You are on page 1of 17

EARTHQUAKE ENGINEERING & STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2015; 44:1863–1879


Published online 23 February 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eqe.2563

Analytical seismic fragility functions for highway and railway


embankments and cuts

Sotiris Argyroudis1,*,† and Amir M. Kaynia2,3


1
Department of Civil Engineering, Aristotle University, Thessaloniki, Greece
2
Department of Structural Mechanics, Norwegian University of Science and Technology, NTNU, Trondheim, Norway
3
Norwegian Geotechnical Institute, Oslo, Norway

SUMMARY
A fundamental tool in seismic risk assessment of transportation systems is the fragility curve, which de-
scribes the probability that a structure will reach or exceed a certain damage state for a given ground motion
intensity. Fragility curves are usually represented by two-parameter (median and log-standard deviation) cu-
mulative lognormal distributions.
In this paper, a numerical approach, in the spirit of the IDA, is applied for the development of fragility
curves for highways and railways on embankments and in cuts due to seismic shaking. The response of
the geo-construction to increasing levels of seismic intensity is evaluated using a 2D nonlinear finite element
model, with an elasto-plastic criterion to simulate the soil behavior. A calibration procedure is followed in
order to account for the dependency of both the stiffness and the damping to the soil strain level. The effect
of soil conditions and ground motion characteristics on the response of the embankment and cut is taken into
account considering different typical soil profiles and seismic input motions. This study will provide input
for the assessment of the vulnerability of the road/railway network regarding the performance of the embank-
ments and cuts; therefore, the level of damage is described in terms of the permanent ground displacement in
these structures.
The fragility curves are estimated based on the evolution of damage with increasing earthquake intensity,
which is described by PGA. The proposed approach allows the evaluation of new fragility curves considering
the distinctive features of the element’s geometry, the input motion, and the soil properties as well as the as-
sociated uncertainties. A relationship between the computed permanent ground displacement on the surface
of the embankment and the PGA in the free field is also suggested based on the results of the numerical anal-
yses. Finally, the proposed fragility curves are compared with existing empirical data and the limitations of
their applicability are outlined. Copyright © 2015 John Wiley & Sons, Ltd.

Received 23 April 2014; Revised 13 October 2014; Accepted 16 January 2015

KEY WORDS: fragility curves; embankments; cuts; site effects; numerical modeling

1. INTRODUCTION

Past earthquakes indicate that highway and railway components such as bridges, tunnels,
embankments, cuts, slopes, and retaining walls are quite vulnerable to earthquake shaking. In
addition to life and material losses, damage to highway and railway elements can seriously affect
the transportation of people and products in both short-term (emergency and relief operations)
and long-term. Recent devastating earthquakes have revealed that damage to geotechnical
structures of highway and railway networks is often as visible and important as the structural

*Correspondence to: S. Argyroudis, Department of Civil Engineering, Aristotle University of Thessaloniki, Thessaloniki,
Greece.

E-mail: sarg@civil.auth.gr

Copyright © 2015 John Wiley & Sons, Ltd.


1864 S. ARGYROUDIS AND A. M. KAYNIA

damage. Therefore, the vulnerability of earth structures is of great interest in the seismic risk
assessment of transportation networks.
Fragility curves constitute one of the key elements of seismic probabilistic risk assessment of the
built-up environment and lifeline systems. They relate the seismic intensity to the probability of
reaching or exceeding a damage state (e.g. minor, moderate, extensive, and collapse) for each
element at risk. Different approaches can be used to develop the fragility curves, including
empirical, judgment, analytical, and hybrid methods. Recently, fragility curve methodologies using
numerical approaches have become more widely adopted as they are readily applicable to different
structural types and geographical regions where damage records are insufficient [1]. The HAZUS
methodology [2] proposes fragility curves for most of the lifeline elements and transportation
infrastructures; however, embankments and cuts are not studied in detail. More specifically, HAZUS
provides fragility curves for roads related to ground failure (i.e. fault offset, liquefaction, and
landslide) without considering the effect of ground shaking. The same functions, which are based in
expert judgment, are also suggested for railway tracks. Nevertheless, the tolerance of railways to
damage is lower, and therefore, these curves are considered unrealistic. Similarly, the REDARS
methodology [3] that was developed for the seismic risk analysis of highway systems does not
provide damage functions for these components. Sasaki et al. [4] presented statistical fragility curves
for tunnels, embankments, and natural slopes based on damage records in major earthquakes that
occurred in Japan (Figure 1). These curves relate different damage ranks as a function of PGA;
however, soil conditions or embankment geometry is not provided in this study.
Another way of damage estimation for extended infrastructures (such as roads and pipelines) is
through functions that estimate the rate of repair or expected number of repairs per unit of length.
Maruyama et al. [5] presented such fragility functions for expressway embankments based on
damage datasets after recent earthquakes in Japan (Figure 2). Two functions are provided, one for
major damage that affects the serviceability of traffic and the other for all damage to expressway
embankments. They are given as a function of peak ground velocity (PGV) and are applicable to
embankments with height of approximately 5–10 m.
The objective of the present paper is to develop analytical fragility curves for highways/railways on
embankments and in cuts due to seismic shaking. The effect of different parameters on the overall
performance of the embankment/cut is examined. In particular, the soil conditions underneath the
geo-construction and the geometry of the embankment/cut are studied. New numerical fragility
curves are derived for different soil types and embankment/cut heights. This methodology is based
on the approach presented by Argyroudis et al. [6] for the fragility assessment of cantilever bridge
abutments.

Figure 1. Empirical fragility curves for embankments (modified Sasaki et al. [4]).

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
FRAGILITY FUNCTIONS FOR HIGHWAY AND RAILWAY EMBANKMENTS AND CUTS 1865

Figure 2. Empirical fragility curves for expressway embankments (modified Maruyama et al. [5]).

2. METHODOLOGY

The procedure for the derivation of analytical fragility curves is described in Figure 3. The effect of soil
conditions and ground motion characteristics on the response of the embankment and the cut is taken
into account by using different, but typical, soil profiles and seismic input motions. The stiffness and
damping parameters are defined based on the results of 1D equivalent linear analysis, in order to be
compatible with the expected strain levels during the earthquake. By defining the damage level
through an appropriate damage index, the fragility curves could be constructed as a function of the
level of seismic excitation. This approach allows evaluating the fragility curves of geotechnical
structures considering the distinctive features of their geometries, the input motion characteristics,
and the soil properties [6].

2.1. Definition of damage states


Earthquake effects on geotechnical structures can generally be grouped into two categories: (1) ground
shaking and (2) ground failure such as liquefaction, fault displacement, and slope instability. The major

Figure 3. Procedure for deriving numerical fragility curves for embankments and cuts.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
1866 S. ARGYROUDIS AND A. M. KAYNIA

factors influencing seismic performance of geotechnical structures include their geometry (shape and
dimensions) and parameters, the properties of the surrounding soil or rock and the severity of the
ground shaking or ground failure. According to Ikehara [7], the factors that are closely related to
damage of embankments are (i) age of the embankment including the method of construction, (ii)
height, and (iii) topographical conditions, including the shape of the embankment. Embankments are
sometimes constructed on liquefaction-susceptible sandy deposits. When the foundation bearing
capacity is lost because of dynamic loading, the embankment might settle and spread (from a few
centimeters to many meters), and cracks open in the road pavement. Typical damage modes include
slide, slip, or slump of embankment, cracks on the surface of the road and/or subsidence of
embankment [8]. In case of cuts, damage is related to slope failure and cracking and
settlement/heave of the road surface.
Different damage levels have been proposed for the various elements of the highway system
infrastructures based on damage description, functionality criteria, or repair costs (e.g. [2–5]). The
damage states that have been defined in the European project SYNER-G [9] for roadway and
railway elements are used in this paper (Table I). These states are described in terms of induced
permanent ground displacement (PGD) of the embankment or the cut surface. In particular, a mean
value of PGD is estimated for minor, moderate, and extensive/complete damage based on a range of
values (min and max). These definitions are based on expert judgment and are consistent with the
existing ones (e.g. [3, 4]). The selected thresholds are also related with the serviceability of the
roadway or railway. These thresholds are also consistent with the study by Tokida [10] who
conducted driving tests using small and large vehicles to establish limit states of settlement for road
embankment in relation with driving speed.

2.2. Derivation of fragility curve parameters


Fragility functions describe the probability of exceeding different limit states given a level of ground
shaking. The level of shaking can be defined using numerous earthquake parameters depending on
the element under consideration. Most common parameters are PGA, PGV, spectral acceleration,
and PGD. Fragility curves are usually described by a lognormal probability distribution function as
follows:

   
1 IM
Pf ds≥dsi SÞ ¼ Φ ln (1)
βtot IM mi

where Pf(·) is the probability of exceeding a particular damage state, ds, for a given seismic intensity
level defined by the earthquake intensity measure, IM (e.g. PGA), Φ is the standard cumulative
probability function, IMmi is the median threshold value of IM required to cause the ith damage
state, and βtot is the total lognormal standard deviation. Therefore, the development of fragility
curves according to Eq. (1) requires the definition of two parameters, IMmi and βtot.
The level of damage (in embankment or cut in the present study) is described by a damage index
expressing the exceedance of certain limit states (Table I), and the fragility curves are estimated
based on the evolution of damage index with increasing earthquake intensity considering the

Table I. Definition of damage states for highway and railway elements.


Permanent vertical
ground displacement (m)

Typology Damage state min max mean Serviceability


Highway ds1. Minor 0.02 0.08 0.05 Open, reduced speeds or partially closed during repair
ds2. Moderate 0.08 0.22 0.15 Closed or partially closed during repair works
ds3. Extensive/complete 0.22 0.58 0.40 Closed during repair works
Railway ds1. Minor 0.01 0.05 0.003 Open, reduced speeds
ds2. Moderate 0.05 0.10 0.008 Closed during repair works
ds3. Extensive/complete 0.10 0.30 0.200 Closed during reconstruction works

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
FRAGILITY FUNCTIONS FOR HIGHWAY AND RAILWAY EMBANKMENTS AND CUTS 1867

associated uncertainties. An example is given in Figure 4, where the different points indicate the results
of analysis in terms of damage index for different levels of earthquake intensity. The solid line is
produced based on a regression analysis, and the median threshold value of the intensity measure
(IMmi) required to cause the ith damage state (dsi) is estimated based on the definition of this
damage state through the damage index. A lognormal standard deviation (βtot) that describes the
total variability associated with each fragility curve has to be estimated. Three primary sources of
uncertainty are considered [2], namely (i) the definition of damage states (βds), (ii) the response and
resistance (capacity) of the element (βC), and iii) the earthquake input motion (demand) (βD). The
total uncertainty is estimated as the root of the sum of the squares of the component dispersions.

3. DEVELOPMENT OF FRAGILITY CURVES FOR EMBANKMENTS AND CUTS

Representative geometries of embankment and cuts are considered for the analyses (Figure 5).
Different heights (h) commonly used in practice are selected, equal to 2.0, 4.0, and 6.0 m for the
embankment and 4.0 and 6.0 m for the cut. Preliminary analyses have shown that negligible damage
is expected for shorter heights.

3.1. Input motions


Five real acceleration time histories from different earthquakes recorded on rock were selected as outcrop
motion for both the 1D ground response and the 2D analyses of the embankments and the cuts. The
earthquakes are Kocaeli (Gebze), Turkey, 1999; Hector Mine (Hector), USA, 1999; Parnitha (Kypseli),
Greece, 1999; Loma Prieta (Diamond Height), USA, 1989; and Umbria Marche (Gubbio-Piana), Italy,
1998. The normalized mean of the acceleration spectra of the selected motions matches EC8 spectrum
for soil class A. Spectral matching is the most commonly proposed earthquake record selection method
providing least variations in the response parameters [11, 12]. In the dynamic analyses, the time
histories are scaled so that their PGAs increase from 0.1 to 0.7 g with a step of 0.1 g.

3.2. Soil properties


Two ideal soil deposits of 50 m depth were considered for the numerical analyses. Soil properties based
on common engineering practice [13] were selected in order to obtain soil types C and D according to
the Eurocode 8 [14]. In particular, an initial value for the undrained shear strength at the top layer, Su0,

Figure 4. Example of evolution of damage with earthquake intensity measure (IM) - definition of thresh-
old median value (IMmi) for the damage state i (dsi) and standard deviation (βD) due to variability of
input motion (demand).

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
1868 S. ARGYROUDIS AND A. M. KAYNIA

Figure 5. Geometry and properties of the embankment and cut under study.

was assumed (equal to 50 kPa for soil type C and 20 kPa for soil type D), while a gradient model for Su
is defined for the next layers (Sun = Su0 + 0.25σvn′, where σvn′ is the average vertical effective stress in
layer n). The variation of Gmax is estimated as Gmaxn = A ∙ Sun where A was taken as equal to 1000 for
soil type C and 800 for soil type D. This corresponds to a clay with plasticity index in the range
25–30%. The values of the other soil parameters were taken as v = 0.35, γ = 19–19.5 kN/m3 for soil
type C and γ = 18–18.5 kN/m3 for soil type D. The variation of initial shear modulus (Gmax) and
shear wave velocity (Vs) with depth is shown in Figure 6. The cut is being excavated in each of the
two soil profiles, while for the embankment, a sandy material is considered (γ = 19.0 kN/m3,
ϕ = 400). The variation of Gmax and Vs with depth for the embankment is shown in Figure 6(b). The
selection of generic soil profiles aims to avoid site-specific issues, in spirit with similar studies for
derivation of fragility curves that are applied for preliminary vulnerability assessment of large
networks and infrastructures.

3.3. 1D equivalent linear ground response analysis


In order to account for the soil nonlinearity in the low to medium strain range, 1D equivalent linear
analyses were carried out to derive realistic parameters for the soil modulus and damping at the
expected shear strain levels for increasing levels of seismic excitation. The resulting average
stiffness and damping parameters were adopted in the finite element (FE) analyses in order to be
compatible with the shear deformation level expected during each utilized earthquake. A similar
approach was adopted by other researchers (e.g. [15–18]).
The 1D ground response analyses were performed using the code EERA [19], which is based on the
traditional equivalent linear method. The variation of the shear modulus G/Gmax and damping ratio D
with the shear strain level γ was defined according to the typical results available in the literature. In
particular, the curves provided by Darendeli [20] as a function of plasticity index and effective stress
were used. For the bedrock, the curves proposed by Schnabel et al. [21] were applied. A total
number of 14 layers were taken to model the 50 m soil profile.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
FRAGILITY FUNCTIONS FOR HIGHWAY AND RAILWAY EMBANKMENTS AND CUTS 1869

Figure 6. Variation of shear wave velocities (Vs) and initial shear modulus (Gmax) with depth for the soil
profiles underneath the embankment and cut and for the embankment (b).
A calibration procedure was followed in order to account for the dependency of both the stiffness
and the damping on the primary shear strain level during the earthquake. In this context, the results
from the 1D analyses for all the input motions scaled at 0.1 g were employed. This assumption was
made because the equivalent linear methods are more reliable for lower strain levels. For higher
strain levels, the effect of nonlinearity is captured by the Mohr–Coulomb yield criterion used for the
soil behavior. The Mohr–Coulomb model is a widely used failure criterion in soil mechanics that
provides a reasonable modeling of soil behavior for stresses exceeding shear strength. The increase
of stiffness with depth is taken into account, while its main limitation is that it does neither include
stress dependency nor stress-path dependency of stiffness or anisotropic stiffness. The effect of the
soil model on the results could be the subject of a specialized study.
For each soil layer in the FE model, a single value of shear modulus, G, was used based on the
average computed G/Gmax ratios. These ratios were computed equal to 0.68 for soil type C and 0.55
for soil type D (Figure 7). The reduction of the Gmax value for the backfill material was estimated in
a similar way. The parameters of the Rayleigh damping were computed based on the average value
of damping for all the layers. The Rayleigh damping parameters were chosen for the frequency
interval 1.4–4.2 Hz for soil profile C and 1.0–3.0 Hz for soil profile D, covering approximately the
first and third modes of the profile. In particular, the average values of damping were about 6% and
8% for soil types C and D, respectively.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
1870 S. ARGYROUDIS AND A. M. KAYNIA

G (MPa)
0 100 200 300
0

10 Soil D
Soil C
15

20

Depth (m)
25

30

35

40

45

50

Figure 7. Average computed values of G with depth for soil type C and D compatible to the developed
ground strains for PGA=0.1 g.

3.4. FE numerical analysis


The analyses were performed with the 2D (plane strain) FE code PLAXIS [22]. The base of the model
was assumed to be rigid, and the lateral sides were equipped with the standard absorbent boundaries.
The FE domain was discretized using 2126 (for the 4.0 m embankment) and 2143 (for the 6.0 m cut)
15-noded plain strain triangular elements. In the area around the embankment and the cut, the mesh
was refined as shown in Figure 8. In this part, the characteristic dimension of the elements (h)
satisfies the condition h ≤ hmax = Vs/(6 ÷ 7)fmax, where Vs is the shear wave velocity and fmax is the
maximum frequency of the seismic motion. Consistently with the 1D ground response analyses, the
foundation soil domain was modeled with 14 horizontal layers to account for the variable stiffness
with depth.
All the analyses started by performing a set of initial analyses to simulate the geostatic stresses and
the construction of the embankment or the cut. Then, the dynamic analyses followed where the seismic
input is applied uniformly at the base of the model. The side slopes of the embankments and the cut
were selected so as to satisfy the stability criteria. All the analyses were performed using the elasto-
plastic soil behavior with Mohr–Coulomb criterion. A representative example of the analysis output
is illustrated in Figure 9 where the vertical displacements in the embankment and the deformed
mesh are illustrated for an earthquake with PGA = 0.4 g.

4. FRAGILITY CURVES

Derivation of fragility curves (i.e. definition of the median threshold value of PGA for each damage
state) is based on the construction of a curve for the computed damage indices in terms of PGD at

Figure 8. Detail of the finite element mesh for embankment.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
FRAGILITY FUNCTIONS FOR HIGHWAY AND RAILWAY EMBANKMENTS AND CUTS 1871

[m]
(a)

(b)

Figure 9. Distribution of the permanent vertical displacements max: 0.182 m (edge), 0.075 m (middle) (a)
and deformed mesh (b). Case: h=4 m, soil type D, input motion Gebze-0.4 g.

the surface of embankment or cut versus increasing seismic intensity in terms of PGA on the ground
surface in the free field. The PGD given in the plot corresponds to an average total displacement
(measured at the edge and middle of the embankment/cut surface). The curves constructed in this
way represent the evolution of damage with increasing earthquake shaking. The curves are
established by a regression analysis, considering PGD as the dependent variable and PGA as the
independent variable. A similar approach has been used for the derivation of the fragility functions
in other studies (e.g. [23, 24]). The first parameter of the lognormal probability function that
describes the fragility curves is the median threshold value of PGA that can be obtained for each
damage state using the regression curve and the definitions of damage states given in Table I for
highway/railway on embankment or in cut. When the computational results do not reach
extensive/complete damage threshold, the fragility curves for this damage state are derived based on
extrapolation of the available results. It is assumed that the ground is not liquefiable, which is in
spirit of other studies focusing on derivation of fragility curves for earthquake excitation. In
addition, the slopes of the embankment are considered not seismically unstable. Therefore, the
permanent deformations of the embankments and cuts are solely due to the nonlinear soil response.
A lognormal standard deviation (β) that describes the total variability associated with each fragility
curve needs to be estimated. Because of the lack of a rigorous estimation, the uncertainty associated
with the definition of damage states (βds) is set equal to 0.4 following HAZUS [2] for buildings. The
uncertainty due to the capacity (βC) is assigned equal to 0.3 based on engineering judgment. The
last source of uncertainty, associated with the seismic demand, is described by the variability in
response (i.e. permanent ground displacement of the embankment or cut) due to the variability of
ground motion. In particular, the variability in the results of numerical simulation that have been
calculated for the different input motions is estimated.
Figure 10 shows the PGD–PGA diagrams for embankment with height h = 6 m and soil profiles
types C and D. Similar diagrams have been produced for all the geometries of the cuts and
embankments. The estimated parameters of the fragility curves are given in Tables II and III for
highway and railway embankments and cuts. The corresponding fragility curves are illustrated in
Figures 11–14. For simplicity and in order to avoid intersection of the different fragility curves for
embankments, the plots are given for an average lognormal standard deviation equal to 0.9 for soil
type C and 0.8 for soil type D.
The plotted fragility curves show that, as expected, the embankment with h = 6.0 m is more
vulnerable than the ones with h = 4.0 and 2.0 m. Similarly, the cut with h = 6.0 m is more vulnerable

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
1872 S. ARGYROUDIS AND A. M. KAYNIA

Figure 10. Evolution of damage in embankment (average displacement) with intensity measure (PGA) for
soil type C and D and h=6.0 m.

Table II. Parameters of analytical fragility curves for embankments.


Soil type C Soil type D
h=2m h=4m h=6m h=2m h=4m h=6m

Typology Damage State μ (g) β μ (g) μ (g) β μ (g) β μ (g) β μ (g) β


Highway Minor 0.65 1.00 0.51 0.90 0.43 0.70 0.47 0.90 0.31 0.70 0.16 0.80
Moderate 1.04 1.00 0.88 0.90 0.78 0.70 0.66 0.90 0.48 0.70 0.29 0.80
Extensive/complete 1.57 1.00 1.42 0.90 1.31 0.70 0.89 0.90 0.72 0.70 0.49 0.80
Railway Minor 0.52 1.00 0.36 0.90 0.30 0.70 0.40 0.90 0.25 0.70 0.12 0.80
Moderate 0.77 1.00 0.57 0.90 0.50 0.70 0.53 0.90 0.37 0.70 0.20 0.80
Extensive/complete 1.17 1.00 0.91 0.90 0.84 0.70 0.72 0.90 0.54 0.70 0.34 0.80

Table III. Parameters of analytical fragility curves for cuts.


Soil type C Soil type D
h=6m h=4m h=6m

Typology Damage state μ (g) β μ (g) β μ (g) β


Highway Minor 0.59 1.00 0.44 1.00 0.38 1.00
Moderate 1.09 1.00 0.92 1.00 0.77 1.00
Extensive/complete 1.90 1.00 1.77 1.00 1.46 1.00
Railway Minor 0.44 1.00 0.31 1.00 0.27 1.00
Moderate 0.74 1.00 0.58 1.00 0.49 1.00
Extensive/complete 1.29 1.00 1.11 1.00 0.93 1.00

than the one with h = 4.0 m. The results also indicate that the vulnerabilities of embankment and cuts
with the same height in soil type D are higher than in soil type C. Furthermore, railway elements are
more vulnerable than highways because of the lower threshold values that define the damage states
(Table I) and describe the lower tolerance of railways to permanent ground displacements. Finally,

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
FRAGILITY FUNCTIONS FOR HIGHWAY AND RAILWAY EMBANKMENTS AND CUTS 1873

Figure 11. Numerical fragility curves for highways on embankments.

the fragility curves show that the cuts are less vulnerable than the embankments when the same soil
conditions and height of the geo-construction are considered.

5. PGA–PGD RELATIONSHIPS FOR EMBANKMENTS

A by-product of the preceding analyses is the relationship between PGA and PGD as a function of the
embankment height (h). This relationship could be useful for design purposes where pseudo-static
stability methods indicate insufficient safety factor. In particular, the expected permanent ground
displacement at the top of the embankment can be predicted for a specific level or range of PGA.
The calculated displacement from these relationships represents an index of seismic performance.
Similar simplified predictive models are proposed for seismic slope displacements by Ambrasseys
and Menu [25], Bray and Travasarou [26] and others. In general, seismic displacement estimates are
approximate in nature because of the complexities of the dynamic response of the soil materials
involved and the variability of the earthquake ground motion. However, when viewed as an index of
potential seismic performance, the calculated seismic displacement is effectively used in practice for
design purposes [26].
The relationships established for the estimation of the PGD at the top of the embankment are given
in Eqs. (2) and (3) for soil types C and D, respectively:

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
1874 S. ARGYROUDIS AND A. M. KAYNIA

Figure 12. Numerical fragility curves for highways in cuts.

soil type C:

PGD ¼ ð0:025h þ 0:089ÞPGA2:1 (2)

soil type D:

PGD ¼ ð0:217h þ 0:113ÞPGA2:5 (3)

where the units of PGD and h are meters, and PGA is expressed as g. These relationships have been derived
for embankments width L = 20.0 m and height from 2.0 to 6.0 m. However, it is believed that they can be
used for typical embankments with dimensions not too different from the aforementioned values.

6. VERIFICATION OF THE PROPOSED FRAGILITY CURVES

Failures of roads and tracks on embankments or in cuts due to ground shaking have been widespread in
recent earthquakes. Cracking of road pavement or displacement of track due to settlement of the
underneath soil were common in recent earthquakes (e.g. [5, 27]). Generally, damage to
embankments and cuts is not a life safety concern; however, depending on the extent of damage, the
duration of repairs can seriously affect the emergency traffic and other activities in the area.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
FRAGILITY FUNCTIONS FOR HIGHWAY AND RAILWAY EMBANKMENTS AND CUTS 1875

Figure 13. Numerical fragility curves for railways on embankments.

Although such damages have been widely observed, complete descriptions of the damage extent and
the embankment properties have rarely been reported. Moreover, information related to the seismic
intensity and soil type at the damage location is usually incomplete. In order to verify the proposed
fragility curves, a comparison is made with the available empirical data reported by Sasaki et al. [4]
(Figure 1) and Maruyama et al. [5] (Figure 2) based on observed damage in embankments in Japan.
The curve provided by Sasaki et al. [4] for damage rank C, which describes the deformation of road
on embankment as a function of PGA, is used for the present verification. Because no information is
given for the soil type or the embankment height, the comparison with the proposed numerical
curves is only qualitative (Figure 15). Moreover, the empirical curve reaches probability of damage
up to 0.1 for PGA up to 0.6 g and it does not follow a lognormal function as the analytical ones. For
the verification purpose, the analytical fragility curves for embankments on soil type C are used.
The plots in Figure 15 reveal that for PGA up to 0.3 g, the empirical curve expresses an average
trend of the analytical curves. For higher values of PGA, the analytical curves give higher damage
probabilities compared with the empirical one. This might be partly due to the way damage is
perceived and interpreted on a qualitative basis.
The fragility functions provided by Maruyama et al. [5] are given in terms of damage ratio Rrate
(number of damage incidents per kilometer) of expressway embankment as a function of PGV. In
order to characterize the probability of having at least one damage, d, along a given length, ℓ, of
embankment, a spatial Poisson process is introduced following the approach by Duenas-Osorio
et al. [28] for pipeline breaks. The probability of road damage occurrence is given by Eq. (4).

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
1876 S. ARGYROUDIS AND A. M. KAYNIA

Figure 14. Numerical fragility curves for railways in cuts.

Figure 15. Comparison between analytical (Soil C) and empirical (Sasaki et al. [4]) fragility curves for
highway embankments.

Pðd > 0Þ ¼ 1  eRrate *ℓ (4)


In the dataset on which these empirical curves are based, damage is recorded with respect to the
kilometer post with the intervals of 10 m. The analytical curve for ‘all damages’ is used for the

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
FRAGILITY FUNCTIONS FOR HIGHWAY AND RAILWAY EMBANKMENTS AND CUTS 1877

Figure 16. Comparison between analytical (Soil C, h=6m) and empirical (Maruyama et al. [5], modified)
fragility curves for highway embankments.

comparison, which is also converted in terms of PGA. In particular, the PGA values are assessed based
on ratios of PGV to PGA at surface as provided by Power et al. [29] for stiff soil, considering M = 7.5
and source to site distance 20–50 km.
The comparison here is made with the analytical fragility curves with h = 6 m as the empirical curves
are applicable to embankments having heights of approximately 5–10 m (Figure 16). Again, because of
lack of information about the soil characterization of the empirical curves, the analytical fragility curves
in soil type C are compared, which is more common in case of high embankments. It is observed that the
reference length (ℓ) that is considered for the conversion of damage ratio to damage probability
according to Eq. (4) has a significant effect in the empirical curves. In particular, the longer the
embankment is, the higher is the probability of damage. The curves derived for length equal to 100
and 10 m for all damages, under the aforementioned assumptions, are plotted in Figure 16. These
curves bracket the analytical ones for the three damage states, especially at higher shaking levels.
There are a number of uncertainties related to the aforementioned comparisons. Namely, the
definition of damage states in the empirical and analytical methods, the unknown soil, and
embankment features in the empirical curves and the assumptions made for the conversion of
damage ratio to damage probability in case of the empirical fragility curve provided by Maruyama
et al. [5]. Nevertheless, the present validation exercise is approximate and has a qualitative nature;
the results are encouraging as they support the general trends in the analytical results. Despite the
uncertainties involved in this comparison, the results indicate that the proposed fragility curves for
highway embankments are generally capable to represent the seismic vulnerability of typical
embankments in road and railway networks.

7. CONCLUSIONS

A methodology to construct numerical fragility curves for geotechnical constructions due to seismic
shaking, which has been presented in a previous study, is briefly described and applied to
embankments and cuts for highways and railways. The response of the embankment and cut is
evaluated based on dynamic 2D numerical analysis using appropriate 2D FE code. The model is
properly considering the dependency of both soil stiffness and damping on the strain level. Seismic
inputs with different frequency content, scaled to different levels of seismic loading (PGA outcrop)
and typical soil profiles, classified as C and D according to EC8, are employed. Defining different
damage states that describe the permanent ground displacements at the surface of embankment or
cut due to the nonlinear soil response and assuming that the ground is not liquefiable, the fragility
curves could be derived as a function of the level of PGA in free field conditions. The peak
acceleration on the ground surface is a more useful parameter in risk assessments, which at the same
time reduces the uncertainties related to the soil profiles. The related uncertainties in the definition
of damage states, the demand, and the capacity of the geo-construction are also considered.
The available fragility models for the vulnerability assessment of highways/railways on
embankments and in cuts are generally limited; they are mainly based on empirical data, without

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
1878 S. ARGYROUDIS AND A. M. KAYNIA

properly considering soil and geometry characteristics. With the proposed numerically derived fragility
curves, the distinctive features of the construction typology, as well as the input motion characteristic
and the soil properties, are considered in a more systematic way. The effect of the foundation soil and
the influence of the height on the performance of the geo-construction are investigated. The reliability
of proposed fragility curves in case of highways on embankments is qualitatively verified using
empirical fragility functions from Japan. Despite the uncertainties in the comparisons, the results
indicate that the proposed fragility curves are generally capable to represent the seismic vulnerability
of typical embankments.
The present study in computing fragility curves is a first attempt to estimate the vulnerability of
highway and railway embankments and cuts based on a numerical approach. The applicability of
the derived fragility curves has some constraints, which are related to the specific assumptions and
selected parameters in the numerical models. These include the thickness of foundation soil
(i.e. 50 m), the properties of soil layers (as described in Section 3.2), the Mohr–Coulomb soil
behavior, the geometry of the embankment and cuts. Therefore, the provided fragility curves can be
used for preliminary vulnerability assessment of similar structures in soil classes C and D.
Finally, several uncertainties are inherent in the modeling approach; however, a standard procedure
to introduce them into seismic fragility functions does not exist. All the common methods are
approximate, while any attempt to reduce the uncertainties requires an abrupt increase of the
computation effort. Future improvements that could result in reduction of uncertainties in the
proposed fragility curves include the randomization of soil properties and geometry (i.e. thickness of
soil profiles), sensitivity analyses regarding the soil failure criterion, use of larger set of input
motions, estimation of liquefaction, and slope failure effects.

ACKNOWLEDGEMENTS
This work has been developed in the framework of the research project SYNER-G: Systemic Seismic Vul-
nerability and Risk Analysis for Buildings, Lifeline Networks and Infrastructures Safety Gain, funded by the
European Commission (FP7-ENV-2009-1-244061).

REFERENCES
1. Pitilakis K, Crowley E, Kaynia A (eds). SYNER-G: typology definition and fragility functions for physical elements
at seismic risk. Geotechnical, Geological and Earthquake Engineering 2014; 27, Springer. DOI: 10.1007/978-94-
007-7872-6
2. HAZUS-MH: User’s Manual and Technical Manuals. National Institute of Building Sciences. report prepared for the
Federal Emergency Management Agency, Washington, D.C. USA, 2004.
3. Werner SD, Taylor CE, Cho S, Lavoie J-P, Huyck C, Eitzel C, Chung H, Eguchi RT. REDARS 2: methodology and
software for seismic risk analysis of highway systems. MCEER-06-SP08, University at Buffalo, State University of
New York, 2006.
4. Sasaki Y, Shimizu Y, Sunasaka Y. Development of fragility curves of civil structures based on observed structural
damages in past great earthquakes. Proceedings of Conference by Institute of Social Safety Science, 2000; 17–20.
(in Japanese)
5. Maruyama Y, Yamazaki F, Mizuno K, Tsuchiya Y, Yogai H. Fragility curves for expressway embankments based on
damage datasets after recent earthquakes in Japan. Soil Dynamics and Earthquake Engineering 2010; 30:1158–1167.
6. Argyroudis S, Kaynia AM, Pitilakis K. Development of fragility functions for geotechnical constructions: application
to cantilever retaining walls. Soil Dynamics and Earthquake Engineering 2013; 50:106–116.
7. Ikehara T. Damage to railway embankments due to the Tokachioki earthquake. The Japanese Geotechnical Society
1970; X(2):52–71.
8. Towhata I. Geotechnical Earthquake Engineering. Springer-Verlag: Berlin Heidelberg, 2008.
9. Kaynia AM (ed.). Guidelines for deriving seismic fragility functions of elements at risk: buildings, lifelines, transpor-
tation networks and critical facilities. Publications Office of the European Union 2013. DOI: 10.2788/19605
10. Tokida K. Seismic potential improvement of road embankment. Advances in Geotechnical Earthquake Engineering -
Soil Liquefaction and Seismic Safety of Dams and Monuments, Moustafa A (ed.). 2012; 269–296. ISBN: 978-953-
51-0025-6 InTech, DOI: 10.5772/28710.
11. Katsanos EI, Sextos AG, Manolis GD. Selection of earthquake ground motion records: a state-of-the-art review from
a structural engineering perspective. Soil Dynamics and Earthquake Engineering 2010; 30(4):157–169.
12. Iervolino I, Maddaloni G, Cosenza E. Eurocode 8 compliant real record sets for seismic analysis of structures.
Journal of Earthquake Engineering 2008; 12:54–90.
13. Andersen KH, Lunne T, Kvalstad T, Forsberg CF. Deep water geotechnical engineering. Proc. of the XXIV National
Conference of the Mexican Society of Soil Mechanics: Aguasclientes, 26-29 November 2008.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe
FRAGILITY FUNCTIONS FOR HIGHWAY AND RAILWAY EMBANKMENTS AND CUTS 1879

14. EN 1998-1 Eurocode 8: design of structures for earthquake resistance-part 1: general rules, seismic actions and rules
for buildings. CEN, Bruxelles, 2004.
15. Gil LM, Hernandez E, De la Fuente P. Simplified transverse seismic analysis of buried structures. Soil Dynamics and
Earthquake Engineering 2001; 21:735–740.
16. Hashash YA, Hook J, Schmidt B, Chiang YJ. Seismic design and analysis of underground structures. Tunnelling and
Underground Space Technology 2001; 16:247–293.
17. Gazetas G, Gerolymos N, Anastasopoulos I. Response of three Athens metro underground structures in the 1999
Parnitha earthquake. Soil Dynamics and Earthquake Engineering 2005; 25:617–633.
18. Amorosi A, Boldini D. Numerical modelling of the transverse dynamic behaviour of circular tunnels in clayey soils.
Soil Dynamics and Earthquake Engineering 2009; 29:1059–1072.
19. Bardet JP, Ichii K, Lin CH. EERA: a computer program for equivalent-linear earthquake site response analyses of
layered soil deposits. Department of Civil Engineering, University of Southern California, 2000; 40.
20. Darendeli MB. Development of New Family of Normalized Modulus Reduction and Material Damping Curves.
University of Texas: Austin TX, U.S., 2001.
21. Schnabel PB, Lysmer J, Seed HB. SHAKE: a computer program for earthquake response analysis of horizontally lay-
ered sites. Report No. UCB/EERC-72/12, Earthquake Engineering Research Center, University of California,
Berkeley, 1972; 102.
22. Plaxis 2D, reference manual, version 9, 2011.
23. Argyroudis S, Pitilakis K. Seismic fragility curves of shallow tunnels in alluvial deposits. Soil Dynamics and
Earthquake Engineering 2012; 35:1–12.
24. Nielson BG, DesRoches R. Seismic fragility methodology for highway bridges using a component level approach.
Earthquake Engineering and Structural Dynamics 2007; 36:823–839.
25. Ambraseys NN, Menu JM. Earthquake induced ground displacements. Earthquake Engineering and Structural Dy-
namics 1988; 6:985–1006.
26. Bray J, Travasarou T. Simplified procedure for estimating earthquake-induced deviatoric slope displacements.
Journal of Geotechnical and Geoenvironmental Engineering 2007; 133(4):381–392.
27. Kayen R, Collins B, Abrahamson N, Ashford S, Brandenberg SJ, Cluff L. Investigation of the M6.6 Niigata-Chuetsu
Oki, Japan, earthquake of July 16. US Department of the Interior & US Geological Survey, Open File Report
2007-1365, 2006.
28. Duenas-Osorio L, Craig JI, Goodno BJ. Seismic response of critical interdependent networks. Earthquake Engineer-
ing and Structural Dynamics 2007 4; 36:285–306.
29. Power MS, Rosidi D, Kaneshiro J, Gilstrap SD, Chiou S-J. Summary and evaluation of procedures for the seismic
design of tunnels. Final Report for Task 112-D-5.3(c). FHWA Contract No. DTFH61-92-C-0012. Multidisciplinary
Center for Earthquake Engineering Research: Buffalo, New York, U.S., 1998.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1863–1879
DOI: 10.1002/eqe

You might also like