Tunnelling and Underground Space Technology 153 (2024) 106001

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

incorporating Trenchless Technology Research
journal homepage: www.elsevier.com/locate/tust

A strain applied method for FEM-2D modelling of TBM tunnels in

coarse-grained soils with comparative analysis of case histories
F. Schiena a , E. Lusini b, * , A. Lembo Fazio a , A. Graziani a
a
Department of Civil, Computer Science and Aeronautical Technologies Engineering - DICITA, Università degli Studi “Roma Tre”, Via Vito Volterra 62, 00146 Roma,
Italy
b
Department of Chemical Engineering Materials Environment – DICMA, Sapienza University of Rome, via Eudossiana 18, 00183 Rome, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: The paper focuses on the prediction of tunnelling-induced settlements in greenfield conditions. The main com-
TBM mechanized excavation ponents of ground deformation are investigated with approaches based on both empirical and numerical solu-
Shallow tunnels tions. In particular, the study analyses five well documented case histories of shield-driven tunnels bored in
Settlements
coarse-grained soil deposits and compares the predicted and measured settlements.
2D finite element analysis
Coarse-grained soils
A specific approach for FEM-2D modelling of soil-shield interaction (called “Strain Applied Method”) has been
Prescribed strains implemented and its potential for the prediction of the settlement trough has been discussed. The analyses were
performed for the green field conditions with the Finite Element software Plaxis. The mechanical behaviour of the
soil has been modelled by both the usual linear elasto-plastic model (Mohr-Coulomb, MC) and the advanced
Hardening Soil model with small strain stiffness (HSs). The comparison of results shows that the settlements
calculated by the HSs model are generally in better agreement with the field measurements. Then, the analysis
has been focused on the volume loss obtained for the five case histories, finding some significant correlations
between volume loss and the specific parameters required by the proposed FEM-2D model. The predictive ca-
pabilities of the 2D model, with parameters estimated by specific correlations, were demonstrated by applying
the proposed method for two new case histories, not included in the initial set utilized to find the correlations.

1. Introduction unavoidable effects of over-excavation. Under this term Lee et al. (1992)
list three major phenomena: i) the over-excavation proper, due to the
The boring process causes stress relief on the tunnel walls as well as larger diameter of the cutting face with respect to the shield; ii) the
on the face. Since tunnelling in urban areas usually takes place at conicity of the shield, tapering toward the tail; iii) the gap between the
shallow depth, easily the induced deformations spread up to the ground soil and the lining behind the shield, normally filled with grout under
surface and affect the existing structures. The surface settlements are pressure. All these phenomena take place behind the face, while other
influenced by the excavation method: they can be effectively controlled additional ground losses can occur in front of the face when the ideal
by using EPB or Slurry-Shield mechanized excavation, able to apply a value of support pressure is not reached, or is lost during machine
continuous support pressure to the tunnel face. The general criteria to maintenance or is difficult to calibrate due to mixed face conditions
select the excavation method and, more specifically, the design pa- (Lambrughi et al., 2012).
rameters of the TBM depend on a large number of variables: soil con- The basic approach to the prediction of settlements is the “semi-
ditions, tunnel depth, hydraulic conditions and many other technical empirical method” introduced by Peck (1969) and subsequently refined
and economic aspects, often site-specific, discussed in various guidelines based on data from a variety of tunnelling projects (e.g., O’Reilly and
and technical papers (e.g., ITA/AITES Guidelines, 2000, Pelizza, 2002, New,1982; Attewell and Woodman,1982; Mair and Taylor, 1997). The
Duhme and Lee 2021, DAUB Recommendations 2022). semi-empirical method, based on the assumption of a “Gaussian curve”
Even if closed-face pressurized TBMs can effectively reduce to a to represent the settlement trough, represents an effective tool for the
minimum the deformation around the cavity and, consequently, the preliminary design of urban tunnels, to be complemented by numerical
settlements at surface, they cannot totally prevent them, due to finite element analyses in the detailed design phase.

* Corresponding author.
E-mail address: [email protected] (E. Lusini).

https://doi.org/10.1016/j.tust.2024.106001
Received 6 February 2024; Received in revised form 15 July 2024; Accepted 29 July 2024
Available online 10 August 2024
0886-7798/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 1. Gaussian curve for transverse settlement trough, nominal tunnel diameter, volume of the settlement trough, V*S and volume loss at the tunnel excavation
profile, V*L.

The Gaussian curve accurately describes the settlement trough in Möller and Vermeer, 2008) in which the initial geostatic stress is
undrained clay but does not always match well to the trough shape in replaced by a grout pressure distribution into an annular ring of soil
drained soils. The discrepancy increases with the intensity of yielding elements around the tunnel lining.
around the tunnel and the formation of vertical shear bands, which tend While pressure-controlled methods exhibit a narrower settlement
to give a chimney-like mechanism (Mair and Taylor, 1997; Celestino et profile with respect to displacement-controlled ones, however, it has
al, 2000). Curves with additional degrees of freedom, such as the been observed that, for both approaches, the numerical analyses tend to
modified Gaussian and the yield density curves, can provide a better fit predict settlement troughs wider than those observed experimentally, in
for a specific monitoring section but comparison of data from different small-scale centrifuge tests (Song and Marshall, 2020) as well as in real
tunnels becomes more complicated (Vorster et al., 2005; Marshall et al., tunnels (Möller, 2006; Schiena, 2020).
2012). In this paper a new procedure for modelling excavation and shield-
FEM modelling overcomes some limitations of the empirical soil interaction in 2D conditions will be presented (called “Strain
methods, particularly for soil-structure interaction analyses. Two ap- Applied Method”). The proposed approach, suitable for implementation
proaches of increasing complexity can be distinguished to this respect. A in FEM codes, brings together some aspects of the Gap and Contraction
preliminary level modelling approach, in which tunnelling effects on methods and involves a two-stage evaluation procedure of the de-
existing structures are evaluated by means of uncoupled analyses: e.g., formations to be imposed around the tunnel section.
the foundation plane of the building is assumed to accommodate the The imposed strains can be easily calibrated based on surface volume
greenfield displacement trough. The influence of building weight and losses.
stiffness on the settlement profile is therefore neglected. These limita- The performance of the method will be analysed by comparing the
tions are removed in the second-level “coupled analyses”, which how- computed results to the measurements in greenfield conditions recorded
ever require the structural model of the building also being included in in five well-documented case histories. The analyses were performed by
the global model of subsoil and tunnel (Boscardin and Cording 1989; two different soil models: the conventional ideal elastoplastic model
Addenbrooke et al., 1997; Burd et al., 2000; Boldini et al., 2021; Xu with Mohr-Coulomb (MC) strength criterion and the Hardening Soil
et al., 2021). model with small strain stiffness (Hardening) model implemented in
FEM models have the additional advantage to make it easy to analyse Plaxis 2D code (Brinkgreve et al., 2017). In fact, it has been already
the complex interaction that takes place, during the excavation advance, observed that a non-linear stiffness constitutive model can significantly
between soil and the support system, represented first by the TBM shield improve the capabilities of FEM models, particularly for coarse-grained
and, then, by the lining (e.g. Breth and Chambosse, 1975; Viggiani and soils (e.g., Addenbrooke et al., 1997; Hejazi et al., 2008).
Standing, 2001; Mair, 2003). This aspect has been investigated in full It will be shown that the settlements predicted by the developed
detail by 3D step-by-step modelling (Kavvadas et al. 2017; Ochmański procedure in combination with the HSs soil model, are generally in
et al., 2018). However, in engineering practice, various methods have better agreement with field data and empirical predictions (Gaussian
long been developed that allow for a simpler analysis of the construction curves), rather than those obtained with the conventional linear elasto-
process, using 2D rather than 3D models. plastic MC model. A specific section of the paper will be devoted to the
It is useful to recall the main approaches utilized for 2D simplified influence of soil dilatancy and volumetric cap, within the HSs model, on
analysis of mechanised tunnel excavation. They can be grouped into the predicted settlements at ground surface.
three main methods. The Convergence-Confinement Method (Panet and The comparative study of the five case histories, in terms of
Guenot, 1983), in which the initial geostatic stress on the excavation maximum settlement and shape of the settlement trough, has led to
profile is progressively reduced as a function of the face distance where identify a strong correlation between overall settlement parameters,
the support annulus is installed. The Gap Method (Rowe et al., 1983) in particularly the surface volume loss, and the specific deformation pa-
which an initial mismatch is considered between the support annulus rameters, to be imposed at the tunnel level, required by the proposed
and the excavation profile, maintaining lining and soil unconnected procedure. Therefore, the obtained results were employed to set up a
until the initial gap is closed. The Contraction Method (Vermeer and framework of relationships useful for the assessment of the parameters
Brinkgreve, 1993) can be considered a variation of the gap method in required by 2D tunnel modelling in coarse-grained soil conditions. The
which the support annulus is uniformly contracted accordingly to the use of such relationships has been demonstrated by their application for
TBM overcut and shield diameter, and therefore forced to maintain a two new cases of tunnel excavated in Italy.
nearly circular shape. Finally, the Grout Pressure Method (Möller, 2006; The comparative analysis has been extended also to other aspects

2
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 2. Typical dimensions of the finite element model (a) and example mesh for greenfield analyses (b).

than solely the vertical settlements, as horizontal displacements of Equation (1).

ground surface and internal forces in the lining, for the case histories for O’Reilly and New (1982), analysing field data from a series of
which these data were also available. tunnelling projects in relatively homogeneous soils, showed that ix is
The Strain Applied Method, developed in this study for the specific proportional to the depth of the tunnel axis z0:
case of closed-face tunnelling in coarse-grained soils, is believed to be
i x = K z0 (3)
applicable, with appropriate adaptations, even in different soil condi-
tions. The special interest in coarse-grained soils was motivated by The “width parameter” K depends essentially on the soil type: for coarse-
several ongoing tunnelling projects in Italy, which involve subsoil con- grained soils varies from 0.25 to 0.45, with an average value K = 0.35
ditions with prevailing alluvial deposits or fine-coarse intermixed strata (Mair and Taylor, 1997).
(e.g., Boldini et al., 2023). The basic features of sandy soil response to O’Reilly and New (1982) also observed that the displacement vector
tunnel excavation have been thoroughly investigated by laboratory is approximately directed toward the axis of the tunnel, or, in other
centrifuge tests, providing new insights into the influence of tunnel words, that the horizontal displacement, uh, may be calculated from the
cover-to-diameter ratio, soil relative density and tunnel volume loss vertical settlement at the same location, w, by the following equation
(Franza et al., 2019; Song and Marshall, 2020). However, the behaviour
of real tunnel is more complex, due to the influence of non- x
uh (x) = − w(x) (4)
homogeneous soil conditions and different TBM characteristics, in z0
contrast with the perfectly controlled and idealized laboratory envi-
A convenient reference parameter often adopted in design practice is the
ronment. Conversely, the focus of this study is on back analysis of real
“ground volume loss” (VS), which represents the volume of the surface
tunnelling projects: a set of 7 case histories is analysed, finding relevant
settlement trough related to the nominal volume of the tunnel cross-
common features, useful to calibrate the proposed Strain Applied Method.
section, while the “tunnel volume loss” (VL), represents the closure of
the excavation profile, normalised in the same way (Df is the nominal
2. Basic definitions and empirical background
diameter of the tunnel)
This section briefly reviews the empirical equations commonly used 4 VS*
for estimating soil settlements induced by tunnelling in greenfield con- VS = (5a)
π D2f
dition, i.e. in absence of buildings and other interacting structures. The
settlement trough induced by tunnel excavation, with constant depth of 4 VL*
the tunnel axis, z0, is well represented by a Gaussian curve: this is the VL = (5b)
πD2f
basic assumptions of the “empirical” approach developed by Peck
(1969) and extensively used in engineering practice. The surface set- In perfectly undrained conditions (i.e., for an incompressible medium),
tlements w(x) in a transversal cross-section can be therefore calculated the ground volume loss (VS) is equal to the tunnel volume loss (VL). For
as (Figure 1): coarse-grained soils, however, the surface volume loss can be lower or
( ) higher than the tunnel volume loss if a dilative or contractive volumetric
x2
−
2ix2
behaviour prevails, depending on the dilatancy angle (ψ ) of the soil.
w(x) = wmax e (1) Thus, the volume losses (VS) is a function of the mechanical properties of
the soil as well as of the construction method (i.e., the specific process of
∫ √̅̅̅̅̅̅ V* excavation and support installation).
VS* = w(x) dx = 2π ix wmax ⇒wmax = √S̅̅̅̅̅̅ (2)
ix 2π Attewell & Woodman (1982) also inferred that the longitudinal
settlement profile, along the tunnel axis, is well represented by the
where wmax is the maximum settlement in correspondence of the tunnel Gaussian cumulative probability distribution. Settlements evolve with
axis, ix represents the half-width of the settlement trough, defined as the the advancement of the excavation face and the steady-state condition is
horizontal distance between the tunnel axis and the point of inflection of reached when the maximum settlement (wmax) is completely developed.
the settlement curve, and V*s is the total volume of the surface settle-
ment trough (per unit length of tunnel), obtained by integrating

3
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 3. Schematic of the two cross-sections considered for settlements prediction: Section A at the face, Section B at the steady-state.

Fig. 4. Calculation stages: Phase 1, Section A (a), Phase 2 (b) and Phase 3, Section B (c).

Table 1
Summary of geometry data for the five case histories.
Tunnel Country z0 [m] Df [m] z0/Df [-] zw [m] Main references

a. 2nd Heinenoord NL 15.6 8.5 1.8 − 1.5 (Möller and Vermeer, 2008)
b. Milan – M1 ext. IT 11.5 6.54 1.8 − 11.5 (Migliazza et al, 2009)
c. Cairo L2 – Lot.12 EGY 18.0 9.4 1.9 − 3.0 (El-Nahhas and El-Mossallamy, 2009)
d. Cairo L2 – Lot.16 EGY 17.8 9.4 1.9 − 1.9 (El-Nahhas et al., 2015)
e. Naples – L6 IT 17.3 8.15 2.1 − 2.9 (Bilotta et al., 2017)

3. Finite element modelling general and flexible approach, suited to model any soil mechanical
behaviour, tunnel geometry and construction process.
In the detailed design phase, the empirical approach is generally This section focuses on FEM analysis of shield-driven tunnels in
complemented by finite element numerical modelling, which is a very plane-strain conditions. All the analyses were performed by the software

4
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Table 2
Summary of lining and shield data for the five case histories.
Tunnel Method sr [m] Lr [m] Lining segments [Nr.] D [m] Ls [m] gap [mm]

a. 2nd Heinenoord Slurry 0.35 1.5 7 8.3 8.0 200

b. Milan – M1 ext. EPB 0.30 1.4 n.a. 6.3 9.0 240
c. Cairo L2 – Lot.12 Slurry 0.40 1.5 7 9.15 9.0 250
d. Cairo L2 – Lot.16 Slurry 0.40 1.5 7 9.15 9.0 250
e. Naples – L6 EPB 0.30 1.7 9 7.85 n.a. 300

Fig. 5. Subsoil profiles and tunnel cross-sections: 2nd Heinenoord (a), Milan − M1 ext. (b), Cairo L2 − Lot.12 (c), Cairo L2 − Lot.16 (d) and Naples − L6 (e).

Plaxis 2D (Brinkgreve et al., 2017). A novel approach for 2D-modelling stiffness and isotropic hardening (Hardening Soil model with small
of tunnel excavation, called the “Strain applied method”, has been strain stiffness, HSs); a more detailed description of this last model is
developed, in which the three-dimensional effects induced by the actual given in the Appendix.
boring process are simulated by applying specific strain distributions
around the tunnel cross-section. The mechanical behaviour of the soil 3.1. Mesh dimensions and boundary conditions
will be modelled by both a linear-elastic perfectly plastic model (Mohr-
Coulomb, MC) and a more complex constitutive model with non-linear The typical situation to be modelled (Figure 2) is that of a single

5
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Table 3
Soil parameters of the MC model for the five case histories.
Tunnel Soil type γsat [kN/m3] c’ [kPa] φ’ [◦ ] ψ ’ [◦ ] υ’ [-] K0 [-]

a. 2nd Heinenoord Dense sand 20 0 35 0 0.3 0.47

b. Milan – M1 ext. Sandy gravel 22 0 30 10 0.3 0.50
c. Cairo L2 – Lot.12 Silty sand/Dense sand 19/20 0 30/37 0 0.35/0.3 0.50/0.40
d. Cairo L2 – Lot.16 Silty sand/Dense sand 19/20 0 30/37 0 0.35/0.3 0.50/0.40
e. Naples – L6 Pyroclastic soil/Marine sand 16/18 0 37/37 0 0.2/0.2 0.40/0.40

soil.
Table 4 The 3D effects of excavation face advance can be accounted for by
Soil parameters of the HSs model for the five case histories.
properly calibrating the tunnel volume losses, obtained by imposing
Tunnel Soil type Gref
0 Eref
50 γ0.7 [-] m suitable strains to a group of soil elements. The magnitude of these
[MPa] [MPa] [-] strains is treated as a “calibration parameter” of the procedure, to be
a. 2nd Dense sand 175 35 5⋅10− 4
0.5 back calculated in order to fit the measured settlements or to impose the
Heinenoord expected ground volume loss VS.
4
b. Milan – M1 Sandy gravel 250 48 1⋅10− 0.4
The main stages of the proposed calculation method are summarised
ext.
c. Cairo L2 – Silty sand/Dense 116/ 20/35 2.8⋅104/ 0.5 in Figure 4. The final effect of the whole excavation process is the result
Lot.12 sand 170 4.5⋅10− 4 of the volume losses imposed at the two control sections.
d. Cairo L2 – Silty sand/Dense 116/ 20/35 2.8⋅10− 4/ 0.5 In order to simulate the stress release at the tunnel face, Sec. A
Lot.16 sand 170 4.5⋅10− 4 (Figure 4a), it is possible to apply an in-plane isotropic deformation εxx
e. Naples – L6 Pyroclastic soil/ 180/90 47/40 1.9⋅10− 4/ 0.5
Marine sand 1.3⋅10− 4
= εzz to the soil volume to be excavated (πD2f /4): a volume reduction
must be produced; therefore, the imposed strains are negative. A similar
effect can be obtained, by applying a negative strain in the out-of-plane
tunnel bored in a coarse-grained soil (depth of tunnel axis, z0), with direction, εyy, as generally done in the following analyses. In the last
negligible disturbance of the initial pore pressure distribution (drained, case, the plane strain model prevents any out-of-plane deformation and,
hydrostatic conditions). Details on the specific case histories considered therefore, the effect is to generate a positive (tensile) stress, σyy, which in
in this study are reported in the next Section 4. turn induces a uniform negative in-plane deformation (shrinking) εxx =
The mesh dimensions (Figure 2a) are chosen in order that the arti- εzz.
ficial boundaries have negligible influence on the solution, particularly Note that for a linear elastic model the in-plane strains are simply
on extent and shape of the settlement trough. Tests with meshes of larger proportional, by the Poisson coefficient v, to the imposed longitudinal
transversal size confirmed that the assumed width (b/2 ≈ 5D) is satis- strain εyy (i.e., εxx = εzz = εyy/v). When the non-linear HSs model is used
factory. The horizontal displacement along the vertical boundaries has for the soil core to be excavated, a further effect must be considered. In
been restrained, while the bottom boundary is completely restrained. fact, the unloading produced by the imposed εyy is accompanied by a
reduction in stiffness of the soil core, proportional to the minimum
principal stress σ’3 ≈ σ ’yy. This effect can be accepted in that it can
3.2. Tunneling simulation – The proposed “Strain applied method”
represent the likely consequence of the remoulding of the soil around
the excavation head.
Plane strain methods currently utilized in engineering practice (i.e.
After the removal of the soil elements inside the tunnel section and
Convergence − Confinement method, Contraction method, Gap method,
the activation of the lining shell (Figure 4b), the soil-lining gap tends to
Grout Pressure, etc.) can only predict the “final conditions” (or steady-
close and the steady-state conditions are reached. This further volume
state situation, away from the tunnel face) and typically require the
loss is simulated by applying a εzz strain to a moon-shaped annulus of
calibration of a single parameter. Instead, in the Strain applied method
“disturbed soil” (Figure 4c). This annulus is of variable thickness, in fact,
herein introduced two different cross-sections, or excavation phases, can
it represents the real “ground-lining gap” (distance between the extrados
be represented: the stress and strain conditions at the excavation face
of the lining and the excavation profile) with maximum value at the
(Section A) and at the steady-state conditions (Section B).
crown and minimum (zero) at the invert. It is instrumental to model the
The schematic of the two modelling phases, corresponding to two
tail void, the effect of the conical shield and of the possible overcutting,
different “control sections” along the longitudinal profile of the surface
that is to say the main sources of tunnel volume loss, VL. A confirmation
settlement, is shown in Figure 3. Attewell et al. (1986) suggested that,
for the particular shape assumed for the gap can be found in Franza et al.
for open-face tunnelling, the settlement directly above the tunnel face
(2019), where a distinction is made between the deformation pattern
(Section A) is almost equal to half the maximum final settlement, 0.5
around the tunnel depending on the soil type: in coarse-grained soil the
wmax (Figure 3, solid line). However, surface settlements in Section A can
ground loss appears particularly localized at the tunnel crown, therefore
be significantly lower for closed-face shield, provided that the face
producing mainly vertical deformations.
pressure is regularly maintained (Figure 3, dashed line). Based on
The lining is represented by a ring of beam elements, having the
monitoring data, Mair and Taylor (1997) found that settlements as low
equivalent-stiffness of the precast segmental lining, connected to the soil
as 0.25–0.3 wmax are typical for closed-face tunnelling in coarse-grained

Table 5
Parameters for the empirical approach (Gaussian curves).
Tunnel wSec.A [mm] wSec.B [mm] ix [m] K = ix/z0 [-] VS Sec.A [%] ΔVS Sec.B [%] VS tot [%] VS Sec.A/VS tot [%]

a. 2nd Heinenoord 4 26 6.2 0.40 0.11 0.61 0.72 17.5

b. Milan – M1 ext. 7.5 13.6 5.2 0.45 0.29 0.24 0.53 61.5
c. Cairo L2 – Lot.12 2 11 7.2 0.40 0.05 0.23 0.29 21.3
d. Cairo L2 – Lot.16 6 18 7.7 0.40 0.17 0.33 0.50 36.0
e. Naples – L6 1.8* 6 7.8 0.45 0.07 0.16 0.23 32.0
*
Estimated as 0.3 wSec.B.

6
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 6. Comparison of measured and computed (MC and HSs model) surface settlement troughs at the tunnel face: 2nd Heinenoord (a), Milan – M1 ext. (b), Cairo L2
– Lot.12 (c), Cairo L2 – Lot.16 (d) and Naples – L6 (e).

clusters by interface elements (same elasto-plastic parameters of the soil Phase 2 − Lining installation: within the extrados of the lining
clusters). (diameter, D), the soil is excavated, the pore pressure is removed and
The modelling procedure can be summarized in the following the lining ring is installed (Figure 4b).
phases: Phase 3 − Section B: a vertical strain εzz (Figure 4c) is applied to the
ground-lining gap simulating the additional volume loss due to the
Phase 0 − Geostatic load: initial equilibrium conditions, characterized gap closure (ΔVL,Sec.B). The final increment in settlement and surface
by given K0 stress ratio and water table elevation. volume loss, far from the excavation face (steady-state conditions),
Phase 1 − Section A: an out-of-plane strain εyy (Figure 4a) is applied to are obtained (wSecB, ΔVS,Sec.B).
the excavation section (diameter, Df) simulating the effects of the
tunnel volume loss (stress release) at face (VL,Sec.A) and the In terms of computational steps, it has been found convenient to
remoulding of the soil. At the end of this calculation step, the cor- separate Phase 2 and Phase 3, although in the effective construction
responding surface maximum settlement and surface volume loss are process, they can be considered as a single phase.
obtained (wSecA, VS,Sec.A). In the following, the two strain parameters (εyy, εzz) will be back-
calculated (“calibrated”) in order to fit the monitoring data of five
well-documented case histories in greenfield condition (see Section 4).

7
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 7. Comparison of measured and computed (MC and HSs models) surface settlement troughs at the steady-state: 2nd Heinenoord (a), Milan – M1 ext. (b), Cairo
L2 – Lot.12 (c), Cairo L2 – Lot.16 (d) and Naples – L6 (e).

The overall results of the numerical modelling can summarized by types were present below or above the tunnel depth. For all the cases a
the following output parameters: the surface volume loss at the face, VS, well detailed description was found in the literature, both for the
Sec.A, and the increment of the surface volume loss at the steady-state geotechnical parameters as well as for the settlements data. A brief
condition, ΔVS.Sec.B. The total volume loss at ground surface, at the description of each case study follows.
end of the boring process, is given by VS.Tot. = VS,Sec.A+ΔVS.Sec.B. The Second Heinenoord Tunnel (2nd Heinenoord), case (a), was
bored by means of a Slurry-Shield TBM in a layered sand deposit over-
4. Case histories considered in this study laying a sandy-clay layer few meters below the tunnel invert (Bakker
et al., 1999; van Jaarsveld et al., 1999; Möller and Vermeer, 2008). The
Five case histories were selected to test the performance of the Milan M1 extension (Milan – M1 ext.), case (b), was bored by an EPB-
aforementioned modelling approach by comparison of computed results shield machine in a homogeneous deposit of sandy-gravels (Migliazza
with field measurements. All the case studies refer to shallow tunnels et al, 2009). Two instrumented sections (Lot.12 and Lot.16) of the Cairo
(depth of tunnel axis, z0, in the range 11.5–18.0 m, diameter, D, in the L2 tunnel represent, respectively, the cases (c) and (d); they were
range 6.3–9.15 m, with a z0/D ratio of 1.8–2.2) bored with EPB or excavated by a Slurry-Shield. The Cairo L2 line passes through medium
Slurry-shield machines in greenfield condition. All tunnels were exca- dense to very dense sands overlaid by a finer cohesive transition layer
vated in coarse-grained soils, although in some cases also different soil (El-Nahhas and El-Mossallamy, 2009; El-Nahhas et al., 2015). The last

8
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 8. Comparison of measured and computed (HSs εxx-εzz and HSs εyy model) horizontal displacements with depth at the steady-state at a distance of 6, 10 and 16 m
from the axis of 2nd Heinenoord tunnel.

case (e) is the Naples L6 tunnel, bored by an EPB-shield TBM in alter- Gs = 0.7G0, parameter for stress-stiffness law (m).
nated pyroclastic soils and marine sands (Bilotta and Russo, 2012;
Bilotta et al., 2017; Celentano et al., 2017). 5. Comparison of settlement results
In all cases, the water table is close to the ground surface, only for
Milan − M1 ext. the water table is at the same elevation of the tunnel The reliability of the proposed FEM modelling approach for the
axis. Moreover, the installed segmental linings are endowed with prediction of settlements was assessed, essentially, by comparing the
watertight joints and can be considered as an impervious boundary. computed settlement troughs with the empirical predictions (Gaussian
Table 1 lists the main geometry data for the five case histories: z0 is curve, see Equation (1)). The Gaussian curve was not used for settlement
the depth of the tunnel axis, Df the excavation diameter and zw the depth prediction but simply for fitting the monitoring data, imposing that the
from ground surface of the piezometric surface. curve passes through the point of maximum settlement measured above
Table 2 summarizes the parameters of the lining and the shield the tunnel axis (wSec.B). However, due to the limited data available in the
characteristics: sr is the lining thickness, Lr the lining ring length, D the cross-section of the tunnel face (Section A), the shape of the settlement
extrados diameter of the lining, Ls the shield length; the “gap” parameter trough in the transversal direction was assumed fixed, i.e., considering
represents the difference between the nominal diameter of excavation also for the tunnel face the same width parameter (K) calibrated for the
and the external diameter of the lining. steady-state condition (Section B).
Figure 5 shows the typical subsoil profile and the tunnel geometry of Table 5 reports the parameters of the Gaussian curves obtained by
the five case studies. field data fitting; wSec.A is the surface settlement measured above the
For each case, it is possible to represent the soil behaviour both by an tunnel face, wSec.B is the maximum final settlement measured in steady-
elasto-plastic model (Mohr-Coulomb, MC) and by the Hardening Soil state condition.
model with small strain stiffness (HSs), first described in Benz (2007). In the following subsections, the effects of the specific soil consti-
The mechanical parameters utilized for the main sandy layers of each tutive model on the induced settlements are discussed; the results are
case study are briefly summarized in the following. The reader should presented for both the face (Subsection 5.1) and the steady-state con-
refer to the mentioned literature for further details and insight into the ditions (Subsection 5.2).
case studies. Table 3 lists the soil parameters of the MC model inferred
from geotechnical investigations: γsat total unit weight, c’ cohesion, φ’ 5.1. Face conditions (Section A)
friction angle, ψ ’ dilatancy angle, υ’ Poisson’s ratio and K0 coefficient of
earth pressure at rest. The additional geotechnical parameters for the Figure 6 plots the results at the tunnel face. The settlement troughs
HSs model are reported in Table 4: small-strain shear modulus (Gref 0 ) at estimated by the FEM modelling are not always in agreement with the
the reference confining pressure (pref = 100 kPa), reference secant empirical profile. In particular, the calculated profiles (MC and HSs
stiffness modulus (Eref50), shear strain (γ 0.7) at which secant shear modulus model) for cases 6a, 6c and 6e tend to overestimate the width of the

9
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 9. Comparison of measured and computed (HSs εyy and HSs εxx − εzz) horizontal displacements with depth at the steady-state at 6, 10 and 16 m from the axis of
Cairo L2 – Lot.16 tunnel.

Fig. 10. Comparison of measured and computed (HSs model) surface horizontal displacements at the steady-state: 2nd Heinenoord (a) and Cairo L2 – Lot.16 (b).

empirical trough, a good agreement is found only for cases 6b and 6d, 5.2. Steady-state conditions (Section B)
particularly if the non-linear HSs model is utilized. However, as already
remarked, the Gaussian curve does not represent an interpolation of Figure 7 shows the settlement profiles obtained by the empirical
measured settlements at different distance from the tunnel axis but is (Gaussian) and numerical (FEM) approaches for the steady-state con-
only anchored to a single point: the maximum settlement wSec.A, above ditions, when the maximum final settlement is completely developed.
the tunnel axis. The numerical results for both the constitutive models are compared

10
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 11. Computed ground volume loss at the surface against tunnel volume loss for each case history: Section A (a) and Section B (b). The equations of linear
regression are given in the VL Sec.A-VS Sec.B, ΔVL Sec.A-εyy, ΔVL Sec.A-ΔVS Sec.B and ΔVL Sec.A-εzz planes.

Table 6
Computed volume losses at the surface and at the tunnel axis for the five case histories.
VS Sec.A [%] VL Sec.A [%] ΔVS Sec.B [%] ΔVL Sec.B [%] VS tot [%] VL tot [%] VS Sec.A/VS tot [-] VS tot/VL tot [-]

a. 2nd Heinenoord 0.15 0.08 0.72 0.70 0.88 0.78 0.17 1.13
b. Milan – M1 ext. 0.32 0.19 0.21 0.24 0.53 0.43 0.60 1.23
c. Cairo L2 – Lot.12 0.07 0.04 0.31 0.34 0.38 0.38 0.18 1.00
d. Cairo L2 – Lot.16 0.20 0.11 0.35 0.36 0.55 0.47 0.36 1.17
e. Naples – L6 0.09 0.04 0.20 0.23 0.29 0.28 0.31 1.04

Fig. 12. Subsoil profile and tunnel cross-section: Brescia Metrobus (a) and Milan – M5 (b).

11
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Table 7
Summary of geometry data for Brescia Metrobus and Milan – M5.
Tunnel z0 [m] Df [m] z0/Df [-] zw [m] Main reference

a. Brescia Metrobus 21.5 9.15 2.3 − 21.0 (Sanzeni et al., 2017)

b. Milan – M5 15.0 6.70 2.2 − 15.0 (Fargnoli et al., 2015)

Table 8
Summary of lining and shield data for Brescia Metrobus and Milan – M5.
Tunnel Method sr [m] Lr [m] Lining segments [Nr.] D [m] Ls [-] Gap [mm]

a. Brescia Metrobus EPB 0.35 n.a. 7 8.85 n.a. 300

b. Milan – M5 EPB 0.30 1.4 6 6.40 10 300

with field data. good prediction of the final settlement trough while the intermediate
The first observation concerns the results obtained by the HSs model, stages present greater challenges. An imposed strain method based on a
that are in better agreement with both the measurement points and the single calibration parameter (e.g., the application of εyy strain, as in the
empirical curve, with respect to the results for Section A. Instead, the previous calculation Phase 1) could be considered as a convenient
settlement profiles obtained by the MC model (in particular, for cases 7a alternative to the proposed method. However, it was found that the two-
and 7d) tend to slightly overestimate the width of the settlement trough, parameters modelling approach, besides representing the actual defor-
and consequently also the volume loss, as already noticed by Adden- mation mechanism better, gives more satisfactory results (Schiena,
brooke et al. (1997) and Hejazi et al. (2008). 2020).
Therefore, it seems quite affordable, based on available data, to get a

Table 9
Soil parameters of the HSs model for Brescia Metrobus and Milan – M5.
Tunnel Soil type γsat [kN/m3] c’ [kPa] φ’ [◦ ] ψ ’ [◦ ] υ’ [-] K0 [-] Gref
0 [MPa] Eref
50 [MPa] γ0.7 [-] M [-]
4
a. Brescia Metrobus Gravel 20 0 36 0 0.3 0.41 210 65 1⋅10− 0.4
4
b. Milan – M5 Sandy gravel 20 0 33 0 0.3 0.45 250 48 1⋅10− 0.4

Fig. 13. Comparison of measured and computed (HSs model) surface settlement troughs at the face cross-section (Section A): Brescia Metrobus (a) and Milan M5 (b).

Fig. 14. Comparison of measured and computed (HSs model) surface settlement troughs at steady-state (Section B): Brescia Metrobus (a) and Milan − M5 (b).

12
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Table 10
Parameters for the empirical approach (Gaussian curve): Brescia Metrobus and Milan – M5 tunnels.
Tunnel wSec.A [mm] wSec.B [mm] ix [m] K = ix/z0 [-] VS Sec.A [%] ΔVS Sec.B [%] VS tot [%] VS Sec.A/VS tot [%]

a. Brescia Metrobus 2.6 13 8.0 0.37 0.08 0.32 0.40 26.5

b. Milan M5 1.2 6.2 6.0 0.4 0.05 0.22 0.27 25.0

Fig. 15. Computed volume loss at the surface against tunnel volume loss for Brescia Metrobus and Milan − M5, compared to the average trend of the previous cases:
Section A (a) and Section B (b).

6. Comparison of horizontal displacements induces a displacement distribution in the soil similar to the Contraction
and the Stress Reduction methods and different from the Grout Pressure
Vertical profiles of horizontal displacements were available only for method.
two of the analysed case studies (i.e., 2nd Heinenoord and Cairo L2 – For the second case (i.e., Cairo L2 – Lot.16), horizontal displacement
Lot. 16). Since the calibration of the ε-method was performed with data are available only for a single vertical located at 9.4 m from the
respect to the maximum settlement at ground surface, an inspection of tunnel axis (Figure 9). The vertical profile from FEM analysis closely
the horizontal displacements within the soil can be considered as a resembles the middle profile of the 2nd Heinenoord case, which is
further validation of the method. located at a similar distance (10 m) from the tunnel axis, but in this case
For the 2nd Heinenoord case, three monitoring sections were the agreement with monitoring data seems less accurate: while the
implemented: located 6, 10 and 16 m apart from the tunnel axis. simulated and measured curves are in very good agreement below 10 m
Figure 8 shows that FEM-simulated and measured displacements are in depth (i.e., 7.8 m above the tunnel axis), they substantially diverge at
very good agreement, at least for the two furthest verticals, while for the shallow depth. However, for the Cairo L2 case, a mismatch with moni-
vertical at 6 m distance, a good agreement is found only below the toring data was observed also in the settlement trough (Figure 7d).
invert. The influence of two possible methods for applying the volu- A possible explanation to these findings can rely on a not-optimal
metric strains in Phase 1 (i.e., the basic εyy and the alternative εxx-εzz estimation of the HSs model parameters or more likely, on the influ-
method, as defined in Section 3) were also investigated. At the tunnel ence of the consolidation process that could have taken place in the
level, the displacement profiles calculated by both the methods exhibit a silt–clay layer, close to the surface, also considering that the water table
bulging towards the excavation, with a maximum at the sidewalls is just 1.9 m below the ground surface. In fact, the non-zero settlements
elevation, which is nearly absent in the real displacement profile. In the observed at a considerable distance from the tunnel axis are consistent
portion between sidewalls and crown elevations the curves of the with the superimposed effect of uniformly distributed consolidation
simulated settlements show an inversion in gradient. Finally, above the settlement of the silt–clay layer In El-Nahhas et al. (2015), the horizontal
crown elevation, the calculated displacements are less than the recorded displacement profile predicted by the Contraction method and the HSs
ones, but the trend is however consistent. In other words, the horizontal model, is compared with the actual displacements. The two curves are
displacements appear more localized around the cavity in the FEM generally in good agreement for what concerns the displacements at
model, while they are more uniformly distributed in the real case. ground surface. On the contrary, the FEM-3D model by El-Nahhas et al.
A significant difference between the two possible methods for (2015) fails to capture the displacement distribution at the tunnel depth,
applying isotropic volumetric strains can be observed in the magnitude well predicted by the 2D model in this paper.
of bulging, with εyy producing the largest displacements owing to the A comparison between calculated and measured horizontal dis-
stiffness reduction induced by the out-of-plane prescribed strains. With placements at ground surface is shown in Figure 10. The Gaussian curve
this respect, the εxx-εzz method seems in better agreement with real data. was calculated from settlement data after Equation (4) in Section 2. For
A comparison with the calculated profiles in Möller and Vermeer (2008) the 2nd Heinenoord case, the FEM model offers a better estimation than
shows that the herein proposed ε-method (either applying εyy or εxx-εzz) the empirical Gaussian approach, but underestimates the horizontal

13
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 16. Application of the “Predictive lines” of Equations (6-9) for obtaining the imposed strain to feed the FEM model after a trial-and-error iterative procedure (a,
b, c, d); verification of the FEM prediction against the monitoring data (Milan – M5 monitoring section S897.3, Fargnoli et al., 2013).

displacements near the tunnel. For the Cairo L2 – Lot.16 case, both the 7. Comparative analysis of the volume losses
FEM model and the Gaussian approach are less satisfactory. It seems that
the Gaussian curve would fit better the monitoring data if a lower ix This section provides a comparative analysis of the different case
value were adopted and the contribution of consolidation settlement histories in terms of computed volume losses. Figure 11 plots the ground
excluded. volume loss at the surface (VS) against the volume loss at the tunnel axis

14
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 17. Surface settlement troughs (a) and normalized surface settlement troughs (b) for different values of dilatancy angle (ψ = 0◦ , ψ = 10◦ , ψ = 20◦ ).

Fig. 18. Volumetric strain εv related to deviatoric strain γs at the crown and at the sidewall for different values of dilatancy angle (ψ = 0◦ , ψ = 10◦ , ψ = 20◦ ).

(VL) after the imposition of prescribed strains. A close linear correlation was also observed between tunnel volume
Figure 11a (face condition, Section A) shows the amount of surface losses and imposed strains (εyy and εzz). The regression lines, Equation
volume loss VS Sec,A as a function of tunnel volume loss VL Sec.A, given by (8) and (9), are plotted as red lines in Figures 11a and 11b, for face and
the imposed εyy. The corresponding parameters for the steady-state steady state conditions, respectively
condition (Section B) are represented in Figure 11b. The excavation of
the soil inside the tunnel diameter and the imposed vertical strain, εzz, at εyy = − 1.79 VL Sec.A [%] (8)
the soil-shield gap (related to the increment in tunnel volume loss, ΔVL
Sec.B) produce an increment in the surface volume loss (ΔVS Sec.B).
εzz = − 25.53 ΔVL Sec.B + 2.91 [%] (9)
By the procedure reported in Appendix B, a specific point in the
graphs of Figure 11 can be obtained, one for each case study. All the Table 6 summarizes the values of the computed surface volume losses
points seem well positioned along a line (the solid black line, hereafter (VS) and tunnel volume losses (VL) for the five case histories, both at the
called “correlation line”): the equations are reported also on the right of face (VSec.A) and at the final steady-state condition (Vtot); the volume
the graphs. In particular, Equations (6) and (7) represent respectively variations ΔVSec.B is also reported. The effects of dilatancy on the volume
the “face” and “steady-state” conditions, by assuming a non-linear losses will be further investigated in Section 9. For coarse-grained soils,
stiffness soil response (HSs model). Very high R2 coefficients were the surface volume loss can be lower or higher than the tunnel volume
found for both the correlations. loss depending on the volumetric response of the soil mass, whether
dilative or contractive behaviour prevails.
VS Sec.A = 1.63 VL Sec.A + 0.01 [%] (6) From the last column of the previous Table 6 it can be noted that the
proportion between surface volume losses before the excavation face
ΔVS Sec. B = 1.12 ΔVL Sec. B − 0.06 [%] (7) (Section A) and in steady-state conditions (Section B) it is not always

15
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 19. Contours of the computed total volumetric strain (εv) at the steady-state: ψ = 0◦ (a) and ψ = 20◦ (b).

Fig. 20. Contours of the computed total deviatoric strain (γs) at the steady-state: ψ = 0◦ (a) and ψ = 20◦ (b).

constant but it varies in a quite wide range, from 17 % to 60 %. The aforementioned correlations, two further case histories of tunnels
mean value from this case selection can be calculated as 32 %, while it recently excavated in Italy were analyzed: the Brescia Metrobus
descends to 25.5 % if the Milan – M1 ext. case is excluded from (Sanzeni et al., 2010, Sanzeni et al., 2017) and the Milan M5 metro line
calculation. (Fargnoli et al., 2013; Fargnoli et al., 2015). Both tunnels were exca-
Farrell (2010) inferred that at low tunnel volume losses (VL<1.0 %) vated by EPB-shields in gravelly-sand soil deposits, with the water table
there is little difference with VS: this fact was particularly noticed in a approximately located at the elevation of the tunnel axis.
series of centrifuge tests. A similar behaviour was observed for the case The soil profile and tunnel geometry of Brescia Metrobus and Milan –
histories considered in this study (Table 6). In fact, the ratio between the M5 are represented in Figure 12. In particular, field measurements and
final VL tot and VS tot is within the 1.00 – 1.23 interval. geotechnical parameters of the sections SCBF8 (Brescia Metrobus) and
S942.2 (Milan – M5) were respectively considered. For additional details
8. Correlations as forecasting tool the reader is referred to the publications mentioned above.
Table 7 and Table 8 summarize the principal geometric data and soil
The correlations obtained in the previous Section can be useful in the parameters considered for each tunnel, Table 9 the mechanical param-
context of 2D modelling for tunnel design. In fact, Equations (6-9), eters of the HSs soil model.
which summarize the average response found in the five case histories, Figure 13 and Figure 14 show the surface settlements of the new case
can be employed as a guideline to estimate model parameters for tunnels histories: measured data, respectively for Section A and Section B, are
bored in similar conditions. To test the predictive capability of the compared with model results and the empirical curves. The parameters

16
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 21. Stress-paths in the t-s’ plane at crown (a) and sidewall (b) and in the q̃-p’ plane at crown (c) and sidewall (d) for different values of the over-consolidation
ratio (OCR = 1, OCR = 10) for the Milan – M1 ext. tunnel.

of the empirical relationship were defined by data fitting, imposing that alarm value, which can be used to analyse and predict the effects
the curve passes through the maximum settlement point, wSec.B, above induced by a new tunnelling project: generally, it is in the 0.5 – 1.0 %
the tunnel axis (Table 10). For the first case, Brescia Metrobus, upper range (Lunardi et al., 2015; Franza et al., 2020).
and lower limits for the settlements at Section A (tunnel face section) An example of the possible use of Equations (6-9) as a forecasting
were available from monitoring data, referred as day 3 and day 6, while tool is given in Figure 16, referring to the geotechnical conditions of the
the passage of the face through the section occurred approximately previously analysed Milan – M5 case.
during day 4. The modelling results are in good agreement with field In short, given a target total volume loss from design specification (e.
data, both in terms of shape of settlement trough and volume loss, and g. VS tot = 0.5 %), one should enter the two graphs in Figures 16a and 16b
generally offer a better approximation than the empirical method. If the with a trail subdivision of VS tot in VS Sec.A and ΔVS Sec.B (e.g. 25–75 %).
volume losses calculated for Brescia Metrobus and Milan – M5 are re- Note that a VS SecA equal to the 25 % of VS tot represents to the mean value
ported with the results of the 5 preceding cases, the new points are obtained for the 5 cases of Table 6. Following the arrows, the corre-
aligned with the correlation lines or very close to them (Figure 15). The sponding values of εyy and εzz with which to feed the FEM model can be
correlation formulas of Equations (6-9) are therefore confirmed in their derived. Thereafter, the computed VS Sec.A and ΔVS Sec.B can be compared
predictive potential. with the trial inputs: whether the two values are in good agreement or
Given the promising results, it is feasible to use the correlations of not, the process ends or a new trial subdivision is attempted until a
Equations –(6-9) to guide the modelling process for Class A predictions, satisfactory matching is found (Figures 16c and 16d). The next trial
i.e., for predictive analyses of the settlement trough performed at the subdivision can be driven by the difference between the output and
design stage, before monitoring data are available. In fact, the strain input ΔVS Sec.B value of the previous iteration by applying the following
parameters εyy and εzz required by the novel method can be immediately relation:
obtained from the “Predictive lines” of Equations (8-9), for a given [ ]
“design value” of volume loss VS, and consequently of VL (Equations (6) (ΔVS Sec.B )ℓinput
+1
= (ΔVS Sec.B )ℓinput − (ΔVS Sec.B )ℓoutput − (ΔVS Sec.B )ℓinput [%]
and (7)). The “design value” of VS tot (the sum of VS SecA and ΔVS SecB) is (10)
typically intended to represent the maximum acceptable value, or an

17
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Fig. 22. Plastic stress points at the steady-state for OCR = 1 (a) and OCR = 10 (b).

Fig. 23. Surface settlement troughs (a) and normalized surface settlement troughs (b) for OCR = 1 and OCR = 10.

where the superscript indicates the iteration number (ℓ) and the according to Rowe’s theory (Rowe, 1962), at least for large values of
subscript distinguishes the trial input from the output of the FEM calcu- mobilized friction, or by the approach proposed by Li and Dafalias
lation. The portion of volume loss to be applied to the Section A phase is (2000) when the Rowe’s formula gives a negative angle.
calculated according to VS Sec.A = VS tot − ΔVS Sec.B. In this example For most of the preceding case histories, the dilatancy angle ψ was set
(Milan – M5 case), the ratio between VS Sec.A and VS tot at the end of the equal to zero, in the absence of specific experimental determinations;
iterative procedure is around 57.5 %, very similar to the result of the only for the Milan – M1 ext. case a small value of ψ = 10◦ was assumed. It
Milan – M1 ext. case, reported in Table 6. is therefore worthwhile to analyze the influence of the dilatancy angle
Fargnoli et al. (2013) analysed a monitoring section (S897.3) of on the volume loss. To this aim, again the case of Milan – M1 ext. was
Milan – M5, close to the previously analysed section (S942.2) and with selected, considering three different values of ψ (0, 10 and 20◦ ).
similar geotechnical conditions. For this section, the monitoring data The first observation concerns the surface settlement profiles: with
suggested a total volume loss of 0.48 %, very close to the input value of increasing values of dilatancy the maximum settlement above the tunnel
0.5 % assumed in the calculation example. Figure 16e shows a nearly axis decreases (Figure 17a). However, the dilatancy effect does not
perfect matching between the trough predicted by FEM modelling and change the shape of the settlement curves: in fact, in each case, the in-
the monitoring data and, more in general, confirms the strength of the flection point is approximately located at the same distance from the
proposed approach. tunnel axis and the scaled settlement curves (Figure 17b), with respect
to the specific wmax, are similar.
9. Influence of dilatancy and volumetric cap Figure 18 shows the calculated path of the total volume strain (εv)
versus the deviatoric strain (γs) for two soil elements, located at a dis-
9.1. Dilatancy tance of only 1.5 m from the crown and the sidewall of the tunnel,
therefore in zones markedly affected by the excavation effect and likely
In the HSs model the mobilized dilatancy angle can be calculated to experience large plastic strains. The initial stage (Phase 0) is located

18
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

distance of 1.5 m from the crown and the sidewall, for OCR = 1 and 10.
Moreover, the initial K0 was always calculated with the Jaki formula, 1 –
sinφ’, not dependent on OCR, to ensure that the initial stress state re-
mains unchanged (same initial point O for the stress paths).
At the crown (Figure 21a), the first calculation stage (point A) de-
termines a reduction in both mean, s’ = (σ1′+σ 3′)/2, and deviator stress,
t = (σ 1′ − σ3′)/2. During the following stages (up to the final point B),
while the mean stress s’ remains almost constant, the deviator stress t
increases; however, the OCR does not affect the stress-paths at crown
significantly. On the contrary, the stress-paths at the sidewall
(Figure 21b) for OCR = 1 and 10 differ remarkably: at the end of first
stage, the increment in mean and deviator stress for OCR = 1 is only half
of that observed for OCR = 10, owing to the reaching of the cap yield
surface. In the following stages, the more relevant variation is the
reduction in s’.
Fig. 24. Surface volume loss (VS,Sec.B) related to the tunnel volume loss (VL,Sec.
The stress paths of the same elements are also represented
B) for different angles of dilatancy (ψ = 0 , ψ = 10 , ψ = 20 ) at the steady-state
◦ ◦ ◦

condition for tunnel Milan – M1 ext. in comparison with all the other cases.
(Figures 21c and 21d) in the plane (q̃, p’), in which the cap yield function
is defined, with q̃= σ 1′ + (α-1) σ 2′ − ασ 3′, α = (3 + sinφ’)/(3-sinφ’), and
p’ = (σ 1′ + σ 2′ + σ 3′)/3. At the crown, the stress paths are similar and
at the origin and represents the undisturbed condition. The intermediate
develop in the elastic domain. At the sidewall, while the stress path for
Phase corresponding to the face passage is labelled with the letter A, the
OCR = 1, moving outwards of the initial yield cap, is influenced by the
final Phase with the letter B.
compaction hardening of the soil, the stress path for OCR = 10 develops
At the first calculation step, the soil element at crown expands
within the very large boundaries of its elastic domain (not represented in
(Figure 18, element C). The path is similar for the three cases up to the
Figure 21). It is worth noting that in this last case the initial state of stress
point A. After that (i.e., when vertical strains are applied to the soil-
does not lie on the initial yield cap, because the same K0 value has been
lining gap), the path separates into three branches: in case of null
assumed, independently of the different OCR.
dilatancy the preceding expansion is almost recovered thanks to a
The effects of OCR on the stress state is also recognizable in
contractive phase; with higher dilatancy angles the volumetric strain
Figure 22, in terms of plastic stress points. Red squares indicate that the
remains stable (ψ = 10◦ ) or keeps increasing (ψ = 20◦ ), reducing the
stress lies on the failure surface, blue triangles represent a stress state
maximum settlement above the tunnel axis. On the opposite, the side-
that lies on the yield cap surface, brown diamonds lie on the shear and
wall element initially exhibits a contraction (Figure 18, element S),
cap hardening envelope and green triangles lie on the shear-hardening
which is partially or almost completely recovered in the next phase,
envelope. When a large OCR is imposed (OCR = 10, Figure 22b) and
depending on the dilatancy angle.
the activation of the cap yield mechanism is excluded, the plastic strains
For the same case study (Milan – M1 ext.), a potential calibration of
are mainly concentrated around the tunnel, where the shear hardening
the strain parameters under the hypothesis of ψ = 0◦ would have led to
activates and the failure condition is reached.
lower values of εyy and εzz as observed for the for Naples – L6 case. In this
Reflecting the different distribution of plastic points, when the cap
respect, a careful assessment of relative density seems crucial for the
yield surface is deactivated, the ground surface settlements tend to
coarse soil layer in which the tunnel is excavated.
reduce (Figure 23a) and the normalized curve becomes (Figure 23b)
Figure 19 shows the contours of the total volumetric strain εv for the
narrower (i.e., the inflection point is closer to the tunnel axis). The in-
two limit values of dilatancy (ψ = 0◦ and 20◦ ). Increasing ψ , the
fluence of the OCR is therefore different from that of the dilation angle,
expansion strains (positive) at the sidewalls increase and positive
as it not only affects the magnitude, but also the shape of the settlement
volumetric strains also occur in a thin band of soil above the tunnel
curve. A proper assessment of OCR is therefore particularly relevant for
crown. When a non-dilative behaviour is assumed, expansion strains can
a correct prediction of the settlement trough.
be observed only in a small wedge of soil above the crown, other than
below sidewalls and invert; the zone with more pronounced contraction
9.3. Effect of plastic deformation on volume losses
strains is located above the springline at some distance from the tunnel
wall. The deviatoric strain γ s (Figure 20) are mainly concentrated in a
It is now possible to thoroughly analyse the correlation between the
wing-shaped zone of soil starting from the sidewalls, in both the
surface volume loss VS and the tunnel volume loss VL at the steady state
considered cases. In the non-dilative case, the deviatoric strains are
condition. To this aim, the total volume losses at the end of the analyses
higher and the wing zone grows up towards the ground surface.
were considered, instead of analysing the partial effects at Section A and
Section B. The final (VL, VS) points in Figure 24 show that the overall soil
9.2. Volumetric cap response remains contractive for a null or low dilatancy angle (ψ = 0,
10◦ ), while it becomes substantially neutral when a larger dilatancy
According to HSs model, the dilatancy angle is not the only param- angle is considered (ψ = 20◦ ). The deactivation of the compaction
eter that may induce volume changes in the soil. In fact, irrecoverable hardening (OCR = 10) causes a dilative behaviour in the soil, for a ψ =
volumetric strains may also be a consequence of compaction hardening 10◦ . The foregoing sensitivity analysis was performed specifically for the
under isotropic compression (see Appendix A), defined by means of a case of the Milan – M1 ext. tunnel. For all the other tunnels considered in
volumetric cap yield surface whose magnitude is a function of the this paper, for which ψ = 0◦ and OCR = 1, the final (VL, VS) points
isotropic preconsolidation stress, pp. The initial value of pp depends on demonstrate that the overall response is generally contractive or, at the
the assigned over-consolidation ratio (OCR). Therefore, it is possible to maximum, neutral.
inhibit the initiation of compaction hardening by fictitiously increasing This general trend is clearly observed in Figure 24, where the dashed
the OCR. line separates two portions of the graph: the upper area characterised by
A new analysis was therefore carried out purposely assuming OCR = a contractive behaviour of the soil volume and the lower area where a
10, sufficient to completely inhibit the cap-yielding mechanism and, dilative behaviour prevails. It is possible to infer that the contractive
therefore, to appreciate its influence by comparison with the case OCR response generally prevails, also for a non-null value of dilatancy, due to
= 1. Figure 20 shows the stress-paths followed by two soil elements at a the effect of the compaction hardening in the soil mass. This behaviour

19
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

should be carefully considered when a dense sand is found, or simply the routinely sequence of lining installation and grouting. In fact, all the
prevented by assuming a high OCR value, as shown in this Section. aforementioned features, although impossible to be explicitly included
in a simple FEM − 2D model, may be relevant for the correct prediction
10. Conclusions of tunnelling effects in coarse-grained soils.
Out of the purposes of this paper, the “Strain applied method” was
The paper presents a specific approach, called the “Strain applied further validated by comparing it with other simplified methods for
method”, for FEM-2D modelling of soil-shield interaction in two calcu- modelling the tunnel face advance in FEM − 2D analyses. In many cases,
lation phases, useful for a simplified prediction of surface settlements the new approach provided better outcomes in terms of surface dis-
caused by closed-face tunnelling. The overall volume loss at the tunnel placements field. Moreover, the practical application of the “Strain
boundary is obtained by applying volumetric strains to specific soil applied method” can take advantage of the “Predictive lines” provided in
clusters. Two reference cross-sections of the tunnel can be analysed: the this study. Being based on a set of case histories sharing specific features
vertical plane passing through the tunnel face (Section A) and the (range of depth, range of excavation diameters, typology of shield, soil
steady-state conditions at distance from the face (Section B). The nu- type) the correlations found cannot be extended to tunnelling situations
merical analyses, carried out by the software Plaxis 2D, were performed that substantially diverge from those analysed.
with two soil models, the classic ideal elasto-plastic Mohr-Coulomb Moreover, the lack of monitoring data for the settlement trough near
model and the Hardening Soil with small strain stiffness model, charac- the tunnel face (i.e., during the transient excavation phase, before the
terized by non-linear stiffness and two hardening mechanisms governing final settlements are reached) did not allow a fully reliable comparison
the shear and volumetric pre-failure deformation of the soil. The influ- between measured and modelled results. In this regard, further re-
ence of dilatancy and compaction hardening on ground deformation finements of the method are also possible: for instance, a non-isotropic
were thoroughly analysed. deformation of the soil to be excavated in the first calculation phase
The study started with the calibration of the surface settlement (Section A, tunnel face cross-section) could be considered.
trough predicted by the proposed approach, through the comparison
with greenfield measurements, recorded in five well documented case CRediT authorship contribution statement
histories of shallow tunnels bored in coarse-grained soil. The settlement
troughs predicted by the HSs model are generally in better agreement F. Schiena: Writing – original draft, Formal analysis. E. Lusini:
with field observations and with the empirical Gaussian curve, both in Writing – original draft, Visualization, Formal analysis, Methodology. A.
terms of shape of the settlement trough and overall volume loss. Also the Lembo Fazio: Supervision, Methodology, Conceptualization. A. Gra-
horizontal displacement distribution can be well predicted by this FEM ziani: Writing – review & editing, Supervision, Methodology,
approach, particularly at some distance from the tunnel axis, while some Conceptualization.
mismatch was found for displacements very close to the tunnel walls.
Specific correlations were determined between the ground volume Declaration of competing interest
loss at surface, Vs, the tunnel volume loss, VL, and the strains, εyy and εzz,
to be imposed at the tunnel boundary. By a comparative analysis of the The authors declare that they have no known competing financial
selected case histories, it was possible to establish a set of trend lines, interests or personal relationships that could have appeared to influence
representative of the typical conditions for the current TBM tunnelling the work reported in this paper.
practice in coarse-grained deposits, which can be used as “Predictive
lines” for tunnel excavated in similar conditions. Two further case his- Data availability
tories of tunnelling in Italy were used as testing situations, to validate
the reliability of the previously found correlations. data used in the paper are derived from published works
These “Predictive lines” take into account the combined effect of a
large number of factors, such as: the typical features of TBM (slurry- and Acknowledgements
EPB- shields) and TBM operation (effective control of the face pressure,
steering mode and muck conditioning parameters, for EPB shields), the This research did not receive any specific grant from funding
generally accepted amount of ground loss for tunnelling in urban areas, agencies in the public, commercial, or not-for-profit sectors.

A. Appendix

A.1. Essentials of the Hardening soil model with small strain stiffness (HSs)

The elasto-plastic with isotropic hardening HSs model (Benz, 2007) is an extension of the Hardening Soil model (Schanz et al., 1999). The model
takes into account important features of soil behaviour, such as: high stiffness at small strain levels, development of plastic strains at first stages of
loading and soil stiffness decrease with increasing strain level. HSs derives from the hyperbolic function based on the hyperbolic stress–strain relation
proposed by Duncan and Chang (1963) in primary triaxial loading. The main parameters are:

• the resistance parameters c’ (cohesion), φ’ (friction angle), and ψ (dilatancy angle);

• the unloading/reloading Young’s modulus, Eur, the secant stiffness modulus, E50, the oedometric tangent stiffness, Eoed.

The nonlinear isotropic elastic behaviour of the soil is considered by using a stress-dependent Young’s modulus, which is a function of the effective
stress and strength parameters according to the following expression:
( )m
cʹ cotφʹ + σʹ3
Eur = Eur
ref
(A.1)
cʹ cotφʹ + pref

where Eref
ur is the unloading/reloading Young’s modulus at the reference pressure at 100 kPa (p
ref
= 100 kPa). Soil behaviour is ruled by the stiffness

20
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

moduli listed above and depends on the minor principal stress σ ’3, which is the confining effective stress, and on the parameter m that controls the
linear or non-linear response of the soil stiffness.
Similar relationships hold for the stress dependency of the secant stiffness modulus, E50, and the tangent stiffness for primary oedometer loading,
Eoed.
The HSs model considers two additional parameters, to those defined for the standard HS model, to take into account the dependency of soil
stiffness on the strain level:

• the small-strain shear modulus, G0;

• the shear strain, γ 0.7, at which the secant shear modulus is Gs = 0.7 G0.

The stress dependency of the small-strain shear modulus, G0, is expressed by a relationship similar to the ones discussed above (Equation A.1) for
the other stiffness.
( )m
cʹ cotφʹ + σʹ3
G0 = G0ref ʹ (A.2)
c cotφʹ + pref

The HSs model considers the modified version (Dos Santos and Correia, 2001) of the stiffness reduction curve proposed by Hardin and Drnevich
(1972):
GS 1
= ⃒ ⃒ a = 0.385
G0 ⃒ γ ⃒ (A.3)
1 + a⃒⃒ ⃒
⃒
γ 0.7

The small-strain stiffness curve is cut off at the unloading/reloading shear modulus Gur, defined as:
Eur
Gur = (A.4)
2(1 + υur )

where υur is the Poisson’s ratio for unloading/reloading.

The HSs model comprises two yield surfaces, not fixed in the principal stress space but which can evolve following a hardening rule. A shear
hardening yield surface, f s, which is a function of the plastic strains due to the primary deviatoric loading and a cap yield surface, f c, which depends on
the plastic volumetric strain due to primary compression. The flow rule is associated for states lying on the surface f c, whereas a non-associated flow
rule is adopted for states on the surface f s, by adopting a formulation based on stress-dilatancy theory.
The shear yield (f s) loci can expand up to the ultimate Mohr-Coulomb failure surface, while the cap surface (fc) can expand as a function of the pre-
consolidation stress, pp, which, in turn, depends on the initial OCR value. Schanz et al. (1999) give the complete description of the model.

B. Appendix

B.1. Volume loss calculation procedure

The deformed volume of soil clusters is internally calculated by the software Plaxis 2D. However, the built-in procedure is not available for the
deactivated soil clusters (e.g., the excavated volume) or the clusters with small volume and distorted elements (e.g., the lunette-shaped volume in
Figure 4).
The first issue was overcome by applying the ‘shoelace’ formula to the initial and deformed nodal coordinates of the plate, which represents the
final lining. The formula (Equation B.1) approximates the shape of the excavation profile with a polygon, connecting the N adjacent nodes, and
calculates the nominal volume, Vexcavated,n, from nodes coordinates (xj,yj) and the deformed volume Vexcavated,f, from the coordinates updated with
displacements (xj + ux, yj + uy). This formula introduce an approximation in the calculated volume, but the error is negligible when considering the
volume variation of Equation (B.2), if the number of nodes (N) is reasonably high.
⃒ ⃒
⃒ x x2 ⋯ xj ⋯ xN ⃒
2 Vexcavated = ⃒⃒ 1 ⃒ (B.1)
y1 y2 … yj … yN ⃒

ΔVexcavated = Vexcavated,f − Vexcavated,n (B.2)

where ΔVexcavated is the volume loss, j = 1…N is the number of the nodes of the plate representing the lining.
The volumetric deformation of the soil clusters was calculated by a Matlab script by the same method utilized in Plaxis 2D (Brinkgreve et al., 2017).
For each cluster the volume loss (or gain) is calculated as follow:
⎛ ⎛⎡ ⎤ ⎞⎞
∂N
M 3
(ξii , ηii ) ⎥
1 ∑⎜∑ ⎜ ⎜⎢
⎜⎢ ∂ξ ⎥
⎟⎟
⎟⎟
ΔVcluster = ⎜ Wii det⎜⎢ ⎥[ Xel Yel ] ⎟ ⎟ (B.3)
2 i=1 ⎝ ii=1 ⎝⎣ ∂N ⎦ ⎠⎠
(ξii , ηii )
∂η

where i = 1…M is the number of elements in each cluster, ii = 1…3 is the element Gauss point, Wii is the weight of the Gauss point, N(1x6) is the vector
of element shape functions, (ξ, η) are the coordinates of the six-node triangular element in the parent domain, Xel(6x1) and Yel(6x1) are the vectors of

21
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

element nodes in global coordinates.

The surface volume loss, VS, is calculated by Equation (B.4), while Equation (B.5) is used for obtaining the tunnel volume loss, VL
⎛ ⎞
∑
⎜ k ∈ Ω ΔVk + ΔVexcavated ⎟
VS = 100⎜
⎝ π D2
⎟[%]
⎠ (B.4)
f
4

⎛ ⎞
∑
⎜ k ∈ Π ΔVk + ΔVexcavated ⎟
VL = 100⎜
⎝ π D2f
⎟[%]
⎠ (B.5)
4

where k are the (active) soil clusters, Ω is the set of all active soil clusters and Π is the set of soil clusters to which the prescribed strain (εyy or εzz) is
applied.

References Fargnoli, V., Boldini, D., Amorosi, A., 2013. TBM tunnelling-induced settlements in
coarse-grained soils: the case of the new Milan underground line 5. Tunn. Undergr.
Space Technol. 38, 336–347.
Addenbrooke, T.I., Potts, D.M., Puzrin, A.M., 1997. The influence of pre-failure soil
Fargnoli, V., Boldini, D., Amorosi, A., 2015. Twin tunnel excavation in coarse grained
stiffness on the numerical analysis of tunnel construction. Géotechnique 47 (3),
soils: observations and numerical back-predictions under free field conditions and in
693–712.
presence of a surface structure. Tunn. Undergr. Space Technol. 49, 454–469.
Attewell, P.B., Yeates, J., Selby, A.R., 1986. Soil movements induced by tunnelling and
Farrell, R.P., 2010. Doctoral thesis: Tunnelling in sands and the response of building.
their effects on pipelines and structures. Blackie, Glasgow.
HYPERLINK “http://www.eng.cam.ac.uk/” Department of Engineering, University
Attewell, P.B., Woodman, J.P., 1982. Predicting the dynamics of ground settlement and
of Cambridge.
its derivatives caused by tunnelling in soil. Ground Eng. 15 (8), 13–22.
Franza, A., Marshall, A.M., Zhou, B., 2019. Greenfield tunnelling in sands: the effects of
Bakker, K.J., de Boer, F., Admiraal, J.B.M., van Jaarsveld, E.P., 1999. Monitoring pilot
soil density and relative depth. Géotechnique 69 (4), 297–307.
projects using bored tunnelling: the Second Heinenoord Tunnel and the Botlek Rail
Franza, A., Marshall, A.M., Zhou, B., Shirlaw, N., Boone, S., 2020. Greenfield tunnelling
Tunnel. Tunn. Undergr. Space Technol. 14 (2), 121–129.
in sands: the effects of soil density and relative depth (Discussion). Géotechnique 70
Benz, T., 2007. Doctoral thesis: Small-strain stiffness of soils and its numerical
(7), 639–646.
consequences. Institute of Geotechnical Engineering, University of Stuttgart.
Hardin, B.O., Drnevich, V.P., 1972. Shear modulus and damping in soils: design
Bilotta, E., Paolillo, A., Russo, G., Aversa, S., 2017. Displacements induced by tunnelling
equations and curves. J. Soil Mech. Found. Div. 98 (7), 667–692.
under a historical building. Tunn. Undergr. Space Technol. 61, 221–232.
Hejazi, Y., Dias, D., Kastner, R., 2008. Impact of constitutive models on the numerical
Bilotta, E., Russo, G., 2012. Ground movements induced by tunnel boring in Naples. In:
analysis of underground constructions. Acta Geotech. 3, 251–258.
Viggiani, C. (Ed.), Geotechnical Aspects of Underground Construction in Soft
ITA/AITES (International Tunnelling Association), 2000. Recommendations and
Ground. Taylor & Francis Group, London, pp. 979–986.
Guidelines for Tunnel Boring Machines – Working Group N.14 (Mechanized
Boldini, D., Spaggiari, C., Abul, J.K., Fuoco, S., Lusini, E., 2023. Class A predictions of
Tunnelling).
damage level in an historical fortress induced by twin tunnelling, in: Expanding
Kavvadas, M., Litsas, D., Vazaios, I., Fortsakis, P., 2017. Development of a 3D finite
underground - Knowledge and passion to make a positive impact on the world:
element model for shield EPB tunnelling. Tunn. Undergr. Space Technol. 65, 22–34.
Proceedings of the ITA-AITES World Tunnel Congress 2023 (WTC 2023), 12-18 May
Lambrughi, A., Medina Rodriguez, L., Castellanza, R., 2012. Development and validation
2023, Athens, Greece. CRC Press, London, pp. 1788-1795.
of a 3D numerical model for TBM–EPB mechanised excavations. Computers Ang
Boldini, D., Losacco, N., Franza, A., DeJong, M.J., Xu, J., Marshall, A.M., 2021.
Geotechnics 40, 97–113.
Tunneling-induced deformation of bare frame structures on sand: numerical study of
Lee, K.M., Rowe, R.K., Lo, K.Y., 1992. Subsidence owing to tunnelling. I. Estimating the
building deformations. J. Geotech. Geoenviron. Eng. 147 (11).
gap parameter. Can. Geotech. J. 29, 929–940.
Boscardin, M.D., Cording, E.J., 1989. Building response to excavation-induced
Li, X.S., Dafalias, Y.F., 2000. Dilatancy for Cohesionless Soils. Géotechnique 50 (4),
settlements. J. Geotech. Eng. 115(1), 1-21.
449–460.
Breth, H., Chambosse, G., 1975. Settlement behavior of buildingsabove subway tunnels
Lunardi, G., Mancinelli, L., Zimbaldi, A., Carini, M., 2015. Copenhagen Cityringen Metro:
in Frankfurt clay. In: Settlement of Structures: Conference Organised by the British
EPB-TBM head pressure definition, in. Proceedings of ITA-AITES World Tunnel
Geotechnical Society at the Lady Mitchell Hall, Cambridge Held in April 1974.
Congress 2015 (WTC 2015)/41st General Assembly.
Pentech Press, London, pp. 329–336.
Mair, R.J., Taylor, R.N., 1997. Theme lecture: Bored tunnelling in the urban
Brinkgreve, R.B.J., Engin, E., Swolfs, W.M., 2017. Plaxis 2D Manual. Balkema,
environment, in: Proceedings of the fourteenth international conference on soil
Rotterdam.
mechanics and foundation engineering, Hamburg, 6-12 September, 1997. Balkema,
Burd, H.J., Houlsby, G.T., Augarde, C.E., Liu, G., 2000. Modelling tunnelling-induced
Rotterdam, pp. 2353-2385.
settlement of masonry buildings. Proc. Inst. Civ. Eng. Geotech. Eng. 143 (1), 17–29.
Mair, R.J., 2003. Research on tunnelling-induced ground movements and their effects on
Celentano, P., Bilotta, E., Russo, G., Aversa, S., 2017. Effetti dello scavo di una galleria
buildings – lessons from the Jubilee Line Extension, in: Jardine, F.M. (Ed.), Response
urbana su un edificio intelaiato, in: Moraci, N., Soccodato, C. (Eds.), La Geotecnica
of buildings to excavation-induced ground movements: Proceedings of the
nella conservazione e tutela del patrimonio costruito: XXVI Convegno Nazionale di
international conference held at Imperial College, London, UK on 17-18 July 2001.
Geotecnica - Roma 2017. AGI, pp. 527-534.
CIRIA, pp. 3-26.
Celestino, T.B., Gomes, R.A.M.P., Bortolucci, A.A., 2000. Errors in ground distortions due
Marshall, A.M., Farrell, R., Klar, A., Mair, R., 2012. Tunnels in sands: the effect of size,
to settlement trough adjustment. Tunn. Undergr. Space Technol. 15 (1), 97–100.
depth and volume loss on greenfield displacements. Géotechnique 62 (5), 385–399.
DAUB (Deutscher Ausschussfür Unterirdisches Bauen), 2021. Recommendations for the
Migliazza, M., Chiorboli, M., Giani, G.P., 2009. Comparison of analytical method, 3D
Selection of Tunnel Boring Machines.
finite element model with experimental subsidence measurements resulting from the
Dos Santos, J.A., Correia, A.G., 2001. Reference threshold shear strain of soil. Its
extension of the Milan underground. Comput. Geotech. 36, 113–124.
application to obtain an unique strain-dependent shear modulus curve for soil, in:
Möller, S.C., Vermeer, P.A., 2008. On numerical simulation of tunnel installation. Tunn.
Proceedings of the Fifteenth International Conference on Soil Mechanics and
Undergr. Space Technol. 23 (4), 461–475.
Geotechnical Engineering, Istanbul, 27-31 August 2001. Balkema, Rotterdam, pp.
Möller, S.C., 2006. Doctoral thesis: Tunnel induced settlements and structural forces in
267-270.
linings. Institute of Geotechnical Engineering, University of Stuttgart.
Duhme, R., Lee, J., 2021. Latest Developments in Shield TBM Selections & Design for
Ochmański, M., Modoni, G., Bzówka, J., 2018. Automated numerical modelling for the
Mechanized Tunneling. in: Proc. of the 2021 World Congress on Advances in
control of EPB technology. Tunn. Undergr. Space Technol. 75, 117–128.
Structural Engineering and Mechanics, Seoul, 24-26 August 2021.
O’Reilly, M.P., New, B.M., 1982. Settlements above tunnels in the United Kindom – Their
Duncan, J.M., Chang, C.Y., 1963. Nonlinear analysis of stress and strain in soil. J. Soil
magnitudes and prediction. in: Jones, M.J. (Ed.), HYPERLINK “https://search.
Mech. Found. Div. 96 (5), 1629–1653.
worldcat.org/it/title/889517498” Tunnelling ’82: papers presented at the third
El-Nahhas, F.M., El-Mossallamy, Y.M., 2009. The role of small strain constitutive model
international symposium . Institution of Mining and Metallurgy, London, pp. 173-
for predicting differential settlement above tunnels, in: Proc. of the 13th Int. Conf. on
181.
Structural and Geotechnical Engineering, Cairo, 27-29 December 2009.
Panet, M., Guenot, A., 1983. Analysis of convergence behind the face of a tunnel. In:
El-Nahhas, F.M., El-Mossallamy, Y.M., El-Shamy, A.A., 2015. 3D analysis of ground
Tunnelling 82, proceedings of the 3rd international symposium, Brighton, 7–11 June
settlement induced by machanized tunnelling, in: Proc. of the 14th Int. Conf. on
1982, pp. 197–204.
Structural and Geotechnical Engineering, Cairo, 20-22 December 2015.
Peck, R.B., 1969. Deep excavations and tunneling in soft ground, in: Proceedings of the
Seventh International Conference on Soil Mechanics Foundation Engineering.
Sociedad Mexicana de Mecánica de Suelos, Mexico City, pp. 225-290.

22
F. Schiena et al. Tunnelling and Underground Space Technology incorporating Trenchless Technology Research 153 (2024) 106001

Pelizza, S., 2002. La scelta del metodo di scavo meccanizzato integrale in ambiente Schiena, F., 2020. Doctoral thesis: Predicting effects induced by shallow mechanized
urbano. in: XXI Convegno Nazionale di Geotecnica, L’Aquila: Opere geotecniche in tunnelling in coarse-grained soils. Department of Engineering, “Roma Tre”
ambiente urbano. Pàtron, pp. 50-66. University.
Rowe, P.W., 1962. The stress-dilatancy relation for static equilibrium of an assembly of Song, G., Marshall, A.M., 2020. Centrifuge modelling of tunnelling induced ground
particles in contact. Proceedings of the Royal Society of London – Series A, displacements: pressure and displacement control tunnels. Tunn. Undergr. Space
Mathematical and Physical Sciences 269(1339), 500-527. Technol. 103 (103461).
Rowe, R.K., Lo, K.Y., Kack, G.J., 1983. A method of estimating surface settlement above Van Jaarsveld, E.P., Plekkenpol, J.W., van de Graaf, C.A., 1999. Ground deformations
tunnels constructed in soft ground. Canadian Geotechnical Journal 20 (1), 11–22. due to the boring of the Second Heinenoord tunnel. In: Barends, F.B.J.,
Sanzeni, A., Giuriani, E., Colleselli, F., 2017. Il consolidamento delle fondazioni del Lindenberg, J., Luger, H.J., de Quelerij, L., Verruijt, A. (Eds.), Geotechnical
palazzo della loggia di Brescia. in: Moraci, N., Soccodato, C. (Eds.), La Geotecnica Engineering for Transportation Infrastructure: Theory and Practice, Planning and
nella conservazione e tutela del patrimonio costruito: XXVI Convegno Nazionale di Design. Balkema, Rotterdam, pp. 153–160.
Geotecnica - Roma 2017. AGI, pp. 745-752. Vermeer, P.A., Brinkgreve, R., 1993. PLAXIS Version 5 Manual. Balkema, Rotterdam.
Sanzeni, A., Zinelli, L., Colleselli, F., 2010. Estimated settlements during the Brescia Viggiani, G.M.B., Standing, J.R., 2001. The Treasury. In: Burland, J.B., Standing, J.R.,
Metrobus tunnel excavation. In: Benz, T., Nordal, S. (Eds.), Numerical Methods in Jardine, F.M. (Eds.), Building Response to Tunnelling: Case Studies from
Geotechnical Engineering. CRC Press, London, pp. 789–794. Construction of the Jubilee Line Extension. London, Thomas Telford, London,
Schanz, T., Vermeer, P.A., Bonnier, P.G., 1999. Formulation and verification of the pp. 401–432.
Hardening-Soil Model. In: Brinkgreve, R.B.J. (Ed.), Beyond 2000 in Computational Vorster, T.E.B., Klar, A., Soga, K., Mair, R.J., 2005. Estimating the effects of tunneling on
Geotechnics. Balkema, Rotterdam, pp. 281–290. existing pipelines. J. Geotech. Geoenviron. Eng. 131 (11), 1399–1410.
Xu, J., Franza, A., Marshall, A.M., Losacco, N., 2021. Role of Footing Embedment on
Tunnel-Foundation Interaction. J. Geotech. Geoenviron. Eng. 147 (9).

23