Professional Documents
Culture Documents
a r t i c l e i n f o abstract
Article history: In this paper a numerical approach is proposed for the construction of fragility curves for shallow metro
Received 4 February 2011 tunnels in alluvial deposits, when subjected to transversal seismic loading. The response of the tunnel is
Accepted 27 November 2011 calculated under quasi static conditions applying the induced seismic ground deformations which are
Available online 15 December 2011
calculated through 1D equivalent linear analysis for an increasing level of seismic intensity. The results
Keywords: of the present numerical analyses are compared with selected closed form solutions, highlighting the
Seismic vulnerability limitations of the latter, while indicative full dynamic analysis are performed in order to validate the
Shallow tunnels results of the quasi-static method. The proposed approach allows the evaluation of new fragility curves
Fragility curves considering the distinctive features of the tunnel geometries and strength characteristics, the input
Numerical analysis
motion and the soil properties as well as the associated uncertainties. The comparison between the new
1D ground response analysis
fragility curves and the existing empirical ones highlights the important role of the local soil conditions,
which is not adequately taken into account in the empirical curves.
& 2011 Elsevier Ltd. All rights reserved.
0267-7261/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2011.11.004
2 S.A. Argyroudis, K.D. Pitilakis / Soil Dynamics and Earthquake Engineering 35 (2012) 1–12
constructed, as a function of the level and the type of the seismic varying along the tunnel sides and compatible to the strains
excitation, considering the related uncertainties. developed during the ground shaking. On the other hand the
The problem of the tunnel’s seismic response has been closed form solutions usually assume elastic soil behavior com-
addressed through analytical closed-form solutions by several patible, at the best, to a single average value of soil deformation
authors [38,25,6,18,24], based on a number of assumptions (i.e. (i.e. at the tunnels mid-depth) estimated through empirical or
linear elastic soil and lining behavior, homogeneous ground semi-empirical expressions.
medium, full-slip or no-slip conditions between soil and lining, The level of tunnel damage is described by a damage index
the effect of structure sequence is not considered). The present expressing the exceedance of the lining strength capacity and the
numerical analyses are compared with these closed form solu- fragility curves are estimated based on the evolution of damage
tions, reaching to similar conclusions and results with other index with the increasing earthquake intensity, considering
studies [13,32]. For completeness, representative indicative full associated uncertainties. The proposed approach allows the
dynamic analysis are performed in order to validate the results of evaluation of new fragility curves considering the distinctive
the quasi-static approach. features of the tunnel geometries and strength characteristics,
the input motion characteristics and the soil properties. In
contrast, the available empirical fragility curves do not consider
2. Procedure for deriving fragility curves adequately soil characteristics and they rather describe an aver-
age behavior of the tunnel in seismic action.
2.1. Overview
2.2. Definition of damage states
The proposed approach is based on the quasi static analysis of
tunnels in the transversal direction, considering the soil-structure The damage states of already existing empirical fragility curves
interaction through 2D coupled numerical analyses. The general for tunnels are based on a qualitative damage description from
flowchart of the procedure is illustrated in Fig. 1 [4]. In general, past earthquakes. Although various damage indexes and related
the transversal seismic analysis of underground structures based parameters have been proposed for the fragility analysis of
on imposed seismic ground displacements is quite common in buildings and bridges, no such information is available for
practice (e.g. [12,27]). Being essentially a static analysis, this tunnels. Considering this serious lack of references, in the present
approach is cost effective compared to more elaborate full approach the damage index (DI) is defined as the ratio between
dynamic time–history analysis. Especially for the needs of a the actual (M) and capacity (MRd) bending moment of the tunnel
comprehensive parametric fragility study like the present one, cross section. A definition based on moments is compatible with
where different soil profiles, input motions and tunnel geometries the use of displacements, according to the equal displacement
are used, the aforementioned approach is adequate. approximation. In line with other approaches, it is assumed that
The effect of soil conditions and ground motion characteristics the tunnel’s behavior is approximated to that of an elastic beam
in the global soil and tunnel response is taken into account by subjected to deformations imposed by the oscillating surrounding
using different typical soil profiles and seismic input motions. The ground due to seismic waves propagating perpendicular to the
response of the free field soil profiles and the induced seismic tunnel axis [12]. The actual bending moment (M) is calculated as
ground deformations are calculated through an 1D numerical the combination of static and seismic loads. The capacity of the
analysis, for an increasing level of seismic intensity. The nonlinear tunnel is estimated based on material and geometry properties of
soil behavior is considered through the variation of shear mod- the beam considering the induced static and seismic axial forces
ulus and damping ratio with shear strain in the framework of a 1D (N) and bending moments (M).
EQL ground response analysis; the Mohr-Coulomb criterion is According to previous experience of damages in tunnels and
used to model the 2D soil-tunnel behavior. The results of the 1D applying engineering judgment, four different damage states are
free field soil response analyses are also employed to define the considered due to ground shaking. They refer to minor, moderate,
appropriate stiffness parameters in the finite element analyses, extensive and complete damage of the tunnel lining and they are
Fig. 1. General flowchart of the procedure for deriving numerical fragility curves for tunnels in alluvial deposits.
S.A. Argyroudis, K.D. Pitilakis / Soil Dynamics and Earthquake Engineering 35 (2012) 1–12 3
described in Table 1. Although the proposed limits of damage HAZUS [19] for buildings; bC is assigned equal to 0.3 according to
index are not yet fully documented, they give a realistic descrip- analyses for bored tunnels of BART system [30]. The last source of
tion for the expected damage. uncertainty, associated with seismic demand, is described by the
average standard deviation of the damage indices that have been
2.3. Estimation of fragility curve parameters calculated for the different input motions at each level of PGA.
Table 1
Definition of damages states for tunnel lining.
1
Table 2
Selected records applied to the bedrock of the soil profiles.
or lower to 0.3 g, the first four records were used, while for 3.4. Estimation of the imposed seismic ground displacements
amplitudes greater of 0.3 g the next six records were applied. This
distinction was made in order to scale the real records to The imposed quasi-static seismic ground displacements have
amplitudes as much as possible consistent with their frequency been computed using a 1D EQL approach with the code EERA [5],
characteristics. assuming an equivalent linear elastic soil behavior. The variations
of shear modulus G/Go and damping ratio D with the shear strain
level g were defined according to the available data in the
3.3. Soil profiles literature as a function of plasticity index and effective stress [8].
Curves with PI¼ 30% for clay and PI ¼0% for sand materials were
Fourteen ideal soil deposits were considered, corresponding to selected in this study (Fig. 4). For the seismic bedrock, the curves
soil types B, C and D of Eurocode 8 [10], ranged according to the proposed by Schnabel et al. [31] were applied.
shear wave velocity (Vs30) values (Fig. 3). Three different thick- Each soil profile is discretised by appropriate number of layers
nesses were assumed, equal to 30 m (profiles: B30sand, B30clay, varying from 2.5 to 10 m thick. In the iterative procedure, the
C30sand, C30clay, D30sand, D30clay), 60 m (profiles: B60a, C60a, ratio of effective and maximum shear strain is assumed equal to
D60a, B60, C60, D60) and 120 m (profiles: B120, C120). Typical 0.65. The cumulative displacements in each soil layer and depth
values of the different soil properties were selected for each soil were estimated for each soil profile and input motion, based on
layer. The 60 m and 120 m profiles constitute of a surface 5 m the computed peak shear strain versus depth. These peak dis-
thick sand layer overlaying clay layers, while the 30 m profiles placements’ profiles are imposed on the lateral boundaries of the
consist solely of sand or clay. plain strain soil model in order to estimate the response of the
Vs (m/s) Vs (m/s)
0 200 400 600 800 1000 0 200 400 600 800 1000
0 0
Β60a B60
C60a C60
10 D60a 10 D60
20 20
Depth (m)
Depth (m)
30 30
40 40
50 50
60 60
Vs (m/s) Vs (m/s)
0 200 400 600 800 1000 0 200 400 600 800 1000
0 0
B30sand/clay B120
C30sand/clay C120
5 20
D30clay/sand
10 40
Depth (m)
Depth (m)
15 60
20 80
25 100
30 120
1.0 25.0
PI = 30%, σο' = 0.25atm
PI = 30%, σο' = 1.0atm
0.8 20.0
PI = 30%, σο' = 4.0atm
PI = 0%, σο' = 0.25atm
PI = 30%, σο' = 0.25atm
0.6 15.0 PI = 0%, σο' = 1.0atm
PI = 30%, σο' = 1.0atm
G/Gmax
D (%)
PI = 30%, σο' = 4.0atm
Rock
0.4 PI = 0%, σο' = 0.25atm 10.0
PI = 0%, σο' = 1.0atm
PI = 0%, σο' = 4.0atm
0.2 5.0
Rock
0.0 0.0
0.000 0.001 0.010 0.100 1.000 0.000 0.001 0.010 0.100 1.000
Shear strain γ (%) Shear strain γ (%)
Fig. 4. Shear modulus reduction (G/Gmax) and variation of damping ratio (D) with shear strain (g).
Maximum Acceleration amax (g) G/Gmax Peak shear strain γmax (%) Cummulative displacement Dn (m)
0.0 0.2 0.4 0.6 0.8 0.0 0.5 1.0 0.0 0.5 1.0 1.5 2.0 0.00 0.05 0.10 0.15
0 0 0 0
10 10 10 10
20 20 20 20
Depth (m)
30 30 30 30
40 40 40 40
50 50 50 50
60 60 60 60
Fig. 5. Example of 1D ground response analysis results with EERA and estimation of displacements versus depth (Soil profile: C60, Input motion: Montenegro, 0.5 g for
outcrop conditions).
tunnel lining under quasi-static conditions. We avoided to apply soil profile is selected as the representative parameter of the
the computed displacements’ pattern directly on the tunnel lining seismic intensity in the fragility curves.
through springs and dashpots because recent studies [27] have The relative stiffness of the structure and the surrounding soil is
proved that there is no accurate enough way to choose these an important parameter describing the soil–structure interaction
parameters, and consequently the associated scatter may be very identity of underground structures [12]. It is quantified by the
important, affecting the final results. Moreover, it was found that flexibility ratio (F), which is a measure of the flexural stiffness
the displacements that are estimated from the displacement time (resistance to ovaling for circular tunnel or racking for rectangular
histories in each layer for the time instance where the shear strain frame) of the medium relative to the lining. The flexibility ratio is the
takes the maximum value at the tunnel depth, are similar or even most important factor that contributes to the soil–structure interac-
lower than the aforementioned ones based on the peak values. tion. It depends on the characteristics of the medium (modulus of
Therefore, the displacements’ pattern used herein, which are elasticity Em or shear modulus Gm) and the geometrical and material
based on the peak shear strains, constitute a more conservative properties of the tunnel. The flexibility ratios that correspond to the
and easier to apply approach. different soil profiles and the two tunnels of the present study are
The computed variation of G versus depth was also used to given in Fig. 6. The modulus of elasticity of the surrounding soil has
evaluate the corresponding modulus of elasticity (E) of each soil been estimated through the 1D ground response analyses for the
layer, which is used in the quasi static analysis of tunnel. In different input motions, as described above. They vary from 0.2 to 55
particular, an average value of E is calculated for each soil layer covering a wide range of soil-structure interaction cases.
based on the average values computed for the different input
motions. Fig. 5 shows a typical example of the computed ground 3.5. Numerical analyses
response in terms of maximum acceleration amax, normalized
shear stiffness G/Gmax, peak shear strain gmax, and cumulative The response of the tunnel is calculated under quasi-static
displacement Dn. The computed PGA value at the surface of each conditions applying the induced free field cumulative displacements,
6 S.A. Argyroudis, K.D. Pitilakis / Soil Dynamics and Earthquake Engineering 35 (2012) 1–12
which were calculated through the 1D EQL analysis. A plane strain shear stiffness moduli, are used for the quasi-static analysis. Fig. 7
ground model with the tunnel cross section is simulated using the shows a representative example of the tunnel response after
finite element code PLAXIS 2D [28]. The lateral extent of the model is imposing the shear ground displacements.
properly selected in order to simulate the soil structure interaction
and to avoid boundary effects. Based on sensitivity analysis it was
found that a distance equal to three times the diameter of the 4. Comparison of the numerical results with other solutions
circular section (i.e. 30 m) from both sides of the tunnel axis is
adequate for the analysis. 4.1. Free field shear strains
The side boundary conditions were fixed in the vertical
direction and free to move in the horizontal direction, while the The computed maximum free field shear strains for the 60 m
nodes at the bottom of the mesh were fixed in both directions. soil profiles (B, C and D soil categories) and the different input
Prior to the application of the imposed displacement, a set of motions, are compared with the simplified analytical solution
initial static analyses was performed to properly model the initial proposed by St. John and Zahrah [36], which corresponds to the
static conditions, the excavation of the tunnel and the construc- S-waves propagation (Fig. 8). The latter one is expressed as the
tion of the lining. The circular tunnel excavation was simulated ratio of the peak ground velocity (PGV) to the velocity of S-wave
through the volumetric contraction of the tunnel section corre- propagation (Vs). The PGV values are estimated according to the
sponding to volume loss equal to 0.2–0.5% for soil type B and 1% suggestion by Hashash et al. [12], based on the ratios of peak
and 2% for the soil types C and D, respectively. ground velocity to peak ground acceleration [29] and the soil
The behavior of the tunnel lining was assumed to be linear
elastic, while the soil was characterized by a Mohr–Coulomb yield Maximum shear strain γmax (-15m)
criterion for all the stages of the analysis. Strain compatible soil 1E-01
60
Numerical analysis (EERA)
<=0.3g, Circular section
50 1E-02
>0.3g, Circular section
Flexibility Ratio F
B30clay
Β60a
B60
B120
C30sand
C30clay
C60a
C60
C120
D30sand
D30clay
D60a
D60
Soil Profile Fig. 8. Comparison between 1D numerical (EERA) and analytical (St. [36]) values
of the maximum shear strains (gmax) at the circular tunnel depth ( 15 m) for the
Fig. 6. Variation of flexibility ratios F for the different soil profiles and the two 60 m soil profiles (B60, B60a, C60, C60a, D60, D60a correspond to the Vs profiles
tunnel sections. described in fig.3 according to EC8).
Axial forces
N max = -692.7 kN/m
Extrem e axial force -1.06*103 kN/m
Deformed Mesh
Extreme total displacement 43.07*10-3 m N max = -1060 kN/m
(displacements scaled up 100.00 times)
Fig. 7. Example of 2D analysis results: deformed mesh (a), total moment and axial forces of the circular (b), and rectangular (c), tunnel lining (Soil profile: B60a, Input
motion: Kypseli, 0.3 g).
S.A. Argyroudis, K.D. Pitilakis / Soil Dynamics and Earthquake Engineering 35 (2012) 1–12 7
type; B and C are considered as stiff, while D as soft soil. For the deflection, is defined as a function of the flexibility ratio F based
moment magnitude it is assumed that PGA values up to 0.3g on analytical solutions provided by Wang [38] for full slip
correspond to Mw ¼6.5, while the higher values correspond to conditions and by Penzien [25] and Park et al. [24] for no slip
Mw ¼7.5. Finally, the source to site distance is taken between 20 conditions between the circular tunnel and surrounding soil. In
and 50 km. The Vs value is referred to the soil layer at the tunnel Fig. 10 the normalized structure deflections that have been
depth (Fig. 3). The different points in Fig. 8 correspond to the computed based on the quasi-static numerical approach are
various seismic excitations. While we observe a large scatter, in compared with the aforementioned analytical solutions for the
general the agreement is rather good for soil type B, but rather circular tunnel. The same solutions are used for the case of
poor for soil type C and soil type D, where the computed strains rectangular cross section considering the approximation of 10%
are clearly larger. This is certainly due to the non-linear effects of reduction in the normalized distortion (i.e. ratio of racking and
soil response. Consequently, it is concluded that in many cases, free field deflection) as it is described by Wang [38] and Hashash
and in particular for strong events, the simplified analytical et al. [12]. In the same diagrams, an average logarithmic regres-
expressions of the seismic ground strains are not conservative sion curve is fitted to the data set, bounded from the correspond-
enough to be applied in engineering applications. ing standard deviation. We observe that the analytical solutions
The limitations of such expressions have been also discussed overestimate the tunnel response, especially in the case circular
by Paolucci and Pitilakis [23]. They concluded that the transient sections. However in the case of the rectangular section the
ground strains are not only affected by wave passage effects, as regression curve fits rather satisfactorily with both analytical
generally assumed in these formulas, but also by site effects, solutions (slightly better that for the Park et al. case, which was
especially in terms of lateral variations of soil properties and developed for no slip conditions and corresponds better to our
spatial incoherency of ground motion. The comparison of the model). The differences between the two approaches (numerical
present numerical results with a recently proposed relationship and analytical) are attributed to the different assumptions of the
by Paolucci and Pitilakis [23] is quite good considering the fact interface conditions previously described, i.e. full slip conditions
that this relationship is referring to the ground surface. and linear elastic soil behavior in analytical solution, no slip
conditions and elasto-plactic soil behavior, together with the
4.2. Moment and axial force in circular tunnel better simulation of excavation in the numerical approach.
According to the mean regression curves of the numerical results,
The values of gmax and modulus of elasticity (Em) obtained at the structure deformation is expected to be higher than the free
the circular tunnel axis depth (i.e. 15 m) with the 1D ground field one, when F435 for circular and for F 41.5 for rectangular
response analyses for each soil profile and input motion, are used tunnel section.
to estimate the bending moment and axial force in the tunnel
lining applying the analytical expressions provided by Wang [38]. 4.4. Comparison with full dynamic analysis for circular tunnel
Then the results are compared with the 2D numerical modeling
using PLAXIS. In both cases no-slip conditions between soil and Representative dynamic analyses were performed with the
lining are assumed. The comparison between the analytical finite element code PLAXIS 2D [28] in order to compare and
solution and the numerical results for the three soil types is validate the results of the quasi-static approach for the circular
shown in Fig. 9. The analytical solutions are generally over- tunnel section. The analyses are performed for the 60 m soil
estimate both moments and axial forces especially at higher profiles (B60a, C60a) considering the acceleration time histories
strain levels. Similar conclusions have been reached by other of Kypseli and Gebze (scaled at 0.3 g). The mesh of the model is
researchers when they compared closed form solutions with characterized by a width equal to eight times its height in order to
numerical analysis results following the same assumptions with minimize the boundary effects [2]. The discretization of the model
the analytical approach [13,32,2]. The differences between the includes a total number of 1786 15-node plane strain triangular
two solutions are larger when elasto-plactic models are applied elements and is refined around the tunnel section.
for the soil in the numerical analyses [22,16,2] as it is the case of The initial static and the following dynamic analysis of the
the present study. In particular, the redistribution of stresses present time–history and quasi-static modeling have been per-
around the tunnel due to the non linear soil behavior and the formed assuming linear elastic range. The seismic induced bend-
plastic deformations that are developed, contribute to the ing moments and axial forces of the tunnel section are compared
reduced structure deflections and consequently to reduced values in Table 3 for the two approaches. In case of the dynamic
of generated loads in the tunnel lining, in contrast to the elastic analyses, the maximum and effective (65% of the maximum)
approach. Additionally, when the elasto-plastic approach is values of the loads are given. In general, the dynamic analysis
adopted, the computed thrusts are generally compressive for results, as expected, in higher values; however an overall satis-
the full range of strains, resulting in a stiffer lining and tending factory agreement is observed when the effective and more
to prevent the tensile strains that would occur using a linear realistic from engineering point of view, values of the dynamic
elastic model. The simulation of the excavation in the numerical analysis are compared to the quasi-static approach.
approach, considering the volumetric contraction of the tunnel
section and the corresponding volume loss that modifies the
initial stress conditions around the tunnel, also contributes to 5. Fragility curves
the different results between the two solutions. Finally, another
possible source of differences is the fact that in our analyses the The derivation of fragility curves (i.e. the definition of the
imposed soil deformations vary with depth, while in the analy- median threshold value of PGA for each damage state) is based on
tical solutions only an average value of strain at the tunnel axis the construction of diagram of the computed damage indices
depth is considered. versus PGA at the ground surface according to the definitions of
Table 1. The diagram is estimated by linear regression analysis,
4.3. Ovaling and racking deformations considering the natural logarithm of the damage index (LnDI) as
the dependent variable and PGA as the independent variable.
The normalized structure deflection i.e. the ratio of lining Similar approaches are used for the derivation of fragility analysis
diametric deflection of circular tunnel and free field ground for bridges (e.g. [15,7,20,26]). Examples of the evolution of
8 S.A. Argyroudis, K.D. Pitilakis / Soil Dynamics and Earthquake Engineering 35 (2012) 1–12
1 1
1 100 10000 1 100 10000
Analytical solution (Wang,1993) Analytical solution (Wang,1993)
Moment [kNm] Axial force [kN]
10000 10000
C30clay C30clay
C30sand C30sand
C60 C60
Numerical analysis (Plaxis)
C60a C60a
C120 C120
1:1 1:1
100 100
1 1
1 100 10000 1 100 10000
Analytical solution (Wang,1993) Analytical solution (Wang,1993)
Moment [kNm] Axial force [kN]
10000 10000
D30clay D30clay
D30sand D30sand
D60 D60
Numerical analysis (Plaxis)
D60a D60a
1:1 1:1
100 100
1 1
1 100 10000 1 100 10000
Analytical solution (Wang,1993) Analytical solution (Wang,1993)
Fig. 9. Comparison between numerical (PLAXIS) and analytical [38] solution of maximum moment and axial forces of circular tunnel lining for the different soil profiles
(type B, C, D according to EC8).
damage with PGA are given in Fig. 11 for the two tunnel sections, The sets of fragility curves derived for each soil type (B, C and D)
where an average linear regression is fitted to the data set, following the procedure that is described in this paper, are given in
bounded by the corresponding standard deviation. Figs. 12 and 13 for the circular and rectangular tunnel section
S.A. Argyroudis, K.D. Pitilakis / Soil Dynamics and Earthquake Engineering 35 (2012) 1–12 9
2.5 2.5
Structure deformation
Structure deformation
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0 10 20 30 40 50 60 0 5 10 15 20 25
Flexibility ratio, F Flexibility ratio, F
Wang, 1993 (full slip) Park et al, 2009 (no slip) Wang, 1993 (full slip) Park et al, 2009 (no slip)
Penzien, 2000 (no slip) Numerical analysis Penzien, 2000 (no slip) Numerical analysis
+1σ -1σ +σ -1σ
Average Average
Fig. 10. Comparison between numerical (all soil profiles – PLAXIS) and analytical [38,25,24] solutions of normalized structure deflection for circular and rectangular
tunnels.
Table 3 2.0
Circular tunnel - Soil type: C
Comparison between full dynamic and quasi-static approach.
1.5 +σ
Soil profile and Dynamic analysis Quasi-static Extensive damage
input motion Maximum (Effective) analysis
Damage Index (lnDI)
1.0
Moderate damage -σ
B60a Gebze (0.3 g) 9DM9max 94.3 (61.3) 40.6
(kNm/m) 0.5
Minor damage
9DN9max 381.1 (247.7) 215.8
(kN/m) 0.0
C60a Gebze (0.3 g) 9DM9max 810.3 (526.7) 535.6
(kNm/m) -0.5
9DN9max 828.1 (538.3) 719.9
(kN/m)
C60a Kypseli (0.3 g) 9DM9max 397.3 (258.2) 184.3 -1.0
(kNm/m)
9DN9max 409.8 (266.4) 251.4 -1.5
(kN/m) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
PGA surface (g)
2.0
respectively. In the same figures the parameters of the lognormal Rectangular tunnel - Soil type: B
distribution in terms of median and standard deviation are
1.5
presented. Comparing the fragility curves derived for the three
Extensive damage
soil types, it is noted that for the same PGA in both tunnel types the
Damage Index (lnDI)
1.0 +σ
vulnerability is gradually increasing from soil type D to C and from Moderate damage
soil C to D. The fragility curves for extensive damages for soil type B 0.5 -σ
Minor damage
in both tunnel sections, and for soil type D in case of rectangular
tunnel, (dashed lines), are derived based on extrapolation of the 0.0
available computation results. The extrapolated values of damage
index for the extensive damages could be attributed to several -0.5
reasons, i.e. higher stiffness in case of soil class B and predomi-
nance of non-linear soil behavior and internal damping expected in -1.0
the case of soil class D.
A practical consequence of these observations is that the -1.5
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
probability of extensive damages is very low for circular tunnels
in stiff soils, even for very important ground accelerations. The PGA surface (g)
same is expected for rectangular tunnels in soft soils. These Fig. 11. Examples of the evolution of damages with PGA at the ground surface for
remarks are consistent with the few available observations from circular and rectangular tunnel cross-sections. Estimation of the median threshold
recent strong earthquakes as it is further discussed in the next values of PGA for each damage state.
section.
Comparing the fragility curves derived for the two tunnel
types, it is observed that the rectangular shallow tunnel, con- deviation values are estimated in the case of the circular section
structed with cut and cover, is more vulnerable than the circular, in soil type C and D, which reflects the variability in the results in
normally bored, tunnels, for all soil types. The higher standard terms of PGA at the surface and damages indices values.
10 S.A. Argyroudis, K.D. Pitilakis / Soil Dynamics and Earthquake Engineering 35 (2012) 1–12
1.00 1.00
Soil type: Β Soil type: Β
Μoderate damage
Probability of damage
Μoderate damage
.
Median: 1.51g, βtot = 0.55 Median: 1.28g, βtot = 0.53
Extensive damage Extensive damage
Median: 1.74g, βtot = 0.55 Median: 1.73g, βtot = 0.53
0.50 0.50
0.25 0.25
0.00 0.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
PGA (g) PGA (g)
1.00 1.00
Soil type: C Soil type: C
0.75 0.75
Probability of damage
Probability of damage
0.50 0.50
1.00 1.00
Soil type: D Soil type: D
0.75
Probability of damage
0.75
Probability of damage
0.50
0.50
6. Comparison between numerical and empirical damage resulting fragility curves with the empirical fragility curves for
data alluvial (soil) and cut and cover tunnels with good quality
construction. For this type of tunnels no empirical curves are
The numerically derived fragility curves are compared with provided for extensive damages, which is consistent with the
the empirical ones that are proposed by ALA [1] based on numerical results. In particular, the threshold PGA values for
observed damage data in tunnels from past earthquakes extensive damages are very high or they are estimated based on
(Fig. 14). In these empirical curves PGA values have been extrapolation of the numerical results since the computed
estimated using available ground motion prediction equations, damage indexes are low, as it is described in the previous section.
with all the uncertainties associated to these models. Moreover, The role of the soil conditions is very important to be
the database includes tunnels of various functions (i.e. highway, neglected. The numerical fragility curves are seriously modified
transit, railroad, water supply and communications). They are with the soil conditions, contrary to the empirical curves, which
classified as tunnels in rock and tunnels in soil, and for poor-to- are rather describing an average response of the tunnels inde-
average and good construction practices and maintenance condi- pendently of soil conditions. Therefore, the empirical curves may
tions. Finally, the definition of the damage states is strictly over or underestimate the probability of damage. It is also
qualitative, based mainly on the extent of the observed cracking interesting to remark that the empirical curves are well compared
of the tunnel liner, independently of the type of damage produ- with the numerical ones for soil types C or D. This is due to the
cing these cracking; for example cracking due to transversal fact that the majority of damages have been reported in ‘‘mod-
stressing is not differentiated from cracking in the longitudinal erate’’ and ‘‘poor’’ soil conditions that correspond to soil type
axis. In the present comparison, we selected to compare our C and D.
S.A. Argyroudis, K.D. Pitilakis / Soil Dynamics and Earthquake Engineering 35 (2012) 1–12 11
Circular (bored) tunnel - Minor damage Circular (bored) tunnel - Moderate damage
1.00 1.00
0.75 0.75
Probability of damage
Probability of damage
0.50 0.50
0.25 0.25
0.00 0.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
PGA (g) PGA (g)
Rectangular (cut & cover) tunnel - Minor damage Rectangular (cut & cover) tunnel - Moderate damage
1.00 1.00
0.75 0.75
Probability of damage
0.25 0.25
0.00 0.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
PGA (g) PGA (g)
Soil type Β Soil type C Soil type D ALA - Good quality construction
Fig. 14. Comparison between numerical and empirical [1] fragility curves for circular (up) and rectangular (down) tunnel cross section.
Table 4
Comparison between observations and estimations from fragility curves for cut and cover tunnels.
Earthquake Tunnel type and PGA on site Soil type Observed Probability of
location (free field) damage damage
Parnitha, 1999 Monastiraki 0.34 g Stiff sandy clays and highly weathered rock formations None None: 0.93
station down to at least 60 m depth, Vs30 ¼ 400 m/s Min.: 0.06
Mod.: 0.01
Ext.: 0.00
Parnitha, 1999 Sepolia station 0.31 g 0–7 m: sandy to silty clay, 7–15 m: stiff sandy clay with None None: 0.95
gravels, 15–25 m: fractured conglomerate, ‘Athenian Schist’, Min.: 0.04
Vs30 ¼ 390 m/s Mod.: 0.00
Ext.: 0.00
Kobe, 1995 Dakai station 0.70 g alternations of mostly saturated loose sandy and soft clayey, Extensive None: 0.14
Vs30 ¼ 300 m/s Min.: 0.42
Mod.: 0.23
Ext.: 0.22
The proposed fragility curves are applied to estimate the while Dakai station was not designed according to modern codes.
damage probabilities of cut and cover tunnels that were subjected Moreover, the curves were developed for one-barrel rectangular
to strong ground motion in recent earthquakes (Table 4). Two section, without the presence of the vulnerable central columns
rectangular stations in Athens (Monastiraki, Sepolia) and one in as it is the case of Dakai station.
Kobe (Dakai) are examined. The details about the structure
geometry and response, site soil conditions and recorded ground
motions are given by Gazetas et al. [11]. The observations are well 7. Conclusions
predicted by the fragility curves in case of Athens metro. The
prediction is less satisfactory in case of Dakai station. This A simple yet comprehensive numerical methodology is pro-
divergence may be attributed to the fact that the present fragility posed to construct fragility curves for shallow metro tunnels in
curves are referring to modern structures designed and con- alluvial deposits, when subjected to transversal seismic loading.
structed according to recent improved seismic code prescriptions, Tunnel response is calculated under quasi static conditions,
12 S.A. Argyroudis, K.D. Pitilakis / Soil Dynamics and Earthquake Engineering 35 (2012) 1–12
applying at a certain distance from the tunnel the induced seismic [11] Gazetas G, Gerolymos N, Anastasopoulos I. Response of three Athens metro
ground deformations, which are calculated through a 1D equiva- underground structures in the 1999 Parnitha earthquake. Soil Dynamics and
Earthquake Engineering 2005;25:617–33.
lent linear analysis. Seismic inputs with different frequency [12] Hashash YA, Hook J, Schmidt B, Chiangyao J. Seismic design and analysis of
content are used, scaled in different levels of seismic loading. underground structures. Tunnelling and Underground Space Technology
Typical tunnel cross sections (circular and rectangular) and soil 2001;16:247–93.
[13] Hashash YA, Park D, Chiangyao J. Ovaling deformations of circular tunnels
profiles, classified as B, C and D according to EC8, are employed. under seismic loading, an update on seismic design and analysis of under-
Defining different damage states, the fragility curves could be ground structures. Tunnelling and Underground Space Technology 2005;20:
derived as a function of the level of peak ground acceleration at 435–41.
[14] Kappos A, Panagopoulos G, Panagiotopoulos Ch, Penelis G. A hybrid method
the ground surface, considering the related uncertainties in the for the vulnerability assessment of R/C and URM buildings. Bulletin of
definition of damage states, the demand and the capacity of the Earthquake Engineering 2006;4:391–413.
tunnel. [15] Karim KR, Yamazaki F. Effect of earthquake ground motions on fragility
curves of highway bridge piers based on numerical simulation. Earthquake
Prior to the construction of the fragility curves the results of
Engineering and Structural Dynamics 2001;30:1839–56.
our numerical analyses are compared and validated with avail- [16] Kramer, G, Sedarat, H, Kozak, A, Liu, A, Chai, J. Seismic response of precast
able analytical expressions as well as dynamic analyses. The tunnel linings, in: Proceedings of Rapid Excavation and Tunneling Confer-
comparisons proved the reliability of the proposed numerical ence, Toronto, 2007.
[17] Moschonas I, Kappos A, Panetsos P, Papadopoulos V, Makarios T, Thanopou-
approach, while at the same time they highlighted the inherent los P. Seismic fragility curves for Greek bridges: methodology and case
limitations of the closed-form solutions. The observed general studies. Bulletin of Earthquake Engineering 2009;7:439–68.
trend of the analytical solutions to overestimate the tunnel [18] Nagar H, Hinchberger SD, Nagar MH. Simplified analysis of seismic in-plane
stresses in composite and jointed tunnel linings. Soil Dynamics and Earth-
response, especially at higher strain levels, is mainly due to the quake Engineering 2008;28:1063–77.
inelastic soil behavior and the simulation of volume loss during [19] NIBS. HAZUS-MH: Technical Manuals. Washington, DC: Federal Emergency
tunnel excavation in the numerical approach, which could not be Management Agency and National Institute of Building Science; 2004.
[20] Nielson BG, DesRoches R. Seismic fragility methodology for highway bridges
properly accounted for in the close form analytical solutions. using a component level approach. Earthquake Engineering and Structural
So far the vulnerability assessment of tunnels is based on Dymamics 2007;36:823–39.
simple empirical fragility curves, without properly considering [21] Owen, GN, Scholl, RE. Earthquake engineering of large underground struc-
tures, in report No. FHWA/RD-80/195, Federal Highway Administration and
soil and tunnel characteristics. With the proposed numerically National Science Foundation, (1981) 279p.
derived fragility curves we consider in a more systematic way the [22] Pakbaz M, Yareevandb A. 2-D Analysis of circular tunnel against earthquake
distinctive features of the tunnel geometric and strength proper- loading. Tunnelling and Underground Space Technology 2005;20:411–7.
[23] Paolucci, R, Pitilakis, K. Seismic risk assessment of underground structures
ties, as well the input motion characteristic and the soil proper-
under transient ground deformations, in: K. Pitilakis (Ed.), Earthquake
ties. The proposed fragility curves should be used for shallow Geotechnical Engineering: 4th International Conference on Earthquake
tunnels, circular or rectangular, in shallow alluvial deposits Geotechnical Engineering – Invited Lectures, Springer, Netherlands, 2007,
classified according to EC8 soil categories. pp. 433–459.
[24] Park KH, Tantayopin K, Tontavanich B, Owatsiriwong A. Analytical solution
for seismic-induced ovaling of circular tunnel lining under no-slip interface
conditions: a revisit. Tunnelling and Underground Space Technology
2009;24:231–5.
Acknowledgments [25] Penzien J. Seismically induced raking of tunnel linings. Earthquake Engineer-
ing and Structural Dynamics 2000;29:683–91.
This work is partially developed in the framework of the [26] Pinto, PE (Ed.), Probabilistic methods for seismic assessment of existing
structures, LESSLOSS Report No. 2007/06, Istituto Universitario di Studi
research project SYNER-G: systemic seismic vulnerability and risk Superiori di Pavia, IUSS Press, 2007.
analysis for buildings, lifeline networks and infrastructures safety [27] Pitilakis, K, Tsinidis, G. Seismic design of large, long underground structures:
gain, funded by the European Commission (FP7-ENV-2009-1- metro and parking stations, highway tunnels, in: Proceedings of International
Geotechnical Conference: Geotechnical Challenges in Megacities (GEO-
244061). MOS2010), Moscow, 2010.
[28] Plaxis 2D, Reference Manual, version 7, 1998.
[29] Power, M, , Rosidi, D, Kaneshiro, J, Gilstrap, S, Chiou, SJ. (1998). ‘‘Summary
References and Evaluation of Procedures for the Seismic Design of Tunnels’’, Final Report
for Task 112-D-5.3(c), MCEER Highway Project, Sponsored by US Department
of Transportation, Federal Highway Administration, FWHA Contract Number
[1] American Lifelines Alliance (ALA), Seismic Fragility Formulations for Water DTFH61-92-C-00112.
Systems, part 1– Guideline, ASCE-FEMA, Reston, VA2001, 104 p. [30] Salmon, M, Wang, J, Jones, D, Wu, Ch. Fragility formulations for the BART
[2] Amorosi A, Boldini D. Numerical modelling of the transverse dynamic system, in: Proceedings of the 6th U.S. Conference on Lifeline Earthquake
behaviour of circular tunnels in clayey soils. Soil Dynamics and Earthquake Engineering, TCLEE, Long Beach, 2003.
Engineering 2009;29:1059–72. [31] Schnabel, PB, Lysmer, J, Seed, HB. SHAKE: a computer program for earthquake
[3] ATC-13, Earthquake Damage Evaluation Data for California, Applied Technol- response analysis of horizontally layered sites, Report No. UCB/EERC-72/12,
ogy Council, Redwood City, 1985. Earthquake Engineering Research Center, University of California, Berkeley,
[4] Argyroudis, S. Contribution to seismic vulnerability and risk of transportation December, (1972)102p.
networks in urban environment, PhD thesis, Department of Civil Engineering, [32] Sedarat H, Kozak A, Hashash YMA, Shamsabadi A, Krimotat A. Contact
Aristotle University of Thessaloniki, Greece, 2010. interface in seismic analysis of circular tunnels. Tunnelling and Underground
[5] Bardet JP, Ichii K, Lin CH. EERA: A Computer Program for Equivalent-Linear Space Technology 2009;24:482–90.
Earthquake Site Response Analyses of Layered Soil Deposits, University of [33] Sharma S, Judd WR. Underground opening damage from earthquakes.
Southern California. Department of Civil Engineering; 2000. 40 p. Engineering Geology 1991;30:263–76.
[6] Bobet A. Effect of pore water pressure on tunnel support during static and [34] Shinozuka, M. The Hanshin-Awaji earthquake of January 17, 1995: perfor-
seismic loading. Tunnelling and Underground Space Technology 2003;18: mance of lifelines, Technical Report NCEER-95-0015, 1995.
377–93. [35] Shinozuka M, Feng MQ, Kim H-K, Kim S-H. Nonlinear static procedure for
[7] Choi E, DesRoches R, Nielson B. Seismic fragility of typical bridges in fragility curve development. Journal of Engineering Mechanics 2000;126:
moderate seismic zones. Engineering Structures 2004;26:187–99. 1287–96.
[8] Darendeli, MB. Development of new family of normalized modulus reduction [36] John St. CM, Zahrah TF. Aseismic design of underground structures. Tunnel-
and material damping curves, PhD thesis, University of Texas, Austin, 2001. ing and Underground Space Technology 1987;2:165–97.
[9] Dowding CH, Rozen A. Damage to rock tunnels from earthquake shaking. [37] Wang JM. The distribution of earthquake damage to underground facilities
Journal of the Geotechnical Engineering Division 1978;104:175–91. during the 1976 Tang-Shan earthquake. Earthquake Spectra 1985;1:741–57.
[10] EC8. Eurocode 8: Design of Structures for Earthquake Resistance. Brussels, [38] Wang JN. Seismic Design of Tunnels. A Simple State-of-the-Art Design
Belgium: European Committee for Standardisation; 2004. The European Approach, Monograph 7. Parsons, Brinckerhoff: Quade & Diuglas Inc.,
Standard EN 1998–1. New York; 1993.