175 225
-100 4 ɶс0.043%

M (kN m)
ɶс0.06%
N (kN)

-200 ɶс0.012%
2 ɶс0.09%
ɶс0.024%
-300 ɶс0.043% 0
ɶс0.06% 25 75 125 175 225
-400 -2
ɶс0.09% (c) (d)
-500 -4
Theta (°) Theta (°)

Fig. 23. Dome forces – (a) axial force in the support, (b) moment in the support, (c) axial force in the lining, and (d) moment in the lining.

200 20
(a) ɶс0.012% (b)
0 15 ɶс0.024%
25 75 125 175 225 10 ɶс0.043%
-200
ɶс0.06%
M (kN m)

ɶс0.012% 5
N (kN)

-400 ɶс0.09%
ɶс0.024%
0
-600 ɶс0.043% 25 75 125 175 225
-5
ɶс0.06%
-800 ɶс0.09% -10

-1000 -15
Theta (°) Theta (°)
Fig. 24. Dome forces in the entire section – (a) axial force, and (b) moment.

1000 5. Conclusions

800 In this study, the use of steel and polypropylene fibers as replace-
SolicitaƟon
ment of welded mesh reinforcement for lining in tunnels has been in-
600
vestigated. A series of tests to characterize the material, as well, as
capaciy with beam tests with axial load are carried out to mimic the gravitational
P (kN)

frame 400
capaciy and seismic actions in the lining of the tunnel. Additionally, analytical
without frame models are constructed that reproduce the test results and are used to
200
model the tunnel lining in a 2D finite element model of a ground-tunnel
model to predict the seismic response under a desangulation given by
0
-250 -200 -150 -100 -50 0 50 100 150 200
vertical shear waves.
The concrete tensile characterization using test of prims under
-200
M (kN m)
bending allows to adequately predicting the response of beams.
However, the direct tensile tests of concrete cylinders with and without
Fig. 25. Pn-Mn interaction diagram. fibers show a high scatter, resulting in especially low strength values for
the cases with fiber, with the case of steel fiber showing the lowest
When the frame is not considered, the capacity is largely reduced, re- value. It is observed that there is a failure plane where the fibers are
sulting in less conservatism, especially when the axial load is modest, oriented parallel to the section, which reduces their capacity.
that is, for shallow tunnels. Regarding the beam with axial load response, it has been found that

24
L.M. Massone, F. Nazar Tunnelling and Underground Space Technology 77 (2018) 13–25

replacement of the steel mesh by steel fibers, under the conditions of Chang, G., Mander, J.B., 1994. Seismic energy based fatigue damage analysis of bridge
the tests carried out, and including the reticulated frame, delivers sec- columns: part 1-evaluation of seismic capacity. Tech. rep. MCEER, University of
Buffalo.
tional capacities approximately equal to the solution with steel mesh, Chiaia, B., Fantilli, A.P., Vallini, P., 2009. Combining fiber-reinforced concrete with
and presents similar ductility indicators. For the case of polypropylene traditional reinforcement in tunnel linings. Eng. Struct. 31 (7), 1600–1606.
fibers, for slender specimens, the solution presents sectional capacities Chilean Ministry of Public Works, 2000. Highway Design Manual.
De la Fuente, A., Pujadas, P., Blanco, A., Aguado, A., 2012. Experiences in Barcelona with
approximately 20% higher than the steel mesh solution, with a lower the use of fibres in segmental linings. Tunnell. Underground Space Technol. 27 (1),
ductility, but a larger distribution of damage. Also, it was found that the 60–71.
reticulated steel frame presents the most important contribution in De la Hoz Alvarez, K., 2007. Estimation of shear resistance parameters in coarse granular
soils (in Spanish). Master thesis, University of Chile.
flexure, in both strength and ductility. The use of non-ductile steel Dias-da Costa, D., Valença, J., do Carmo, R., 2014. Curvature assessment of reinforced
mesh, which presents fracture around 1.5% strain, results in low de- concrete beams using photogrammetric techniques. Mater. Struct. 47 (10),
formation capacities in the beam with axial load tests. This is observed 1745–1760.
EN-14651, 2007. Test method for metallic fibre concrete e measuring the flexural tensile
in the specimens without reticulated frame. Short beam specimens
strength (limit of proportionality (LOP), residual), European Committee for
show little differences for both fiber solutions. The overall response of Standardization.
all beam specimens with axial load was well captured by flexural Hashash, Y.M., Hook, J.J., Schmidt, B., Yao, J.I.-C., 2001. Seismic design and analysis of
models. underground structures. Tunnell. Underground Space Technol. 16 (4), 247–293.
Kang, T.H.-K., Kim, W., Massone, L.M., Galleguillos, T.A., 2012. Shear-flexure coupling
The tunnel model, with the defined properties based on the carried behavior of steel fiber-reinforced concrete beams. ACI Struct. J. 109 (4), 435–444.
out tests, indicates that for the solicitations estimated for the soil gravel Kolymbas, D., 2005. Tunnelling and tunnel mechanics: A rational approach to tunnelling.
in Santiago, the response remains in a low range of moment and axial Springer Science and Business Media.
Kort, I., Musante, H., Fahrenkrog, C., 1979. In situ mechanical properties measurements
load with a practically linear behavior in the seismic phase. of gravelly soil used in an interaction and foundation model for the Santiago Metro.
In: Proceedings of the 6th Panamerican Conference on Soil Mechanics and
Acknowledgements Foundation Engineering, Peru, pp. 217–224.
Kuesel, T.R., 1969. Earthquake design criteria for subways. J. Struct. Div. 95 (6),
1213–1231.
The authors would like to acknowledge the support and aid from the Mazzoni, S., McKenna, F., Scott, M.H., Fenves, G.L., 2006. OpenSees command language
Civil Engineering department at Universidad de Chile, where the testing manual. Pacific Earthquake Engineering Research (PEER) Center.
Möller, S. C., 2006. Tunnel induced settlements and structural forces in linings. Ph.D.
was done. They would also like to thank Metro de Santiago, and in Thesis. Univ. Stuttgart, Inst. f. Geotechnik.
particular to Mr. Edgardo González, for their cooperation and funding Nazar, F., 2016. Analytical and experimental evaluation of the use of polypropylene and
provided to build the specimens. steel fibers as reinforcement in shotcrete supports for tunnel interstation of Santiago
(in Spanish). Civil engineering thesis. University of Chile.
Nežerka, V., Antoš, J., Sajdlová, T., Tesárek, P., 2016. Use of open source DIC tools for
Appendix A. Supplementary material analysis of multiple cracking in fiber-reinforced concrete. Appl. Mech. Mater. 827,
336–339.
Supplementary data associated with this article can be found, in the Potts, D., 2002. Guidelines for the use of advanced numerical analysis. Thomas Telford.
Taucer, F., Spacone, E., Filippou, F.C., 1991. A fiber beam-column element for seismic
online version, at https://doi.org/10.1016/j.tust.2018.03.027. response analysis of reinforced concrete structures, vol. 91-17. Earthquake
Engineering Research Center, College of Engineering, University of California
References Berkeley, California.
Wallace, J.W., Elwood, K.J., Massone, L.M., 2008. Investigation of the axial load capacity
for lightly reinforced wall piers. J. Struct. Eng.-ASCE 134 (9), 1548–1557.
Beer, G., 2012. Numerical simulation in tunnelling. Springer Science & Business Media. Wetzig, V., Weiss, R., di Prisco, M., Felicetti, R., Plizzari, G., 2004. Fibre reinforced
Belletti, B., Cerioni, R., Meda, A., Plizzari, G., 2004. Experimental and numerical analyses shotcrete for long tunnel projects in Switzerland. In: 6th International RILEM
of FRC slabs on grade. In: Proceedings of FRAMCOS5 Conference, Vail Colorado, pp. Symposium on Fibre Reinforced Concretes, RILEM Publications SARL, pp. 545–552.
973–980. Yang, Z., Elgamal, A.-W. M., 2000. Numerical modeling of earthquake site response in-
Bentur, A., Mindess, S., 2006. Fibre Reinforced Cementitious Composites, second ed. cluding dilation and liquefaction. Report (Structural Systems Research Project),
Taylor & Francis, Modern Concrete Technology. SSRP-2000/01. University of California at San Diego, Dept. of Structural Engineering.
Blaber, J., Adair, B., Antoniou, A., 2015. Ncorr: Open-Source 2D Digital Image
Correlation Matlab Software. Exp. Mech. 55 (6), 1105–1122.

25