Accepted Manuscript

Prediction of the seismic behavior of an underground railway station and a tunnel

in Napoli (Italy)

S. Fabozzi, V. Licata, S. Autuori, E. Bilotta, G. Russo, F. Silvestri

PII: S2467-9674(16)30041-1
DOI: http://dx.doi.org/10.1016/j.undsp.2017.03.005
Reference: UNDSP 16

To appear in: Underground Space

Received Date: 31 December 2016

Revised Date: 14 March 2017
Accepted Date: 15 March 2017

Please cite this article as: S. Fabozzi, V. Licata, S. Autuori, E. Bilotta, G. Russo, F. Silvestri, Prediction of the seismic
behavior of an underground railway station and a tunnel in Napoli (Italy), Underground Space (2017), doi: http://
dx.doi.org/10.1016/j.undsp.2017.03.005

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Prediction of the seismic behavior of an underground railway station and a
tunnel in Napoli (Italy)
S. Fabozzi, V. Licata, S. Autuori, E. Bilotta, G. Russo, F. Silvestri (corresponding author)
University of Napoli Federico II, Napoli, Italy

Abstract

The assessment of the seismic safety of underground structures, either tunnels or large station
boxes, should not be overlooked especially in densely populated areas, even with low to moderate
seismicity.
For underground structures, an important issue is the estimation of the seismic actions acting on the
structure; only few experimental evidences are available for multi-level propped walls. For tunnels,
it is generally assumed that their seismic behavior in soft ground is governed by the surrounding
soil, while the inertial load contribution of the underground structure itself is negligible. In both
cases, recent numerical studies proved that advanced dynamic analyses can provide satisfactory
interpretation of non-linear soil-structure interaction during earthquakes.
In this paper, a real case study, represented by a large open multi-propped excavation and a circular
segmented tunnel in a densely urbanized area of the city center in Napoli, has been used to
investigate some of the mentioned aspects.
Accurate geotechnical characterization and choice of the reference input motions lead to a first
estimate of the free-field ground motion, which was subsequently used for pseudo-static decoupled
analyses.
For the complexity of both excavation geometry and staged construction, a full dynamic analysis
was considered neither affordable nor reliable for the multi-propped station box; thus two
conventional pseudo-static analyses, applying either a displacement-based or a force-based
approach, were carried out.
In the case of the tunnel, the seismic increments of internal forces in the lining could be calculated
through both a simplified pseudo-static analysis and a full dynamic analysis, showing a satisfying
agreement.
Overall, the results of the study demonstrated that the seismic increments of internal forces in the
diaphragm walls of the station and in the segmented lining of the tunnel were quite significant. The
case study encourages improving the reliability of simplified methods based on the more advanced
dynamic approaches.

Keywords: underground structures; tunnels; diaphragm walls; seismic behavior; pseudo-static

analysis; dynamic analysis; soil-structure interaction
1. Introduction

The behavior of tunnels or large underground excavations in urban areas under seismic actions is
rather complex and until two decades ago it was generally not considered as a major design issue.
Only recently, experimental and numerical research has made some significant steps towards the
comprehension of the mechanisms governing soil-structure interaction for embedded structures
under seismic actions.
As far as large underground infrastructures are concerned, there are only a few cases where such
structures were heavily damaged. It is worth mentioning the impressive case of Dakai Station, along
the Kobe Rapid Railway line, that collapsed during the Kobe Earthquake (January, 17, 1995). Iida
et al. (1996) and Yoshida (1999) described the damage suffered by this station: the large span
ceiling collapsed just above the platforms, as some supporting columns buckled, bringing down part
of road running above it. The analysis of the case study showed that on one hand the relative
displacement between the bottom and top slabs caused a large horizontal shear force on the central
column, on the other hand the overburden soil mass also affected the response, adding inertial
forces to the structure.
The main problem in the seismic assessment of such underground structures is the estimation of
the earth pressure acting on the structure during an earthquake, that is influenced by many factors,
among which the relative soil-structure flexibility that is often neglected in design.
As a matter of fact, only few experimental evidences are available for flexible retaining
structures, among which case studies on physical models. They are generally limited to cantilevered
walls or diaphragm walls with a single level of prop at the top (Conti et al., 2012), while actual
excavations in urban environment often include multi-level propped walls. In the latter case, it has
been demonstrated (Pitilakis & Tsinidis, 2014) that the earth pressure calculated for rigid walls is
only an upper limit of the dynamic earth pressure, while stress concentration at the slab levels
cannot be captured, without a proper dynamic soil-structure interaction analysis. Dynamic analyses
on multi-propped diaphragm walls are generally limited to a couple of propping levels (e.g.
Tropeano & Soccodato, 2014), i.e. at the wall head and at the bottom of the excavation, and show a
strong dependency of the seismic behaviour on the simulation of the construction stages as well as
on the characters of the input motions (Soccodato & Tropeano, 2015).
As far as tunnels are concerned, it is generally assumed that their seismic behavior in soft ground
is governed by the surrounding ground, while the inertial load contribution of the underground
structure itself is negligible (Hashash et al., 2001). Hence changes of internal forces in the lining
due to seismic shaking are generally calculated from the transient response of the ground. The
tunnel lining structure undergoes deformation both in the transverse section (Hashash et al., 2005)
and in longitudinal direction (St.John & Zahrah, 1987; Kawashima 1999). The variation of internal
forces induced in the tunnel lining during earthquakes can be calculated following several
approaches that model in different ways the soil-structure interaction (Hashash et al. 2001, Pitilakis
and Tsinidis, 2014).
Although existing guidelines suggest pseudo-static or uncoupled dynamic analyses, usually
adopted for a preliminary stage of design, it has been shown (Bilotta et al., 2014; Fabozzi & Bilotta,
2016) that full dynamic analyses must be performed of the soil-structure interaction to take into
account the influence of the pre-existing stress state around the tunnel and of permanent ground
deformation during shaking. Indeed, assessment of numerical predictions against centrifuge model
tests proved that such advanced analyses can provide satisfactory interpretation of non-linear soil-
structure interaction during earthquakes (Bilotta et al., 2014).
For both types of underground structures, this paper wishes to analyse from the staged
construction to a design earthquake a real case of a metro tunnel and a multi-level propped station
embedded in the subsoil of the city of Napoli. The study is focused on the effects of ground shaking
on the seismic increment of internal forces in the underground structures with respect to those
acting in static conditions; the effects in terms of transient and permanent deformation induced on
the surrounding subsoil and buildings were not considered in this stage of the research.
2. The case study

During last three decades, the metropolitan area of Napoli underwent an intense development of
construction and expansion of two underground railways (Figure 1a). Metro Line 1 connects the
inner and hilly part of the city with the historical centre and the Central railway station; Metro Line
6 starts from the western district of Fuorigrotta and terminates in the city center, close to the famous
Maschio Angioino castle in the Municipio square, where the two lines are inter-connected.

(a)

(b)

(c)

Figure 1. Metro lines 1 and 6 superimposed to a schematic geological map of Napoli (a); stratigraphic profile and
groundwater conditions for (b) Line 1 lower stretch and (c) Line 6 western stretch (after Russo et al., 2012)
Uphill, most of the twin-tunnel stretches of Line 1 develop within the Neapolitan Yellow Tuff
formation, dominating the subsoil of the urban area; on the other hand, close to the coastline both
lines (i.e. the lower stretch of Line 1 and the western stretch of Line 6) are mainly excavated
through marine, alluvial and volcanic soils overlying the tuff, and often underwater (Figure 1b,c).
San Pasquale station is one of the new large stations of Line 6, located a few tens of meters from
the shoreline. It occupies an area of approximately 2000 m2, quite close to a series of residential
buildings. Figure 2 shows the plan view (a) and the two main cross sections (b, c) of the station; in
both sections the location of the tunnel with respect to the diaphragm walls of the station is
illustrated.
The main body of the station has a rectangular shape in plan of 85.524.1 m; the maximum
ground excavation depth is about 27 m, 26 m of which under the water table. The excavation is
supported by large T-section diaphragm walls made by reinforced concrete. The depth of the
reinforced concrete panels is about 50 m, in order to reach the underlying bedrock constituted by
the Neapolitan Yellow Tuff. As shown by the plan in Figure 2a, two minor shafts for the subway
entrance are adjacent to the long sides of the main station chamber; they are about 12 m deep and
are supported by reinforced concrete rectangular panels, with a thickness of 1 m and a total length
of about 22 m. Such peripheral diaphragm walls were installed before the tunnel excavation crossed
the area of the station, thereafter the construction of the inner panels took place.
In the plan in Figure 2a, two rectangular chambers located on the opposite short sides of the
station are also sketched. They were built inside the diaphragm walls and the soil inside was treated
with jet grouting, in order to allow a safe entrance of the tunneling boring machine (TBM) into the
main shaft of the station after the preliminary demolition of the diaphragm panels.
The total time needed to reach the designed depth of the excavation and to build the underground
structures was approximately 5 years and 10 months.
The underground excavation of the tunnel was carried out below the groundwater table, using an
earth pressure balance (EPB) tunnelling boring machine. The lining has an external excavation
diameter of 8.150 m, with an internal and external radius of 3.625 m and 3.925 m, respectively, for
a total thickness of 0.3 m. The annular void between the extrados and the external excavation
diameter is backfilled with a two-component grout layer 0.15 m thick.
The tunnel lining system is made of pre-cast reinforced concrete segments. Each ring is 1.7 m
wide, assembled in nine precast C35/45 concrete elements: eight segments with trapezoidal shape,
corresponding to a central angle of 41.5, plus the keystone. The segments are jointed in both
longitudinal and circumferential directions with unbolted flat joints.

3. Geotechnical characterization and subsoil model

The geotechnical investigations carried out during the different stages of design and construction
of the tunnel and the station aimed at characterizing the bedrock morphology, the groundwater level
and the geotechnical properties of the shallow deposits relevant for soil-structure interaction
(Lamante et al., 2012). The investigation spots around the station are reported in Figure 2a. In
detail, two boreholes with continuous coring (S1 and S2), respectively 42 m and 49 m deep, were
located at two opposite corners of the station shaft, while the borehole drillings, SG1-SG4 and SG5-
SG8, 28 m to 40 m deep, were evenly distributed along the Northern and Southern sides,
respectively.
The layering obtained from the boreholes is drawn in Figures 2b-c together with the simplified
sections of the station and the tunnel. Like in most of the seaside areas of Napoli, the subsoil is
characterized by alternating layers of volcanic and seashore deposits, lying above a sub-horizontal
tufaceous bedrock. Starting from the ground surface, below a negligible thickness of man-made
ground cover, about 17 m of seashore sands (SS) lie upon 27 m of volcanic products, divided into
24 m of pyroclastic silty sand (Pyr) and 3 m of the fractured facies of Neapolitan Yellow Tuff
(Tuff1). The top of the bedrock (Tuff2) was intercepted at the depth of approximately 44 m.
 (a)

Buildings

PZ5
PZ16 PZ15
CPT
PZ6
PZ4
CH SG1 SG2
SG3
SDMT S1 CPT1 P74 SG4
CH
CPT2
N
Piezometer 0 40m
Borehole with
SG5
P13
continous coring + SPT SG7 SG8
SG6
Borehole drilling + SPT PZ1 S2

PZ3 SDMT2 SDMT1

PZ2

(b) (c)
0

SS Jet grouting SS

20m

Pyr Pyr

Tuff1 Tuff1

Tuff2 Tuff2
Tuff2
85.5 m 24.1m

Figure 2. (a) Schematic plan of the station area with in situ investigations; simplified sections of the station
shaft and tunnel along the (b) longitudinal and (c) transversal directions.

The different nature of the marine and volcanic materials is reflected by their specific gravity,
Gs, and saturated unit weight, sat, measured on undisturbed samples, reported in Table I. The
mechanical parameters adopted for each layer in the numerical analyses will be later shown in
Section 4.

Table I. Layering and mean physical properties of the subsoil

Layer z H Gs sat
(m) (m) (kN/m3)
Seashore sands (SS) 0-17 17 2.6 18
Pyroclastic silty sand (Pyr) 17-41 24 2.5 16
Fractured tuff (Tuff1) 41-44 3 2.5 16
Intact tuff (Tuff2) >44 - 2.5 16

The installation of a significant number of piezometers (PZ in Figure 2a) revealed neither
appreciable variations with the location nor marked oscillations with time of the groundwater level,
which remained stable within 1 m below the ground level before, during and after the excavations.
Standard penetration tests (SPT) carried out in all the boreholes (Figure 3b), as well as two cone
penetration tests (CPT) along the northern side (Figure 3c), reached depths down to 35m in the
pyroclastic soil formation. In-depth measurements of shear wave velocity, V S, were undertaken
down to 45 m in the tuff bedrock by a cross-hole test (CH) carried out along the northern side
(circles in Figure 3d). Two seismic dilatometer tests (SDMT), located along the opposite side of the
excavation area, yielded comparable measurements of VS, although limited to 25m depth (triangles
in Figure 3d).
The profiles of the number of blow counts, NSPT, and cone tip resistance, qc, reported in Figures
3b-c, show similar irregular variations of penetration resistance without any particular trend with
depth. Notwithstanding the apparent soil heterogeneity, a good agreement can be observed between
the results of the two CPT tests: on the average, the cone penetration resistance increases in the first
10 m of the marine sands. Thereafter, in the pyroclastic layers its variability is related mostly to the
change of grading than to the increasing confining pressure with depth (Figure 3c). The individual
NSPT profiles (Figure 3b) are consistent with this picture, although with lower resolution. This
evidence, which is recurrent in the cohesionless volcanic deposits in Napoli, may be related to the
breakage of soil particles under compression, leading to underestimation of shear strength if based
on the penetration resistance (Rippa & Vinale, 1983).
The values of shear wave velocity, VS, measured by Cross Hole and Seismic Dilatometer, are
plotted in Figure 3d. As above anticipated, a quite good agreement can be noted among the different
profiles, even if in the first 25 m the VS values from the SDMTs are, on the average, slightly higher
than those obtained by the CH test. The little difference may be referred to opposite disturbance
effects due to the different test procedures: in fact, with SDMT, the soil is laterally displaced (and
compressed) by the dilatometer, while in the CH test the soil can undergo a stress release following
the drilling. All in all, this latter test was considered as more reliable and conservative. The values
of shear wave velocity are nearly constant and equal to approximately 200 m/s in the marine sand
layer, thereafter they gradually increase up to 400 m/s in the topmost 5 m of the pyroclastic layer.
The VS profile resulting from the CH tests keeps almost constant down to the top of the Tuff1,
reaching values of 700m/s, and rapidly, once again, increases to 1250 m/s in the underlying
bedrock (Tuff2). The mean VS profile, reported in Figure 3d with thick solid line, was adopted in
the dynamic analyses.

(a) (b) (c) (d)

NSPT qc (MPa) Vs (m/s)
0 20 40 60 0 20 40 0 500 1000 1500
0 0 0

CH
SDMT1
-10 -10 -10 SDMT2
SS Vs-ave

-20 -20 -20

z (m)

-30 -30 -30

Pyr

-40 -40 -40

Tuff1 S1 S2 SG1 SG2 CPT1
SG3 SG4 SG5 SG6 CPT2
Tuff2 SG7 SG8
-50 -50 -50
Figure 3. Soil layering (a), profiles of SPT blowcounts (b), CPT tip resistance (c) and shear wave velocity
(d).

To complete the subsoil model assumed in the dynamic analyses, Figure 4 reports the curves
describing the variation of the equivalent shear stiffness, G, normalized with respect to the small-
strain value, G0, and of the damping ratio, D, with the shear strain amplitude, . For the pyroclastic
layer (Pyr), the curves were determined by performing Resonant Column Torsional Shear (RCTS)
tests on undisturbed block samples taken from the excavation front (Licata, 2015). For the SS and
Tuff1 formations, the curves were inferred from previous laboratory tests performed on samples of
the same lithotypes taken in other sites (Vinale, 1988).
 0.0001 0.001 0.01 0.1 1
1 20
18
0.8 16
14
Tuff1

G/G0
0.6 12

D (%)
SS 10
0.4 Pyr 8
6
0.2 4
2
0 0
0.0001 0.001 0.01 0.1 1
(%)
Figure 4. Variation with shear strain of normalized stiffness and damping ratio for SS, Pyr and Tuff1.

4. Static analyses

In order to calculate the internal forces acting in the walls supporting the excavation for the station
and in the tunnel lining, plane strain finite element analyses were carried out using the Plaxis FE
code (Brinkgreve et al, 2015). As shown in the following, different domain extension and
discretization were used to solve the soil-structure interaction problem for the station and for the
tunnel. However, the same constitutive models, and relevant mechanical parameters, were used for
the four layers described in Figures 2b,c and listed in Table I.
Table II summarizes the main geotechnical parameters adopted for the different soil layers.

Table II. Soil parameters for FE analyses

Soil Model sat c E E50ref Eoedref Eurref 0.7 G0ref pref
(kN/m3) - () (kPa) (MPa) (MPa) (MPa) (MPa) (%) (MPa) (kPa)
SS HSss 18 0.3 37 0 - 40 40 80 0.013 88 115
Pyr HSss 16 0.3 37 0 - 47 47 95 0.026 182 170
Tuff1 MC 16 0.3 27 200 2400 - - - - - -
Tuff2 MC 16 0.3 27 500 6700 - - - - - -

Both marine sand (SS) and pyroclastic silty sand (Pyr) were modelled using Hardening Soil
small strain (HSss), an elastic-plastic model with isotropic hardening, available in Plaxis model
library (Benz, 2007). The friction angle, , and most stiffness parameters were estimated on the
basis of the static penetration resistance, as suggested by different authors (e.g. De Beer, 1965;
Jamiolkowski et al., 1988; Mayne, 2007). In particular, the reference axial stiffness for
unloading/reloading, Eurref, was assumed as equal to twice the reference Youngs modulus, E50ref,
corresponding to the reference confining pressure, pref; the latter modulus, in turn, was set as equal
to 6 times the mean value of qc (as suggested by Schanz et al., 1999). As usually adopted in
practice, the reference constrained modulus, E oedref, was set equal to E50ref, while the reference small
strain shear modulus, G0ref, was obtained by the shear wave velocity corresponding to a mean
lithostatic effective stress equal to pref. The HSss model can take into account the stiffness decay
with strain level from the early stages of loading, by introducing the shear strain level 0.7 at which
the secant shear modulus, G, is reduced about to 70% of its initial small-strain value, G0.
For the underlying Neapolitan Yellow tuff, a linear elastic - perfectly plastic model obeying to
the Mohr-Coulomb strength criterion was used, with the stiffness, E, and the strength parameters, c
and , defined on the basis of previous experimental investigation by Evangelista & Pellegrino
(1990).
For all materials, the elastic Poissons ratio, , was set equal to 0.3; the initial stress state was
calculated by assuming lithostatic conditions, with k0=1-sin .
 4.1. Numerical modelling of the excavation of the station

On the basis of the monitoring data, with a strong simplification effort, the construction process
of the station could be subdivided into three main stages:
- stage I: execution of archeological excavations and dewatering tests. The excavation proceeded
slowly due to the possibility to find archeological remains. A dewatering test, which was judged as
a preliminary mandatory activity at the design stage, was carried out satisfactorily; it allowed for the
determination of the number of the pumping wells and of the total hydraulic capacity of the
pumping system, in order to guarantee the dewatering throughout the whole area;
- stage II: reaching of the design depth. The excavation was carried out by a relatively quick and
straightforward procedure, fully using the installed dewatering wells;
- stage III: shutdown of the pumping wells. The internal structural box of the station was
completely realized, including the protection of the waterproof membrane; later on, the closing of
the pumping wells allowed the groundwater level to recover the nearly hydrostatic initial
conditions.
The transversal cross section B-B, sketched in Figure 2c, was chosen to predict the behavior of
the staged excavation and the seismic response. Figure 5 shows the geometrical model adopted for
the 2D analyses, where plane strain conditions were assumed. The size of the calculation domain
was set as 250m 90m, in order to minimize the edge effects. The mesh was made of 3277
triangular elements with 15 nodes each. The average element size was 2.62 m. It must be noted that
the model is not symmetrical (differently from the sketch in Figure 2c), being the depths of the
grouted layer and of the tuff formation different on the opposite sides.

Buildings 24 m Sea
27 m

50 m
90 m

250 m
Figure 5. Geometrical model for the numerical analyses of the station.

The diaphragm walls, due to the huge section dimensions, were modeled as soil standard
triangular elements, rather than using the standard options available for structural elements, such as
plates. Linear elastic behavior was assumed for such elements; in detail, for the T section, two
different values of the Youngs modulus were adopted, i.e. 30 MPa for the portion of full concrete
and 13 MPa for the rib obtained by homogenizing soil with concrete.
The slabs were modeled with node-to-node anchors with unit spacing, by adopting an axial
stiffness EA equivalent to that of the actual plates (Table III). The bending behavior of the slabs
was not considered, due to their extremely low bending stiffness, compared to that of the T-shaped
diaphragm walls.
The jet grouting treatment was also modeled with soil-like elements, as an elastic perfectly
plastic material with Mohr-Coulomb strength, assuming stiffness and strength parameters suggested
in the literature (Croce et al., 2004) as summarized in Table IV.
 Table III. Axial stiffness of slabs
Depth of slabs EA
(m) (MN/m)
0 35.6E3
6.4 13.8E3
10.8 3.5E3
14.45 16.6E3

Table IV. Material parameters for the jet grouting

k c t c Eref = 300c
Material
(kN/m3) (m/s) () (MPa) (MPa) (MPa) (MPa)
Jet-grouting 16 1E-8 36 5 0 1.2 1500

At the soil-diaphragm interface, a limiting shaft friction was imposed, by considering a null
adhesion and a friction angle equal to 2/3 of that of the surrounding soil.
The static calculations consisted of 24 steps, in order to reproduce as realistically as possible the
true construction sequence (Autuori, 2016).
The final results are summarized in Figure 6; note that, being the construction and the subsoil not
symmetric, the profiles of kinematic and static variables on the two opposite sides, facing the sea
and the buildings, are not expected to be the same.
The plots in Figures 6a show the comparison of the horizontal displacements predicted (solid
line) at the end of the excavation with those measured (dots) at the locations of the T-shaped panels
P13 (sea side) and P74 (building side), respectively. The comparison of the results with the
monitoring data confirms that the deformation behavior of the diaphragm walls is greatly influenced
by the embedment in the tuff bedrock and by the top slab, which act as rigid constraints to the base
and the top, respectively. The presence of the three intermediate slabs does not affect significantly
the deformed shape of the T diaphragm walls, due to the relatively low axial stiffness of the props.
The numerical simulation of such a complex construction process included: excavations in
several steps, soil grouting, installation of four temporary prop levels (modeled as node-to-node
anchors, activated after the excavation immediately below the elevation of the installation), and
groundwater lowering of nearly 25 m. It can be appreciated that the above procedure was capable to
predict a fairly close agreement with the observed horizontal displacements of the diaphragm walls,
with the difference never exceeding 20%. A comparable satisfying agreement was obtained also for
vertical displacements at the ground surface (see Autuori, 2016).

(a) (b)
Horizontal displacement (mm) Horizontal displacement (mm) Bending moment (MNm/m) Bending moment (MNm/m)

-10 0 10 20 30 40 40 30 20 10 0 -10 -10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10

0 0

10 SS 10 SS

20 20
z (m)

z (m)

30 30
Pyr Pyr

40 40

Tuff1 Tuff1 Static

analysis
50 50
Tuff2 Monitoringdata
data Tuff2 Tuff2 Tuff2
Monitoring
Building side Calculated data Sea side Building side Sea side

Figure 6. (a) Horizontal displacement profiles calculated by the static analysis along the inner diaphragm wall vs. those
measured by inclinometer P74 (building side, left) and P13 (sea side, right); (b) bending moment predicted along the
same panels.
 In Figure 6b the bending moments accumulated in the large T-shaped panels are plotted against
depth. The maximum value of the bending moment occurs on the building side at the depth of about
26 m (practically corresponding to the maximum excavation depth) and reaches nearly 9.9 MNm/m.
On the sea side, the maximum value occurs at the same depth, but it is significantly lower, reaching
about 8.3 MNm/m.

4.2. Numerical modelling of the excavation of the tunnel

The segmental section of the tunnel (Figure 7a), the geometry of which was already described in
Section 2, was modelled as a continuous concrete ring (Figure 7b) in the numerical analyses,
adopting an equivalent inertia of the lining section, Ieq, defined as proposed by Wood (1975), by the
following formula:

2
4
I eq I j I (1)
n

where n is the number of joints, Ij the inertia of a single joint, I that of the continuous lining.

Figure 7. (a) Actual vs. (b) simplified geometry of the segmental tunnel lining.

A linear elastic behaviour was assumed for both concrete and grout, with the relevant parameters
summarized in Table V.
Table V. Tunnel lining parameters
Material E I ti Ieq teq
MPa kN/m3 m4/m m m4/m m
Concrete 33500 0.2 25 0.00225 0.30 0.00101 0.23
Grout 26000 0.2 15 0.00028 0.15 - -

The FE mesh of the subsoil and the tunnel developed in Plaxis 2D is shown in Figure 8. The
lateral boundary conditions during static analysis consisted of fixed displacements in the horizontal
direction at the vertical sides of the model and fixed displacements in both directions at the bottom.
Displacements along the upper surface were set free.
An interface between the grout and the soil was introduced, by using interface elements having
reduced strength and stiffness parameters compared to the surrounding soil. A reduction factor of
0.7 was adopted.
 (a) (b)

Figure 8. (a) 2D finite element model of the soil-tunnel system; (b) detail of the mesh around the tunnel.

The mechanical properties of each soil layer have been previously shown in Table II. In this
numerical model, only a single layer of tuff (Tuff1) was introduced, while the underlying intact tuff
was assumed as a rigid bedrock.
In order to reproduce the state of stress around the tunnel cavity, taking into account the 3D
arching effect that occurs within the soil and the deformations before lining installation, the initial
lithostatic stress, 0, was proportionally reduced around the cavity to:

1 0 (2)

where the stress relaxation coefficient was assumed equal to 0.3, according to Panet & Guenot
(1982).
The excavation was then simulated through three phases (Fabozzi et al., 2016):

Phase 1. Generation of the initial lithostatic stress field;

Phase 2. Stress relaxation;
Phase 3. Lining installation and grouting, in drained conditions.

Figure 9 shows the results of calculations in terms of bending moment M and normal force N
induced in the tunnel lining by such a simplified 2D excavation process. Starting from this static
state of stress, pseudo-static and full dynamic analyses were performed, as detailed in Sections 6-7.

Figure 9. Static internal forces in the tunnel lining

5. Free-field seismic response analyses

According to the probabilistic seismic hazard approach (PSHA), adopted by the Italian Building
code (NTC, 2008), the reference value of the peak ground acceleration, a g, for a given design limit
state, is related to the probability of exceedance, P r, of the earthquake along the reference life cycle
of the building or infrastructure, VR. For the Riviera di Chiaia site, the median hazard curve
shown in Figure 10a provides a value of ag=0.168g, by assuming a life safety limit state (i.e.
Pr=10%), and VR=50y, which correspond to a return period, T R, of the design earthquake as high as
475y.
 Figure 10b reports the corresponding de-aggregation histogram, providing the relative
contribution, w, to the selected hazard value as a function of the magnitude, M S, and site-source
distance, Repi. The plot reflects the major influence of near-field seismic sources with relatively low
potential magnitudes, located in the volcanic areas surrounding the city, compared to the minor
dependence of the hazard curve of peak acceleration from far-field high-magnitude sources,
pertaining to the Apennine chain fault systems.
The selection of natural seismic input motions used in the dynamic analyses was performed
using REXEL code (Iervolino et al., 2009). Seven seismic signals recorded on rock outcrop were
selected in the ranges of 4.5<Mw<7 and 0<Repi<100km, encompassing the whole de-aggregation
histogram (see Figure 10b). The average spectrum of the records resulted compatible with the class
A target spectrum, as specified for this site by the seismic hazard map, in accordance with the
NTC (2008). Finally, the signals were scaled to the design amplitude of ag=0.168g.
Figures 10c-d respectively show the selected time histories and the corresponding elastic
response spectra, these latter plotted together with the average and the target spectra.

(a) (b)
0.3 30

Pr=10% 25
TR=475y Pr=10%

w (%)
ag= 0.168g 20 TR=475y
0.2 ag= 0.168g
15
ag (g)

10

0.1 5
8.5-9.0
7.5-8.0
0 6.5-7.0

10
5.5-6.0

30
4.5-5.0

50

70
3.5-4.0

90

110
0
10 100 1000 Repi (km)
TR (y)

0.2 (c) (d)

0.1 0.2 Val Comino 0.8
Bingol
0 0.1 Friuli
0 50 Bingol Montenegro
-0.1 0.2 0 Aquila
0 10 20 30 40 50 0.6
-0.2 0.1 -0.1
Umbria
ValComino
Friuli
0.2 0 0.2 SouthIceland
-0.2
0 10 20 30 40 50 Average
0.1 -0.1 0.1
Sa (g)
a (g)

Target Spectrum
Montenegro 0.4
0 0 0.2
0.2 -0.2
0 10 020 30 40 50 50
-0.1 -0.1 0.1
0.1
Umbria
-0.2 0.2 -0.2 South Iceland 0
0
0 50
0.2
0 0.1 lAquila
50 -0.1
-0.1
0 0.2 -0.2
-0.2 0 50
-0.1 0.1 0
-0.2 0 0.01 0.1 1 10
0 10 20 30 40 50 T (s)
-0.1
t (s)
-0.2

Figure 10. (a) Seismic hazard curve for Riviera di Chiaia site; (b) de-aggregation chart; (c) selected accelerograms; (d)
input and target spectra.

Table VI. Main features of the selected seismic signals

Earthquake Earthquake Fault Station Repi Waveform
Date Mw
ID name mechanism ID [km] ID
2309 Bingol 01/05/2003 6.3 strike-slip ST539 14.0 7142
34 Friuli 06/05/1976 6.5 thrust ST20 23.0 55
93 Montenegro 15/04/1979 6.9 thrust ST64 21.0 198
291 Umbria Marche 06/10/1997 5.5 normal ST236 5.0 651
2142 South Iceland 21/06/2000 6.4 strike-slip ST2558 5.0 6349
71 Val Comino 11/05/1984 5.5 normal ATQ 17.4 276
178 l'Aquila 06/04/2009 6.3 normal FMG 19.3 806
 Table VI reports the main features of the selected signals, all recorded at European seismic
stations, which were subjected to a band-pass filtering in the range of 0.1-25 Hz prior to the seismic
response analyses.
Given the flatness of the area and the relatively homogeneity of the subsoil, characterized by
horizontal soil layering (see Figure 2), any topographic effect was excluded and the stratigraphic
amplification of the free-field motion was evaluated by means of one-dimensional seismic response
analyses. The analyses were carried out using the EERA code (Bardet et al., 2000), operating in
total stresses and in the frequency domain by a linear equivalent approach. The acceleration time
histories of the selected records were applied as reference input motions at the outcropping bedrock,
with a time step t=0.01s.
The dependency on shear strain amplitude of the equivalent parameters (shear stiffness and
damping ratio) of the materials SS, Pyr and Tuff1 was assumed as described by the curves shown in
Figure 4; the seismic bedrock (Tuff2) was assumed as a half-space with a linear visco-elastic
behavior.
Figure 11 shows the results obtained for the seven input motions in terms of vertical profiles of
maximum acceleration, amax, maximum shear strain, max, maximum horizontal displacement
relative to the bedrock, umax, and mobilized value of normalized shear stiffness, Gmob/G0.

(a) (b) (c) (d)

amax (g) max (%) umax (m) Gmob/G0

0.0 0.2 0.4 0.0 0.2 0.4 0.6 0 0.05 0.1 0 0.5 1
0 0 0 0

-10 -10 -10 -10

SS

-20 -20 -20 -20

z (m)

-30 -30 -30 -30

Pyr
Umbria
Bingol
-40 Montenegro -40 -40 -40
Tuff1 SouthIceland
Friuli
Aquila
Tuff2 ValComino
-50 -50 -50 -50

Figure 11. Results of 1D linear equivalent response analyses in terms of vertical profiles of: (a) maximum acceleration,
(b) maximum shear strain, (c) maximum displacement, (d) mobilized value of normalized shear stiffness.

It is clear that, even though some amplification of acceleration occurs across the Pyr formation,
the highest shear deformations develop within the more deformable SS surface layer. Moreover, a
different trend is observed in terms of maximum accelerations and strains among the different
earthquakes. As a matter of fact, the largest accelerations are related to Montenegro earthquake
(dark blue line) as far as the Pyr layer is considered, while lAquila earthquake (light blue line)
induces the highest horizontal displacements and, consequently, the largest strains throughout the
whole soil layering.
In order to explain these observations, Figure 12 illustrates a comparison between Montenegro
and lAquila records in terms of:
- frequency spectra, F(f), obtained by Fast Fourier Transform of the input accelerograms
(bedrock), of the accelerations computed at the top of Pyr layer (z=17m) and at surface
(z=0);
 - amplification functions, A(f), computed at the top of Pyr layer and at surface, both relative to
the bedrock.
By inspection of Figures 12a,b it is clear that the frequency content of Montenegro input motion
is spread out on the higher frequencies, up to about 10Hz, while LAquila record is characterized by
a lower frequency content, approximately in the range of 0.1-5Hz, with a dominant frequency of
0.8Hz. At the depth of z=17m, the 1 st and the 2nd modes of the Pyr layer tend to amplify lAquila
(Figure 12b) slightly more than Montenegro (Figure 12a), but in both cases the frequency content is
not substantially modified
90 by amplification phenomena. 5 90 At this depth, because 5 of the higher
Montenegro LAquila
frequency content, Montenegro
80 induces a value of80 amax larger than
FFT surface
(a) that caused by lAquila (see
FFT surface
(b)
4
Figure 11a). Conversely,
70 due to the low frequency4 content,
FFT Bedrock 70 the latterFFT motion
Bedrock tends to produce larger
Amplitude Amplitude
displacements and, thus, larger strains (see Figures3 11b-c).
60 60
3
Moving to the surface
50 (z=0, Figures 12c-d), the amplification
50 phenomena are clearly ascribed to

FFT
FFT

A
A
the deformability of 40the SS layer, the fundamental 2
40
frequency of which was2 checked to range
between 1.7 and 2.5Hz, depending on thez=0m
30 30
strain level mobilized by the selected z=0m input motion. As a
result, the signal characterized by the lowest frequency content, i.e. lAquila earthquake, was that
20 1 20 1
10
most significantly amplified by the uppermost layer.10
0 0 0 0
0 5 10 15 0 5 10 15
90 f (Hz) 5 90 f (Hz) 5
Montenegro (a)
(c) LAquila (d)
(b)
(g s)

80 80
FFT top-pyr FFT top-pyr
70 4 70 4
FFT Bedrock
FS top Pyr FS top PyrFFT Bedrock
Fourier amplitude

Amplitude Amplitude
FS bedrock FS bedrock

Amplitude
60 60
TF 3 TF 3
50 50
FFT

FFT
A

A
40 40
2 2
30 z=17m 30 z=17m
20 1 20 1
10 10

0 0 0 0
0 5 10 15 0 5 10 15
90 5 90 5
Montenegrof (Hz) (a)
(c) LAquila f (Hz)
(b)
(d)
(g s)

80 80
FFT surface FFT surface
70 FS surface
FFT Bedrock 4 70 FS surface
FFT Bedrock 4
Fourier amplitude

FS bedrock
Amplitude Amplitude
FS bedrock
60 60
Amplitude

TF 3 TF 3
50 50
FFT
FFT

A
A

40 40
2 2
30 z=0m 30 z=0m
20 1 20 1
10 10

0 0 0 0
0 5 10 15 0 5 10 15
90 5 90 5
f f(Hz)
(Hz) f f(Hz)
(Hz)
80
Montenegro (c) 80 LAquila (d)
Figure 12. Fourier spectra
70
and amplification functions of4 Montenegro
FFT top-pyr
FFT Bedrock 70
(a, c) and lAquila (b, 4d) earthquakes.
FFT top-pyr
FFT Bedrock
Amplitude Amplitude
60 60
3 3
This latter input motion
50 was therefore viewed as50 the most conservative loading condition for
FFT

FFT
A

predicting the seismic40performance of both the station

2
40 and the tunnel, with the 2analyses which will
be described in the following sections. z=17m
30 30 z=17m
20 1 20 1
10 10

0 0 0 0
6. Pseudo-static analyses
0 5 10 15 0 5 10 15
f (Hz) f (Hz)

In order to simulate the seismic performance of both the station and the tunnel, although in an
approximate way, a simplified decoupled approach was first followed. With such a kind of
approach, the effect of an earthquake is simulated with an equivalent seismic load, statically applied
to the structure as a distribution of inertia forces or displacements (e.g. Argyroudis & Pitilakis,
2012, Do et al., 2015). The results of the free-field seismic response analyses described in the
previous section were therefore used as pseudo-static loading of the same subsoil-structure models
adopted for the static analyses reported in Section 4.
 For both the station and the tunnel, the profile of peak horizontal displacements corresponding to
the free-field soil response to lAquila earthquake (light blue line in Figure 11c) was statically
applied to the boundaries of the numerical model. The soils were modeled as linearly elastic, by
adopting the profile of the equivalent shear modulus, G mob, mobilized in the corresponding free-
field analysis (light blue line in Figure 11d). As well-known, the results of elastic analyses are
independent of the initial stress state, which makes them not affected by the static conditions
considered.

6.1. Pseudo-static analyses of the station

Figure 13 compares the bending moments along the two peripheral diaphragm walls, as obtained
from the static analysis (black lines), already reported in Figure 6b in Sec. 4.1, with the seismic
increments obtained through the pseudo-static analysis (red lines); this latter was carried out by
assuming, for simplicity, the lithostatic stress state as initial condition. The seismic action was
simulated as horizontal displacements acting along the two opposite vertical boundaries of the same
model shown in Figure 5, with the verse of the displacements pointing towards the sea side, which
represents the worst condition for the most flexible diaphragm, i.e. that located at the buildings side.

Bending moment (MNm/m) Bending moment (MNm/m)

-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
0

10 SS

20
z (m)

30
Pyr Static analysis
Pseudo - Static
40 (umax)
Pseudo - Static
Tuff1
(Kh)
50
Tuff2 Tuff2

Building side Positive Sea side

displacements
Positive
Inertial forces
Figure 13. Comparison of the bending moments obtained from static analysis of the excavation of the station vs. the
seismic increments resulting from the pseudo-static analysis using imposed displacements and inertia forces.

It can be noted that the seismic increments of the bending moment are comparable to the static
values from surface to 5m depth, i.e. beneath the first slab; they keep low values moving from the
surface down to the depth of the lowest slab (17m), the maximum bending moment being attained
at an intermediate depth between the lowest slab and the very bottom of the excavation (i.e. about
27 m). In this depth range, they have an opposite sign with respect to those induced by the
excavation, while the are increasing the static values along the foot of the diaphragm walls. This
discordance can be explained recalling that the sign of the static moments does not change across
the excavation bottom, since the slab was cast in place after they have occurred reaching the
maximum excavation depth.
 As a consequence, the static and the pseudo-static maximum bending moments cannot be simply
added for the structural assessment of the panels, because they occur at different depths. In fact, the
maximum value of the additional seismic moments is about 3 MNm/m on both buildings and sea
sides, and occurs at a depth of about 24m, slightly above the excavation bottom (27m),
corresponding to the diaphragm section mostly stressed by the excavation, with static moments
equal to 9.9 or 8.3 MNm/m on the buildings and sea side, respectively. In such a section, the
additional pseudo-static bending moments are approximately equal to 2 MNm/m on both sides of
the excavation; although smaller than the above mentioned maximum values, they are high enough
to make this depth as that to be considered as the most critical in seismic conditions, with a resultant
peak value of the bending moment equal to about 12 MNm/m on the buildings side.
It was not possible to run reliable dynamic analyses for the station, since the FEM mesh had
been optimized, in terms of boundary distance and mean grid spacing, for reproducing the observed
static deformation following the complex construction stages; the same model was therefore not
adequately suited for full dynamic analyses. Being aware of the inherent limitations, a pseudo-static
analysis was thus performed, by using a horizontal seismic coefficient, k h, uniformly applied on the
whole domain as a multiplier of the soil and structural weights. This simplified assumption was
considered to be reasonable, by looking at the moderate variability of the profiles of a max reported in
Figure 11a and after checking the negligible effects of non-synchronous motion. Therefore, the
coefficient kh was assumed as equal to the weighted average of the profile of a max, the weights being
the thickness of the layers times the soil unit weight. For all the input accelerograms considered, the
value of kh resulted constrained between 0.17 and 0.20, with the maximum value pertaining to
lAquila input motion; this latter was considered - once again - the most conservative seismic
loading.
The bending moments resulting from the alternative pseudo-static analyses with k h=0.20 are
plotted with blue lines in Figure 13; it can be observed that the seismic increments computed with
this inertial pseudo-static approach result, at least in the most loaded sections, higher than those
predicted by the kinematic pseudo-static approach above described. Note that, by using a seismic
coefficient uniformly acting throughout the whole domain, the sign of the seismic increments along
the excavation is the same of that predicted with the kinematic approach, while the trends appear
oriented towards the same verse along the foot of the walls. Nevertheless, the worst combination of
static and seismic moments is still located at the bottom of the excavation and approaches 13
MNm/m on the building side.

6.2. Pseudo-static analyses of the tunnel

Figure 14 shows the FEM mesh adopted for the pseudo-static analysis of the tunnel, with a
schematic pattern of the boundary conditions. As suggested by Argyroudis & Pitilakis (2012), the
distribution of displacements corresponding to the input motion was applied at a distance of three
diameters from the tunnel. The numerical mesh consists of 135273 15-node triangular elements,
with an average size of 0.5m. The lateral boundary conditions during pseudo-static analysis allow
for free displacements along the vertical direction; the bottom nodes of the domain are fixed, while
those at surface are free to move in vertical direction, and constrained to a constant horizontal
displacement equal to the maximum free-field value, as also assumed in previous literature studies
(e.g. Do et al., 2015; Tsinidis et al., 2016).
 ux ux
(a) (b)

d=3D d=3D

Figure 14. (a) Mesh and boundary conditions of the numerical model for pseudo-static analysis of the tunnel, (b) detail of the
mesh around the tunnel.

Figure 15 shows the contours of horizontal displacements predicted with the above numerical
model in free-field conditions (a) and with the presence of the structure (b). The first drawing
confirms that the reduced size of the domain was enough to reproduce a homogeneous field of
displacements, while the second contour shows the shadowing effect induced by the presence of
the tunnel on the deformation of the soil around.

(a) (b)

Figure 15. Contours of horizontal displacements in free-field conditions (a) and with soil-tunnel interaction (b).

The effect of the boundary distance from the structure was investigated in terms of increments of
internal bending moments, M, and hoop forces, N, induced in the transverse tunnel section.
Figure 16 shows the comparison for the cases of boundary distance, d, equal to three, two and one
time the tunnel diameter, D. Differences are negligible if not null.

(a) (b)

Figure 16. Influence of boundaries on the pseudo-static increments of bending moments (a) and hoop forces (b).

7. Full dynamic analysis of the tunnel

The above mentioned difficulty to use in dynamic conditions the same numerical model
reproducing the history of excavation was much less in the case of the tunnel-subsoil system. A
coupled approach was followed, being the soil-structure interaction inherently included in the step-
by-step solution of the equations of motion in the time domain, this time accounting for the pre-
failure non-linear behavior of soil through the HSss constitutive model.
It is worth mentioning that the same FE mesh used for static analysis (see Figure 8) was
originally calibrated for the full dynamic analysis. As a matter of fact, the width of such numerical
model was originally optimized by a series of sensitivity analyses, in order to reach a true free-field
condition at the lateral borders and to minimize the influence of vertical boundaries, modeled with
viscous dashpots as suggested by Lysmer & Kuhlemeyer (1969). Also, the mesh was discretized
aiming at a reliable propagation of the maximum significant frequency of the input signal
(Kuhlemeyer & Lysmer 1973), leading to an average size of 0.65m.
The reference input motion of lAquila earthquake was applied at the rigid base of the model as a
time history of acceleration. An additional small-strain viscous damping was introduced in the
dynamic calculation by means of the well-known Rayleigh formulation, which considers a linear
combination of the mass [M] and the stiffness [K] matrices as follows:

C R M R K (3)

The damping coefficients, R and R, were calculated following to the double frequency approach,
assuming as target values the first natural frequency of the deposit and the main frequency of the
input motion.
The full dynamic analyses of the tunnel-subsoil system were carried out in drained conditions,
since previous analyses of liquefaction potential by Licata (2015) showed that the soil layers
surrounding the tunnel are unlike to suffer for neither liquefaction nor significant pore water
pressure build-up under the seismic actions considered.
The non-linear soil constitutive model allowed for predicting the evolution of the internal forces
in the lining from the values following the excavation stages (see Section 4.2) until the end of
shaking. Figure 17, indeed, shows the time history of the bending moment and the hoop force at the
section =283, where the ratio M/N reaches its maximum value over time. The residual bending
moment is equal to 8.5 kNm/m, with a peak value of 54 kNm/m, while the residual normal force is
equal to 630 kN/m, with a peak value of 692 kN/m.

Figure 17. Time histories of (a) bending moment, M, and (b) hoop force, N, at the section located at =283 in the lining

Figure 18 shows the increments of internal forces around the whole lining resulting from the
pseudo-static and full dynamic analyses; in this latter case, the values plotted are those predicted at
the time of peak acceleration, for consistency with the assumptions made for applying the pseudo-
static approach.
The comparison shows that, at least for this particular case, the pseudo-static analysis
underestimates the increment of forces with respect to the full dynamic analysis.
 (a) (b)

Figure 18. Comparison between pseudostatic and full dynamic analyses in terms of seismic increments of (a) bending moment, M,
and (b) hoop force, N.

Literature shows that the magnitude of such a difference between the two approaches is strongly
dependent on the soil-structure relative stiffness and the interface behavior, with the tendency to
increase with the lining flexibility and for a condition closer to no slip interaction (Tsinidis et al.,
2016). In the case of Metro Line 6 of Napoli, characterized by a rather flexible lining with a soil-
structure interface closer to no slip rather than a full slip condition, the two approaches show a
difference up to about 60% in the most loaded tunnel sections. This result is in agreement with
previous studies (Bilotta et al., 2007; Argyroudis & Pitilakis, 2012; Tsinidis et al., 2016) showing,
in almost all cases of flexible lining, an underestimation of the dynamic increment of the internal
forces with the simplified pseudo-static approach, as a function of the interface behavior and lining
flexibility.

8. Conclusive remarks

The assessment of the seismic safety of underground structures is becoming more and more
important, especially in densely populated areas, even with low to moderate seismicity.
Notwithstanding the recent updates in the use of numerical methods and the gathering of experience
provided by experiments on physical models, significant doubts still arise in the engineering
practice when accounting for studies published in the literature or design guidelines, if any.
In this paper, a case study represented by a large open multi-propped excavation and a circular
segmented tunnel in a densely urbanized area of the city center in Napoli has been used to
investigate some of the mentioned aspects.
For instance, for the complexity of both excavation geometry and staged construction such as
those characterizing San Pasquale station, a full dynamic analysis was considered neither affordable
nor reliable. The two conventional pseudo-static analyses, applying either displacements along the
boundaries and, alternatively, inertial forces throughout the whole subsoil-structure domain,
provided rather different profiles of bending moments, the larger values coming from the latter
analysis.
The incremental bending moments calculated in the diaphragm walls supporting the deep
excavations were not negligible when compared to the rather large static values calculated during
the excavation steps. For the kinematic approach, the incremental bending moment in the most
loaded section at the bottom of the excavation was a bit less than 20%, while that predicted at the
same depth by the inertial method approached 40% of the static value.
In the case of the tunnel, due to the relatively simpler geometry and staged excavation, the
seismic increments of bending moments and axial hoop forces could be calculated through both a
simplified pseudo-static analysis and a full dynamic analysis, showing a satisfying agreement. The
kinematic approach was followed, showing a poor sensitivity to the extension of the FE domain, in
this particular case. It is worth noting that a pronounced influence of the size of the model might be
found in more 'complex' subsoil conditions, or in cases where significant soil plasticization occurs
in the equivalent pseudo-static analysis.
 By an engineering point of view, both approaches show that significant seismic increments of
internal forces arise in the lining. In fact, the various analyses yielded incremental bending moments
of the same order of magnitude and incremental axial thrusts about 20% of the corresponding
maximum static values. It should be kept in mind that the segmental tunnel lining is a typical
solution suited for reducing the bending moments due to its relatively low flexural stiffness, thus
both the static and earthquake-induced bending moments are relatively small and not particularly
demanding for the structural section.
As a final remark, the example of the San Pasquale station and tunnel highlights that seismic
effects on underground structures should not be overlooked, even in low to moderate seismicity
areas. Furthermore, the lessons learned from this pilot study on a real case history were enough
stimulating for assessing the limitations of simplified methods of pseudo-static analysis versus more
advanced dynamic approaches. While the case of the station highlighted the difficulty of finding
mesh optimization criteria satisfying the needs of accuracy for both static and dynamic analyses of
multi-propped excavations, that of the tunnel confirmed that, even with relatively simple lining
geometries, the approximation of the pseudo-static analyses can result still too rough.

Acknowledgements

This work was carried out as part of WP3 Tunnels of the sub-project on Earthquake Geotechnical
Engineering, in the framework of the research programme funded by Italian Civil Protection
through the ReLUIS Consortium. Dr. Lorenza Evangelista is warmly acknowledged for the help
provided in preparing the map in Figure 1.

References

Argyroudis A., Pitilakis K. (2012). Seismic fragility curves of shallow tunnels in alluvional deposits. Soil
Dynamics and Earthquake Engineering; 35,1-12.
Autuori S. (2016). La stazione San Pasquale della metropolitana di Napoli. Costruzione, monitoraggio ed
analisi dei comportamenti osservati. PhD Thesis. University of Napoli Federico II.
Benz T. (2007). Small-strain stiffness of soils and its numerical consequences. PhD thesis, University
Stuttgart.
Bardet J., Ichii K., & Lin C. (2000). EERA a Computer Program for Equivalent-linear Earthquake site
Response Analyses of Layered Soil Deposits. Univ. of Southern California, Dep. of Civil Eng.
Bilotta E., Lanzano G., Madabhushi S.P.G., Silvestri F. (2014). A round robin on tunnels under seismic
actions. Acta Geotechnica 9 (4), pp. 563-579. doi:10.1007/s11440-014-0330-3.
Bilotta E., Lanzano G., Russo G., Santucci de Magistris F., Silvestri F. (2007). Methods for the seismic
analysis of transverse section of circular tunnels in soft ground. ISSMGE-ERTC12 Workshop at XIV
ECSMGE Geotechnical Aspects of EC8, Chapter 22, Patron Editore.
Brinkgreve R.B.J., Kumarwamy S., Swolfs W.M. (2015). Plaxis, Material Models, Netherlands; 65-80.
Conti R., Madabhushi G.S.P., Viggiani G.M.B. (2012). On the behaviour of flexible retaining walls under
seismic actions. Geotechnique; 62, 1081-1094.
Croce P., Flora A., Modoni G. (2004). Jet grouting. Technology, Design and Control. CRC Press; 24-26.
DM 14/1/2008. Norme Tecniche per le Costruzioni. S.O. n. 30 - Gazzetta Ufficiale della Repubblica Italiana,
No. 20 - 4/2/2008.
De Beer E. (1965). Bearing capacity and settlement of shallow foundations on sand. Proc. Symp. On Bearing
Capacity and settlement of Foundations, Duke University; 15-33.
Do N.A., Dias D., Oreste P., Maigre I. D. (2015). 2D Numerical investigation of segmental tunnel lining
under seismic loading. Soil Dynamics and Earthquake Engineering 72 (2015) 66-76.
Eurocode2 (2004). Design of concrete structures. Part 1-1: General rules and rules for buildings. EN 1992-1-
1:2004.
Evangelista A., Pellegrino A. (1990). Caratteristiche geotecniche di alcune rocce tenere italiane. III ciclo di
Conferenze di Meccanica e Ingegneria delle Rocce, MIR90, Torino.
Fabozzi S., Bilotta E. (2016). Behaviour of a Segmental Tunnel Lining under Seismic Actions. Procedia
Engineering, 158, pp. 230-235.
Fabozzi S., Bilotta E., Russo G. (2016). Numerical interpretation of monitoring data of an instrumented
tunnel segmental ring. 1st IMEKO TC-4 International Workshop on Metrology for Geotechnics.
Benevento, Italy, March 17-18.
Hashash Y.M.A., Hook J.J., Schmidt B., Yao J.I.-C. (2001). Seismic design and analysis of underground
structures. Tunn and Undergr Space Technology 16, 247293.
Hashash Y.M.A., Park D., Yao J.I-C. (2005). Ovaling de-formations of circular tunnels under seismic
loading, an up-date on seismic design and analyses of underground structures, Tunnelling and
Underground Space Technology, 20, 435-441.
Iervolino I., Galasso C., Cosenza E. (2009). REXEL: computer aided record selection for code-based seismic
structural analysis. Bulletin of Earthquake Engineering, 8, 339-362.
Iida H., Hiroto T., Yoshida N., Iwafuji M., (1996). Damage to Daikai subway station, Soils and Foundations,
Japanese Geotechnical Society, Special issue on geotechnical aspects of the 17 January 1995
Hyogoken-Nambu Earthquake, p. 283-300.
Jamiolkowski, M., Ghionna V.N., Lancellotta R., Pasqualini E. (1988). New application of penetration tests
in design practice. Proc. I Int. Symp. On Penetration Testing (ISOPT-1), 1:263-296. Rotterdam, The
Netherlands: Balkema.
Kawashima K. (1999). Seismic design of underground structures in soft ground: a review, Geotechnical
Aspects of Underground Construction in Soft Ground, Kusakabe, Fujita & Miyazaki (eds). Balkema,
Rotterdam, ISBN 90 5809 1 066.
Kuhlemeyer R.L, Lysmer J. (1973). Finite Element Method Accuracy for Wave Propagation Problems.
Journal of the Soil Mechanics and Foundation Division, 99(5), 421- 427.
Lamante D., Flora A., Russo G., Viggiani C. (2012) Displacements induced by the installation of diaphragm
panels. Acta Geotechnica 7(3) 203-218;doi: 10.1007/s11440-012-0164-9, 2012
Licata V. (2015). A laboratory and field study on cyclic liquefaction of a pyroclastic soil. PhD Thesis in
Geotechnical Engineering. University of Napoli FedericoII.
Lysmer J., Kuhlemeyer R.L. (1969). Finite dynamic model for infinite media, J. of Eng. Mech. Div., ASCE:
859-877.
Mayne P.W. (2007). Cone Penetration Testing (NCHRP Synthesis). Transportation Research Board, National
Research Council, Washington, DC.
Panet M., Guenot A. (1982). Analysis of convergence behind a face of a tunnel. Tunnelling, 82, Institute of
mining and metallurgy, 197-204
Pitilakis K., Tsinidis G. (2014). Performance and seismic design of underground structures, State-of-Art, In
Earthquake geotechnical engineering design (eds M. Maugeri and C. Soccodato), Geotechnical
Geological and Earthquake Engineering, 28, 279-340. http://dx.doi.org/10.1007/978-3-319-03182-
8_11. Geneva, Switzerland: Springer.
Plaxis 2D AE. (2015). Brinkgreve R.B.J., Kumarwamy S., Swolfs W.M. Reference manual, Netherlands,
Plaxis bv.
Rippa F., Vinale F. (1983). Effetti del terremoto del 23 Novembre 1980 sul patrimonio edilizio di Napoli. Atti
XV Convegno Italiano di Geotecnica, Spoleto. AGI, Roma.
Russo G., Viggiani G.M.B, Viggiani C. (2012). Geotechnical design and construction issues for Lines 1 and 6
of the Naples Underground. Geomechanics and Tunnelling vol. 5, No. 3 DOI:
10.1002/geot.201200016 300-312
Schanz T., Vermeer P.A., Bonnier P.G. (1999). The hardening-soil model. Formulation and verification. In
R.B.J. Brinkgrave, Beyond 2000 in Computational Geotechnics, Balkema, Rotterdam, 281-290.
Soccodato F.M., Tropeano G.. (2015). The Role of Ground Motion Characters on the Dynamic Performance
of Propped Retaining Structures. Proc. 6th International Conference on Earthquake Geotechnical
Engineering, 1-4 November 2015, Christchurch, New Zealand (paper no. 645).
St.John C.M., Zahrah T.F. (1987). Aseismic design of underground structures, Tunnelling and Underground
Space Technology, 2 (2), 165-19.
Tropeano G.., Soccodato F.M. (2014). Dynamic analyses of propped retaining structures. Proc. 8th European
Conf. on Numerical Methods in Geotechnical Engineering (NUMGE 2014), 2:1193-1198.
Tsinidis G., Pitilakis K., Anagnostopoulo C. (2016). Circular tunnels in sand: dynamic response and
efciency of seismic analysis methods at extreme lining exibilities. Bull Earthquake Eng, 14:2903
2929 DOI 10.1007/s10518-016-9928-1.
Vinale F. (1988). Caratterizzazione del sottosuolo di unarea campione di Napoli ai fini di una
microzonazione sismica. Rivista Italiana di Geotecnica, 2(1988), 77-100.
Wood A. M. (1975). The circular tunnel in elastic ground. Geotechnique, 25(1), 115-127.
Yoshida N., (1999), Underground and buried structure, Earthquake Geotechnical Engineering, Seco e Pinto
(ed.) 1999 Balkema, Rotterdam.