FEW THOUGHTS ON THE NUMERICAL SIMULATION OF

SOIL, STRUCTURE AND EARTHQUAKE INPUT

Anastasios Sextos a
a
Assistant Professor, Aristotle University of Thessaloniki, Department of Civil Engineering,
Division of Structural Engineering, asextos@civil.auth.gr

1 INTRODUCTION

The importance of soil-structure-interaction (SSI) for the assessment of the dynamic response
of bridges has been widely recognized during the last decades in numerous research studies.
Nowadays, there is strong analytical and experimental evidence to support that soil-structure
interaction phenomena depend on soil properties, soil stratification and topography as well as
on the earthquake input frequency content and amplitude, while SSI may drastically affect the
overall inelastic dynamic response of bridge structures. Moreover, a large variety of very
powerful computational tools is available which utilize an enormous amount of computational
power that was certainly not available just few years ago. On the other hand, despite the
significant progress made regarding both the understanding of the nature of this phenomena
and the availability of modern tools, a large discrepancy is observed in the estimation of the
structural response using different simulation strategies. As a result, the prediction of the
dynamic behaviour of a bridge while considering the presence of the supporting soil and/or
the embankment-abutment system, is found to be very sensitive to a large number of
parameters inclusive of seismic motion itself, which is the most significant source of
uncertainty especially in case of bridges with significant length or supported by soil profiles
that are abruptly changing along their axis. Along these lines, the scope of this paper is
twofold: (a) to comparatively assess the reliability of the various computational simulations of
different complexity that are currently available and (b) to attempt to quantify the inherent
physical uncertainty of soil-structure interaction compared to the uncertainty attributed to the
numerical simulation of the soil-structure system and the estimation of the earthquake input.

Keywords: soil-structure interaction, computational modelling, abutment–embankment

system, multi-platform analysis, spatial variability of ground motion, performance-based
earthquake engineering.

2 COMPUTATION MODELLING OF EMBANKMENT-ABUTMENT--BRIDGE

SUPERSTRUCTURE AND SUBSOIL SYSTEMS

Consideration of the contribution of bridge lateral boundary conditions in the overall seismic
response of bridges, has illustrated the significant role played by the embankment-foundation-
abutment system not only in terms of the dynamic characteristics and response of the bridge
(Goel and Chopra, 1997, Mackie and Stojadinovic, 2002, Dicleli, 2005, Kotsoglou and
Pantazopoulou, 2007) but also regarding the modification of the incoming seismic motion
(Zhang and Makris, 2002). Earthquake damage reports and laboratory tests have also
2 A. Sextos – Earthquake Engineering by the Beach Workshop, July 2-4 2009, Capri, Italy.

indicated that abutment failure commonly caused by rotational and/or translational outward
movement of the toe or even loss of subsoil bearing capacity is fairly common, hence refined
analysis of the overall system is required. As a result, it is indeed a challenge to implement
the computational tools and resources required to simulate the multi-parametric and complex
nature of both the dynamic pier-foundation-subsoil and deck-abutment-embankment
interaction as well as the shear deformation and failure of RC members (i.e. piers and piles),
since coupled modelling of all these systems still requires extensive computation effort due to
the model size and/or behavior complexity. It can be also argued that given the above
complexity and computational demand, it is rather subjective whether a single software
package exists that could possibly combine all the features required for advanced simulation
of the non-linear response of bridge, foundations and abutments and their supporting soil.
Along these lines, recent research has investigated the application of distributed
computational simulation as a means to comparatively assess the limitations and challenges of
the most advanced modelling approaches currently available for the study of complex SSI
systems. It is noted that, multi-platform simulation is one of the most promising approaches of
this kind and was initially developed to accommodate multi-site hybrid simulation (Spencer et
al., 2006). In particular, the dynamic response of full scale specimens that are physically
separated is properly controlled with the use of purpose-specific coordination software that
made feasible the incorporation of various numerical analysis platforms in the sub-structuring
process. This concept has also been successfully applied (Kwon and Elnashai, 2008) for the
coordination of purely numerical analysis modules (in contrast to the hybrid simulation
application) for the case of real bridges in the U.S. for various soil conditions, as well as for
the study of the potential impact of liquefaction susceptibility (Kwon et al., 2008). The
advantage of this approach is that the appropriate selection and combination of different
analysis packages, enables the concurrent use of the most sophisticated models and features of
each package for each corresponding part of the system. In other words, different software
can be used for different system components (i.e. abutments, superstructure and supporting
pile groups) depending on the foreseen material constitutive laws and geometry.
In order to investigate herein the range of applicability of the advanced computational tools
and methods currently available for simulating the embankment-abutment-bridge interaction,
a typical, real and already built, overcrossing in Greece is chosen to serve as a benchmark and
four different alternative modelling approaches are explored, namely: 1) a bridge frame model
supported on complex dynamic impedance matrices that are specifically calculated for pile
foundations and abutments; 2) a 3-Dimentional spring-supported frame model consisting of
the bridge, its abutment and its foundation, 3) a refined 3-Dimentional solid model of the
overall superstructure-abutment-embankment system and; 4) a multi-platform scheme (Kwon
and Elnashai, 2008) using appropriate system sub-structuring. The analysis is performed both
in the linear and the non-linear range. An overview of the bridge structure studied and the
comparative assessment of the aforementioned approaches is presented in the following.

2.1 Overview of the bridge studied

The particular bridge adopted for study (Figure 1) is an overpass (overcrossing) along the
Egnatia highway, a large road network that has been constructed in northern Greece with
more than 646 bridges built of a total of 40km length most of which are structures of
relatively small dimensions (i.e. L<100m). It is a three-span, symmetric structure of 70m
length (span lengths are 19, 32 and 19m respectively), curved in elevation (maximum camber
of 8%), that intersects the highway axis at an angle of 75.3°. The deck is 11m wide and 1.60m
high. The prestressed deck has a hollow T-beam-like section and is supported on two circular
piers of 1.70m diameter and 8.50m height which are monolithically connected to the
Few thoughts on the numerical simulation of soil, structure and earthquake input 3

superstructure and the foundation. At the abutments (which have a 10.501.20m wall section
of 5.0m height), the deck is connected through two pot bearings that permit sliding along the
two principal bridge axes and a sliding joint separates the deck from the backwall. Seismic
forces are also resisted by the activation of stoppers (in the transverse direction) which are
constructed at the seating of the abutments. The foundation on the other hand is deep, due to
the soft clay formations characterizing the overall area. The pier foundation consists of a 22
pile group of 28.0 to 32.0m long piles, connected with a 1.605.05.0m pile cap, while the
abutments are supported on a 14 pile row 27 to 35.0m long at 2.80m axial spacing, all piles
having equal diameter of 1.0m. The bridge was designed for normal loads according to the
German Norms (i.e. DIN 1055, 1045, 1072, 1075, 1054, 4227, 4085, 4014) while the seismic
design was carried out according to the Greek Seismic Code EAK 2000 and the relevant
Greek standards E39/99 for the seismic design of bridges. The bridge site is located in the
Seismic Zone I which is equivalent to a peak ground acceleration of 0.16g. The behaviour
factors of the system adopted for design according to the E39/99 document were qx=2.50,
qy=3.50 and qz=1.00 for the response in the three principal directions, respectively. The target
displacements of the bridge under study for the two directions, the two alternative soil
conditions and the two earthquake levels (i.e. design earthquake and twice the design
earthquake) are also depicted in Figure 2 (the complete calculation process can be found in
Potikas, 2006). It is noted that for twice the design earthquake in the longitudinal direction,
the joint is expected to close. Consequently, the overall bridge system stiffness in the
longitudinal direction is significantly increased due to the activation of the backfill-abutment-
foundation-soil subsystem. It is also noted that in the transverse direction, although damage is
indeed minor for the case of soft foundation soil even for displacements corresponding to
twice the level of the design earthquake, the abutment piles were found to suffer significant
damage due to shear failure at their head when the supporting soil is stiff (Kappos et al.,
2007). This situation is apparently detrimental because the abutments can no longer resist
their own earth pressure, hence the bridge stability is jeopardized and the high ductility of the
middle piers is never utilized. Given the above observations it is clear that for the particular
bridge under study, the role played by the abutment is crucial and hence the appropriate
modelling of the bridge lateral boundary conditions is necessary.

Figure 1: Longitudinal cross-section of the bridge (above) and indicative overview of a typical
overcrossing along Egnatia Highway (bottom).
4 A. Sextos – Earthquake Engineering by the Beach Workshop, July 2-4 2009, Capri, Italy.

Figure 2: Pushover curve and seismic assessment of the overall system studied in the logitudinal direction
for two different soil categories (after Kappos et al., 2007).

In order to investigate and demonstrate the current capabilities of the various analysis
approaches, four different models were developed. The assumptions made in each cases and
the performance of all models is summarized in Sextos and Taskari (2008), while a brief
description of the overall concept is described in the following:

2.2 Frame bridge on Spring and Dashpot systems (Model 1)

First, a bridge frame model supported on complex dynamic impedance matrices that are
specifically calculated for pile foundations and abutments is developed. This consists of the
superstructure (Figure 3) whose pies are assumed to be connected to 6-DOF spring and
dashpot systems with dynamic properties computed using the computer code ASING (Sextos
et al., 2003) for coupled translational and rocking modes of vibration and given the
foundation and soil properties described in Sextos and Taskari (2008). Pile-to-pile interaction
was accounted through the formulation of the particular computer code. The abutment
dynamic stiffness and damping is computed according to Zhang and Makris (2002).
Kinematic interaction was ignored. The analysis was performed using the widely used FE
program Sap2000 and represents the most refined approach that could be expected to be
implemented in the design practice.

2.3 Frame bridge on spring supported abutment and foundation (Model 2)

This approach involves the FE model illustrated in Figure 4 inclusive of the superstructure,
the abutments as modelled with 2D shell elements in 3D space, as well as the pile foundations
modelled using beam-on-dynamic springs. Spring and dashpot values were computed as in
Model 1 but distributed based on the area of influence of each particular spring. The analysis
was also performed using the FE program Sap2000.

2.4 Frame bridge on 3D solid embankment-foundation-abutment (Model 3)

Bridge superstructure is discretized using 3D frame elements which are then connected to a 3-
D (solid) abutment-foundation-embankment system at both lateral supports of the deck
(Figures 5 and 6). The piers are assumed to be supported on the 6-DOF dynamic impedance
matrices described above while retaining the same properties as previously. Soil was assumed
Few thoughts on the numerical simulation of soil, structure and earthquake input 5

as linear elastic for comparison purposes. The analysis was performed with the advanced FE
software Abaqus and the pushover curve for the longitudinal direction is depicted in Figure 7.

2.5 Distributed simulation: frame bridge on 3D soil embankment-foundation-abutment

system (model 4a)
In this approach, which is the most refined compared to the previously described ones, the
structure was subdivided into several modules that are computationally simulated using
different computer codes. The analysis of the distributed modules was coordinated with the
aid of UI-SimCor (Spencer et al., 2006), an enhanced Matlab based script with its own GUI
that was developed by University of Illinois in order to coordinate either software or hardware
supporting NEESgrid Teleoperation Control Protocol (NTCP) as well as TCP-IP connections
outside of the NEES system. The basic concept of the framework is that analytical models
associated with various platforms or experimental specimens are considered as super-
elements with many DOFs. The main routine enforces static equilibrium during gravity load
application and conducts dynamic time integration thereafter. Each of these elements are
solved on a single computer or on different computers connected through the network.
Interface programs for analytical platforms have been developed for Zeus-NL (Elnashai et al.
2002), OpenSees (McKenna and Fenves 2001), FedeasLab (Filippou and Constantinides
2004), and ABAQUS. In the particular analysis of the Egnatia Highway overcrossing, two
different analysis packages were coordinated by UI-Simcor, corresponding to three
distributed modules, namely: (1) the bridge sub-system, which was modelled using the
verified inelastic dynamic analysis program FedeasLab (Filippou and Konstantinides, 2004),
the left (2) and right (3), pile-supported, abutment - embankment system that was
modelled using 3D solid solid elements and the commercial FE package Abaqus (Model 4a
in Figure 8). In order to minimize computational time, a relatively simpler 3D abutment-
embankment system was adopted after appropriate calibration of damping and stiffness along
the overall height with the refined 3-Dimentional soil Model 3.

Figure 3: Overview of dynamic spring-supported Model 1

Few thoughts on the numerical simulation of soil, structure and earthquake input 7

Longitudinal direction
14000

12000

10000
Force (kN)

8000

6000

4000

2000

0
0 0.02 0.04 0.06 0.08 0.1
Displacement (m)

Figure 7: Comparison of the pushover curves derived for the spring supported abutment (Kappos et al.,
2007) and 3-D model of the Egnatia Highway bridge along the longitudinal direction
(Sextos and Taskari, 2008).

Figure 8: Distributed computational simulation in Model 4a (Sextos and Taskari, 2008).

2.6 Comparative assessment of single- and multi-platform approaches in the linear

elastic range
Given the detailed data available for the particular case-study, an effort was made to use
common assumptions regarding earthquake excitations and solution algorithms. Along these
lines, the Kozani, Greece earthquake (PGA=0.19g) was uniformly applied in all cases, while
the Hilbert-Hughes-Taylor integration method was used, with time step Δt=0.01s and a total
of 1000 steps (10 seconds of input). A uniform damping value of 5% was assumed for the
first and second modes of vibration, defined through the Rayleigh alpha and beta
corresponding factors. Gaps and stoppers that have been designed for the particular structure
were ignored to ensure maximum possible activation of the embankment-abutment system.
Backfill and foundation soil properties were also taken identical between models 3 and 4
based on the actual soil properties described in section 2. All analyses were conducted in the
elastic range and the excitation was performed in the longitudinal direction. Parametric
analysis was also performed to investigate the relative influence of various assumptions that
inevitably varied between the four approaches i.e. spring and dashpot constants of Models 1
and 2 in contrast to Poisson’s ratio and modulus of elasticity for soils in Models 3 and 4,
Few thoughts on the numerical simulation of soil, structure and earthquake input 7

Longitudinal direction
14000

12000

10000
Force (kN)

8000

6000

4000

2000

0
0 0.02 0.04 0.06 0.08 0.1
Displacement (m)

Figure 7: Comparison of the pushover curves derived for the spring supported abutment (Kappos et al.,
2007) and 3-D model of the Egnatia Highway bridge along the longitudinal direction
(Sextos and Taskari, 2008).

Figure 8: Distributed computational simulation in Model 4a (Sextos and Taskari, 2008).

2.6 Comparative assessment of single- and multi-platform approaches in the linear

elastic range
Given the detailed data available for the particular case-study, an effort was made to use
common assumptions regarding earthquake excitations and solution algorithms. Along these
lines, the Kozani, Greece earthquake (PGA=0.19g) was uniformly applied in all cases, while
the Hilbert-Hughes-Taylor integration method was used, with time step Δt=0.01s and a total
of 1000 steps (10 seconds of input). A uniform damping value of 5% was assumed for the
first and second modes of vibration, defined through the Rayleigh alpha and beta
corresponding factors. Gaps and stoppers that have been designed for the particular structure
were ignored to ensure maximum possible activation of the embankment-abutment system.
Backfill and foundation soil properties were also taken identical between models 3 and 4
based on the actual soil properties described in section 2. All analyses were conducted in the
elastic range and the excitation was performed in the longitudinal direction. Parametric
analysis was also performed to investigate the relative influence of various assumptions that
inevitably varied between the four approaches i.e. spring and dashpot constants of Models 1
and 2 in contrast to Poisson’s ratio and modulus of elasticity for soils in Models 3 and 4,
8 A. Sextos – Earthquake Engineering by the Beach Workshop, July 2-4 2009, Capri, Italy.

embankment finite element mesh dimensions and size, among many others. It was concluded
that the parameter related to the maximum level of uncertainty was the critical embankment
mass that was expected to be activated during the particular earthquake excitation and most
importantly, the means to simulate its effect in the framework of the four different analysis
strategies adopted. In contrast to the validation case (section 4) the value of the single point
mass that was used for Models 1 and 2 at the lateral boundaries of the bridge to represent the
‘active’ embankment-abutment system, was predicted independently (blindly) based on the
concept of critical embankment length (Zhang and Makris, 2002) and without any calibration
to 3-D solid Models 3 and 4, where the activated embankment mass was inherently
considered. Next, the dispersion in the dynamic response of the bridge due to the assumptions
and modelling approach adopted is illustrated in Figure 9. In particular, it is seen that
following four different approaches to consider the effect of embankment-abutment-
superstructure interaction, the maximum longitudinal displacement of the deck lies in the
range of 0.6-1.0cm whereas the fundamental period of the overall system may also differ by
more than 100% despite the effort to use compatible properties where available. Further
response measures (i.e. middle pier stresses) are not presented herein due to lack of available
space; however, it is noted that the dispersion is of the same order. It is also seen (Sextos and
Taskari, 2008) that multi-platform analysis is a very promising concept since it provides
stable results within the envelope of the response produced by the other three approaches
while enabling the consideration of 3-Dimentional geometry without exceeding the
computational time required for a conventional single-platform 3-D modelling of the entire
embankment-abutment-bridge system.

2.7 Comparative assessment of single- and multi-platform approaches in the nonlinear

range
Following the assessment of the four different approaches examined in the linear range, an
effort was made to compare the response of the bridge in the non-linear range. The dynamic
impedance matrices of Model 1 which represent the abutment-embankment as well as the
dynamic pile group stiffness were properly modified based on intensity soil properties
according to the EC8-Part 5 provisions. The same procedure was applied for the case of the
Model 2 through appropriate multi-linear springs and distinct dashpots that were also based
on the aforementioned modified soil properties along the pile length. On the other hand, the
Mohr-Coulomb constitutive model implemented in Abaqus was utilised to simulate the non-
linear soil behaviour in the case of the, more refined Model 3. As for the model 4, secant
stiffness based on the detailed pushover analysis results along the longitudinal direction
(Figure 7) was applied within the framework of an equivalent linear analysis. It is noted that
the lateral assumption is not introducing any additional uncertainty to the problem since it
essentially introduces an ‘exact’ (i.e., derived through refined pushover analysis) non-linear,
force-displacement relationship at each abutment control point that is used thorugh a multi-
platform process that is also inherently pseudo-static. The dispersion of the results is
presented in the Figure 10 where it is seen that multi-platform analysis is not only feasible in
the non-linear range, but it leads to comparable results with those derived with more
conventional approaches both in terms of longitudinal displacement amplitude of the deck and
of the fundamental period increase of the system as a whole. Moreover, it is noted that the
multi-platform simulation coordinated by UI-Simcor, may also have significant advantages
regarding for capturing the inelastic response of the R/C piers, as it can combine a number of
specialised constitutive models and software currently available.
Few thoughts on the numerical simulation of soil, structure and earthquake input 9

Figure 9: Linear elastic dynamic response of the bridge deck for various embankment-abutment-
foundation-superstructure interaction modelling approaches (linear elastic range)

Figure 10: Nonlinear dynamic response of the bridge deck for various embankment-abutment-
foundation-superstructure interaction modelling approaches (non-linear range)

3 EARTHQUAKE INPUT MOTION FOR THE CASE OF LONG BRIDGES

From all the parameters that define the non-linear dynamic response of complex and extended
structures, such as bridges, the input motion has by far the highest level of uncertainty. The
last two decades, different approaches, methodologies and tools have been developed to deal
with this uncertainty and put it in a framework that can be quantified and thus uniformly
interpreted by the practicing engineers and the scientific community (Harichandran, and
Vanmarcke, 1986, Luco and Wong, 1986, Hao, 1989, Zerva, 1990, Deodatis, 1996, Monti et
al., 1996, Tzanetos et al., 2000, Shinozuka et al, 2000, Pinto et al., 2002, Sextos et al., 2003a
and 2003b., Sextos et al., 2004, Lupoi et al., 2005, Nuti and Vanzi, 2005, Lou and Zerva,
2005, Burdette et al., 2008, Burdette and Elnashai, 2008, Lupoi, 2009, Zerva, 2009). The
10 A. Sextos – Earthquake Engineering by the Beach Workshop, July 2-4 2009, Capri, Italy.

extensive work on predicting or producing refined response spectra as well as the large data
of actual ground motions recorded on different soil and seismotechtonic conditions that are
currently available are a precious source of information that has allowed a better
understanding of both the characteristics of seismic motion and its implications on the
earthquake performance of such structures. Additionally, the increasingly enhanced
capabilities now available for inelastic dynamic analysis provide a very good estimate of their
expected response under earthquake loading. Nevertheless, it is not obvious that all this
progress in developing seismic hazard scenarios, predicting suitable seismic motions and
performing advanced numerical analyses of the superstructure is indeed reflected to the
overall design reliability of a system, since the uncertainty of the inherent hypothesis made
that a (long) structure is uniformly excited with an identical (natural or artificial) earthquake
motion cannot be easily justified.
It is true that the critical question is not whether seismic motion is indeed different along an
extended structure; this is almost self-evident, physically justified and the sources of spatial
and temporal variations of seismic motion have been well identified (Der Kiureghian and
Keshishian, 1997) as: a) waves that travel at a finite velocity, so that their arrival at each
support point is out of phase b) wave coherency loss in terms of gradual reduction of their
statistical dependence with distance and frequency, due to multiple reflections, refractions and
superpositioning during propagation and c) local site effects. As a result of all the above
sources, both peak ground acceleration and frequency content of the motion may be
significantly different among the various foundation points. Moreover, although often
neglected, the potential filtering at the foundation level that results from the relative flexibility
of the foundation-soil system components, is an additional source of seismic motion
variability (Sextos et al., 2003a). In addition to the above theoretical justification, Spatial
Variability of Earthquake Ground Motion (SVEGM) has also been recorded in various
densely instrumented arrays all over the world (SMART-1 and LSST-Lotung in Taiwan,
Chiba in Tokyo, USGS-Parkfield and Imperial Valley in California, as well as Euroseis-Test
in Greece among others), hence the fact that a long structure is expected to be excited with
asynchronous and partially uncorrelated seismic forces is evident and well documented.
The main question therefore, is how the designer may produce ‘reasonable’ spatially varying
suites of ground motion, what the response of a structure would be under such an
asynchronous excitation, whether the final response is detrimental compared to the prediction
made assuming a structure uniformly excited in the time domain, and especially whether it
can be indeed predictable in advance during the design process. The answer to this question is
difficult not only due to the complexity in predicting incoherency patterns but also, due to the
significant coupling between earthquake input, dynamic characteristics of the soil-structure
system (particularly in terms of foundation compliance and energy dissipation) at the soil-
foundation interface. What follows is therefore an effort to investigate the above phenomena
with emphasis on numerous bridge structures studied. The methodology was also applied for
the case of extended Byzantine city Walls (Stylianidis and Sextos, 2009) as a means to
examine more general trends of the effect of asynchronous excitation to large structures that
are not limited by the typical bridge configurations and structural systems. The overall
methodology developed for the study of the particular problem is described in detail
immediately after and the results from the study of the 27 bridges studied are presented in a
macroscopic, qualitative way in order to permit drawing of more general conclusions. The
section concludes with some general observations and recommendations as regard to the
assessment and design of extended structures.
Few thoughts on the numerical simulation of soil, structure and earthquake input 11

3.1 Overview of the parametric analysis scheme

The spatially variable earthquake input has been generated for various combinations and
scenarios using the computer code ASING (Sextos et al., 2003b). With the aid of this
software, which is based on the methodology presented by Sextos et al. (2003a), 27 bridges
(including 5 actual, already built, structures) that present many different structural
characteristics were studied and analysed with the assumptions described in Table 1; they
were subjected to a large number (271) of spatially variable earthquake ground motion
scenarios as presented in detail elsewhere (Sextos and Kappos, 2008).
For all the bridges studied a uniform excitation analysis (base excitations assumed identical)
was performed as a reference level hence the influence of spatial variability was expressed in
terms of the SVEGM-effect ratio:

maximum displacement or bending moment in time for asynchronous excitation

 (1)
maximum displacement or bending moment in time for synchronous excitation

For each of the 271 cases, 3 or 5 realizations of the input motion have been generated and the
average value of ρ was obtained. It is herein recognised that ideally, a Monte Carlo scheme
should have been applied, however, the large number of structures studied and the phenomena
simultaneously considered (i.e., spatial variability, soil-structure interaction, site effects)
render such an approach essentially prohibitive. It is also noted that the 27 bridges studied
have been initially excited with records compatible with different target earthquake motion as
an effort to quantify the importance of the ground motion frequency content in the light of
considering the effect of spatial variability. Therefore in order to account for both the above
issues, for all cases studied, the effect of spatial variability is expressed solely in relative
terms, that is, using the ratio of eq. (1), and no conclusions are drawn based on absolute terms.
Moreover, aiming to additionally evaluate the new EC8 provisions, an additional set of
analyses has been performed using the Eurocode 8 design spectrum corresponding to a peak
ground acceleration equal to 0.24g, ground type B and a behaviour factor q=1 as the common
reference frequency content for all bridges specifically studied. The spatially varying ground
motion time-histories have equal duration of 20 sec, and reflect an identical prescribed
coherence function. In particular, the widely used Loco and Wong (1986) loss of coherency
pattern was adopted, assuming moderate coherency drop (i.e. parameter α was set to 2.510-4
sec/m) while the apparent propagation velocity Vapp was set equal to 1000 m/sec.

3.2 Impact of analysis and design assumptions

Based on the analyses of the 27 bridges briefly presented in Table 2, an effort was made to
identify and quantify the parameters of SVEGM analysis that have the maximum impact on
the action effects, with focus on two critical response quantities, deck displacements and pier
bending moments. In Figure 11, the variation of the average value of ρ (eq. 1) due to various
analysis and modelling assumptions is illustrated. In brief, for each specific assumption or
decision regarding the consideration (or not) of a particular phenomenon (for example,
kinematic or inertial interaction, site effects, different frequency content, or direction in which
the excitation is applied), SVEGM analysis was performed and deck displacements and pier-
base bending moments were computed and compared with the ones derived through
synchronous excitation dynamic analysis. The ratios ρ were thus calculated for each pier of
each particular bridge for which the specific analysis assumption was ‘switched on and off’.
As a result, each bridge is characterised by a maximum and a minimum value of ρ due to the
12 A. Sextos – Earthquake Engineering by the Beach Workshop, July 2-4 2009, Capri, Italy.

specific decision that was made during the SGEVM analysis. Next, the mean value of the
maximum ρ values among all bridges, plus one standard deviation (μ+σ) and the mean value
of the minimum ρ value among all bridges minus one standard deviation (μ-σ) are derived and
plotted in Figure 11 to represent the variation of the impact of each decision in terms of pier
base bending moments and deck displacements. As previously, it has to be stressed that
processing statistically sets of data resulting from such different cases both in terms of
structural characteristics and analysis complexity is an inevitably subjective effort, especially
given the fact that the study of the potential importance of each decision made is based on
samples of unequal size (there are some parameters whose effect has been studied for two
bridges only). Therefore, the illustration of Figure 11 can only be interpreted as a rough,
qualitative indication of the tendencies observed in a problem for which enough and
completely consistent data could only be gathered with great difficulty (and they are not
currently available). As an effort, therefore, to take advantage of the wide variety of the
bridge configurations studied and given the above significant limitations, three very general
comments can be possibly made:
First, it seems that accounting during the design process for local site response, multiple
reference (target) earthquake motions, coherency loss, and the coupling between soil-structure
interaction and site effects, has an impact on the accuracy of the procedure for tackling
SVEGM effects, which is higher than that of the variation of the parameters assumed for the
coherency models to be used, the abutment flexibility, or the stiffness of concrete members
(the corresponding range of variation of the value of ρ is significantly smaller in the latter
cases). In other words, it seems that errors resulting from completely neglecting significant
physical phenomena during the analysis are higher than those due to improper selection of the
parameters involved in the description of these phenomena; furthermore, uncertainty in these
parameters could be tackled through parametric analysis. This observation is also reflected on
Figure 12 where the calculated rotational ductility demand μθ at the base of Bridge 08 piers
according to the ‘classic’ and the ‘comprehensive’ approach is presented (Sextos et al., 2002).
It is seen that by considering spatial variability, soil structure interaction and local site
response phenomena the ductility demand is modified significantly at almost all piers,
compared to the ‘classic’ approach where all the above phenomena are ignored, even for fixed
values for the coherency models used and the stiffness of abutments, piers and piles.
Secondly, it could be claimed that consideration of spatial variability alone may be proven
inadequate, independently of the method used, if a reliable estimate of the target earthquake
characteristics and foundation flexibility is not obtained in advance. Finally, a tendency is
observed that (as expected) stresses are more sensitive to the decisions made than
displacements, a fact that is important for the design process. In any case, ignoring spatial
variability of earthquake ground motion does not necessarily lead to conservative design since
the ratio of ρ can be higher than unity for a large number of scenarios studied. Moreover, by
plotting the SVEGM-effect ratio ρ with the overall length of each bridge (Sextos and Kappos,
2008) it is seen that the above findings are in agreement with the more conservative limits
introduced in Eurocode 8.
Based on the analyses results of the dynamic response of 27 (real and idealised) bridges under
synchronous and asynchronous excitation, an effort was made to quantify the relative
importance of the various parameters involved on the on the action effects (deck
displacements and bending moments) of the superstructure. Given the complexity of the
problem and the infinite combinations of bridge dynamic characteristics, seismotectonic
environment, and ground conditions, it is very difficult to derive general design or analysis
rules to account for spatial variation of ground motion that can be used without any exception.
Few thoughts on the numerical simulation of soil, structure and earthquake input 13

Figure 11: Impact factor of the various decisions made during the study of 27 bridges under SVEGM, SSI
and site effects. Top: deck displacements. Bottom: pier base bending moments (Sextos and Kappos, 2008).

Figure 12: Calculated rotational ductility demand μθ at the base of Bridge 08 piers according to the
‘classic’ and the ‘comprehensive’ approach (Sextos et al., 2002).
 14 A. Sextos – Earthquake Engineering by the Beach Workshop, July 2-4 2009, Capri, Italy.

Table 1: Configuration and analysis details of the bridges studied (Sextos & Kappos, 2008)

Structural Configuration 19 alternative

and analysis bridges

Bridge03 Bridge04 Bridge05 Bridge06 Bridge07 Bridges

Bridge ID # Bridge01 Bridge02 Bridge08
(Warth) (G11 Polymilos) (Krystallopigi) (Kavala) (Lissos) 09-27*
Location Idealized Idealized Austria Greece Greece Greece Greece Idealized Idealized
Spans 04 04 07 03 12 04 11 12 4, 6, 8
2x63.9 and 50m, 100m,
Span length 30.5m 36.6-73.1m 62-67m 44-54m 45m 26.5-44.3m 50m
118.4m 150m
200m, 400m,
Total length 152.4m 292.6m 459m 246.2m 638m 180m 433.3m 600m
Bridge 600m
geometry Pier-deck Monolithic / Monolithic / Monolithic / Monolithic / Monolithic /
Monolithic Monolithic Bearing type Monolithic
connection Bearing type Bearing type Bearing type Bearing type Bearing type
Minor curvature
Curvature In height No No In plan (r=488m) No No No No
in plan
Foundation Fixed Fixed Footing Caissons Pile Group Caissons Pile Group Pile Group
Loss of
Yes Yes Yes Yes Yes Yes Yes Yes Yes
coherence
Wave-
Spatial Yes Yes Yes Yes Yes Yes Yes Yes Yes
passage
variability
Site
parameters No No No No No No No Yes Yes
response
Kinemaric Kinemaric Kinemaric
SSI No No No No No No
/Inertial /Inertial /Inertial
Total cases analysed 8 8 8 8 16 7 4 6 207
TH** TH TH TH TH, NLTH TH TH
Type of analysis TH TH
RS** RS RS NLTH** RS NLTH NLTH
Transverse Transverse Transverse
6 different angles
Direction of Seismic input Transverse Transverse (one case Transverse Transverse Transverse (one case (one case
of incidence
longitudinal) longitudinal) longitudinal)
EC8
EC8 artificial and Athens and
Reference Point Loma Prieta Greek Code artificial,
EC8 artificial EC8 artificial EC8 artificial Kozani EC8 artificial Loma Prieta
Seismic Motion (Gilroy record) artificial Athens
earthquake earthquake
earthquake
Der Kiureghian, (Sextos et al.,
References (structural (Sextos and (Sokol and (Sextos and (Lekidis et al., (Kappos et (Sextos et (Sextos et al.,
& Neuenhofer 2004, Paraskeva
configuration) Kappos, 2008) Flesch, 2005) Kappos, 2008) 2004) al., 2006) al., 2003b) 2003b)
(1992) et al., 2002)
Notes: * Bridges 09-27 represent a stepwise modification of the reference 200m long Bridge09 in terms of stiffness reduction due to cracking (Bridges10-13), geometry, symmetry and
regularity (Bridges14-18), abutment-backfill interaction (Bridges19-21), angle and frequency of excitation (Bridges22-23) as well as overall length (400, 600m) and span length (50, 100,
150m), hence different fundamental period (Bridges24-27); see details in Sextos et al. (2003b). ** RS : Response Spectrum Analysis, TH : Linear Time History Analysis, NLTH : Nonlinear
Time History Analysis
Nevertheless, the results presented herein lead to the conclusion that ignoring spatial
variability of earthquake ground motion does not necessarily lead to conservative design since
its effect can be very often detrimental at least for overall bridge length that exceeds the
newly set (more conservative) limits introduced in Eurocode 8. The above observations were
also verified for the case of a long and straight circuit of the Byzantine Walls of the city of
Thessaloniki (Stylianidis and Sextos, 2008), a fact that additionally highlights the necessity to
study the particular problem for other kinds of extended structures as well (i.e. pipelines,
tunnels, dams). However, despite the significant research effort on the physical modelling of
the problem, it is deemed that the present state-of-the-art has to be enriched with additional
analytical and experimental results, focusing on the study of the response of a larger number
of real structures. This effort can be also supported by utilizing the growing number of
available recorded data within the ground and on the ground surface and, even more
importantly, on the foundation and the superstructure.

4 ACKNOWLEDGEMENTS

This work was been accomplished with the valuable contribution of various colleagues which
are not reported as co-authors herein due to the nature of the particular workshop but their
contribution is evident in the reference list below. Apart from that, the author would like to
thank Prof. A. Kappos and P. Potikas at Aristotle University Thessaloniki for kindly
providing all the data and results derived during the non-linear static assessment of the bridge
studied. It is also notable that most of the results reported herein have been obtained within
the framework of the Research Project ‘Seismic Protection of Bridges’ (ASProGe) funded by
the General Secreteriat of Research and Technology, of Greece (2004-2007) in collaboration
with the scientific responsible Professor Andreas Kappos. Acknowledgements are also due to
Prof. Amr Elnashai and Assist. Prof. Oh-Sung Kwon for their precious assistance regarding
the application of the analysis coordinator UI-Simcor, developed at the University of Illinois,
as well as the Earthquake Engineering Research Center (EERC) at University California
Berkeley and Prof. B. Stojadinovic for their overall support during the first author’s visit in
the framework of which part of this work was conducted. The contribution of Professor
Kosmas Stylianidis as regard to the investigation of the dynamic response of monumental
structures under asynchronous excitation is also equally appreciated.

5 REFERENCES

Burdette, N. and Elnashai, A.S. (2008) “The effect of asynchronous earthquake motion on complex
bridges. Part 1: Results and Implications on Assessment”, Journal of Bridge Engineering, 13(2):
166-172.
Burdette, N., Elnashai, A.S., Lupoi, A. and Sextos, A. (2008) “The effect of asynchronous
earthquake motion on complex bridges. Part 1: Methodology and Input Motion”, Journal
of Bridge Engineering, 13(2): 158-165.
CEN (2004) Eurocode 8: Design of Structures for Earthquake Resistance-Part 1: General
Rules, seismic action and rules for buildings, EN 1998-1:2004, European Committee of
Standardisation.
Deodatis G. (1996) “Simulation of ergodic multi-variate stochastic processes”, Journal of
Engineering Mechanics, 122(8): 778-787.
Der Kiureghian, A. and A. Neuenhofer (1992) “Response Spectrum method for multiple
support excitations”, Earthquake Engineering and Structural Dynamics, 21: 713-740.
16 A. Sextos – Earthquake Engineering by the Beach Workshop, July 2-4 2009, Capri, Italy.

Der Kiureghian, A. and Keshishian (1997) “Effects of incoherence, wave passage and
spatially varying site conditions on bridge response”, FHWA/NCEER Workshop on the
National Representation of Seismic Motion, Report 97-0010, NY, 393–407.
Dicleli, M. (2005) “Integral Abutment-Backfill Behavior on Sand Soil—Pushover Analysis
Approach”, Journal of Bridge Engineering, 10(3): 354-364.
Elnashai, A. S., Papanikolaou, V., and Lee, D. (2002) “Zeus NL – A System for Inelastic
Analysis of Structures”, Mid-America Earthquake Center, Univ. of Illinois at Urbana-
Champaign.
Filippou, F. C., and Constantinides, M. (2004) “FEDEASLab Getting Started Guide and
Simulation Examples”, Technical Report NEESgrid-2004-22: www.nees-grid.org
Goel, R.K., and Chopra, A. (1997) “Evaluation of bridge abutment capacity and stiffness
during earthquakes”, Earthquake Spectra, 13(1): 1-23.
Hao H. (1989) “Effects of spatial variation of ground motions on large multiply-supported
structures”, UBC/EERC-89/06, Berkeley: EERC, University of California.
Harichandran, R.S. and Vanmarcke, E. H. (1986) “Stochastic variation of earthquake ground
motion in space and time”, Journal of Engineering Mechanical Division, 112: 154-174.
Kappos, A., Potikas, P. and Sextos, A. (2007) “Seismic assessment of an overpass bridge
accounting for non-linear material and soil response and varying boundary conditions”,
Computational Methods in Structural Dynamics and Earthquake Engineering,
COMPDYN, Rethymnon, Greece, CD-ROM Volume.
Kappos, A.J., Moschonas, I.F., Paraskeva, Th. and Sextos, A.G. (2006) “A methodology for
derivation of seismic fragility curves for bridges with the aid of advanced analysis tools”,
13th ECEE, Geneva, Paper no. 275.
Kotsoglou, A. and Pantazopoulou, S. (2007) “Bridge–embankment interaction under
transverse ground excitation”, Earthquake Engineering and Structural Dynamics,
36:1719–1740.
Kwon, O. S. and Elnashai, A. S. (2008) “Seismic Analysis of Meloland Road Overcrossing
Using Multiplatform Simulation Software Including SSI”, Journal of Structural
Engineering, 134(4): 651-660.
Kwon, O.S., Sextos, A. and Elnashai, A. (2008) “Liquefaction-dependent fragility
relationships of complex bridge-foundation-soil systems”, International Conference on
Earthquake Engineering and Disaster Mitigation, Jakarta, Indonesia, 14-15 April.
Lekidis, V. , Karakostas, C., Christodoulou, K. , Karamanos, S., Papadimitriou, K. Panetsos,
P. (2004) “Investigation of dynamic response and model updating of instrumented R/C
bridges”, 13th World Conference on Earthquake Engineering, Vancouver, Canada, Paper
No. 2591.
Lou, L., and Zerva, A. (2005)” Effects of spatially variable ground motions on the seismic
response of a skewed, multi-span, RC highway bridge”, Soil Dynamics and Earthquake
Engineering, 25(7-10): 729-740.
Luco, J.E. and Wong, H.L. (1986) “Response of a rigid foundation to a spatially random
ground motion”, Earthquake Engineering and Structural Dynamics, 4: 891-908.
Lupoi, A., Franchin, P., Pinto, P. E., and Monti, G. (2005) “Seismic design of bridges
accounting for spatial variability of ground motion”, Earthquake Engineering and
Structural Dynamics, 34(4-5): 327-348.
Lupoi, A. (2009) “The Response of Isolated Bridges Accounting for Spatial Variability of
Ground Motion”, Journal of Earthquake Engineering, no. 13: 814-834.
Mackie, K., and Stojadinovic, B. (2002) “Bridge Abutment Model Sensitivity for
Probabilistic Seismic demand evaluation”, Proceedings of The 3rd National Seismic
Conference & Workshop on Bridges & Highways, April 28-May 1, Portland.
Few thoughts on the numerical simulation of soil, structure and earthquake input 17

McKenna, F., and Fenves, G. L. (2001) The OpenSees command language manual, version
1.2., Pacific Earthquake Engineering Research Center, Univ. of California at Berkeley.
Ministry of Environment, Physical Planning, and Public Works (1999) Circular 39/99:
Guidelines for the Seismic Design of Bridges, Athens (in Greek).
Ministry of Public Works (2000). Greek Seismic Code—EAK 2000, Athens (amended 2003,
in Greek).
Monti G., Nuti C. and Pinto P.E. (1996) “Nonlinear response of bridges under multi-support
excitation”, Journal of Structural Engineering, 122(10): 1147-1159.
Nuti, C. and Vanzi, I. (2005) “Influence of earthquake spatial variability on differential soil
displacements and SDF system response”, Earthquake Engineering and Structural
Dynamics, 34(11): 1353–1374.
Paraskeva, T., Kappos, A. and Sextos, A. (2002) “Extension of modal pushover analysis to
seismic assessment of bridges”, Earthquake Engineering and Structural Dynamics, 35:
1269–1293.
Pinto, A., Pegon, P. Magonette, and Tsionis, G. (2002) “Pseudo-dynamic testing of bridges
using non-linear substructuring”, Earthquake Engineering and Structural Dynamics,
33(11): 1125 – 1146.
Potikas, P. (2006) Seismic Design and Assessment of an Overpass Bridge, MSc Thesis,
Aristotle University Thessaloniki, Greece (in Greek).
Sextos, A., Kappos, A. and K. Pitilakis (2002) “Effect of analysis complexity on the
calculated ductility demand of RC bridges”, 12th European Conference on Earthquake
Engineering, London, U.K., 7-13 September, CD-RΟΜ Volume, paper 653, 2002.
Sextos, A., Pitilakis, K. and Kappos, A. (2003a) “Inelastic dynamic analysis of RC bridges
accounting for spatial variability of ground motion, site effects and soil-structure
interaction phenomena. Part 1: Methodology and analytical tools”, Earthquake
Engineering and Structural Dynamics, 32(4): 607-629.
Sextos A., Kappos A. and Pitilakis K. (2003b) Inelastic dynamic analysis of RC bridges
accounting for spatial variability of ground motion, site effects and soil-structure
interaction phenomena. Part 2: Parametric study”, Earthquake Engineering and Structural
Dynamics, 32(4): 629-652.
Sextos, A., Kappos, A. and Mergos P. (2004) “Effect of Soil-Structure Interaction and Spatial
Variability of Ground Motion on Irregular Bridges: The Case of the Krystallopigi Bridge”,
13th World Conference on Earthquake Engineering, Vancouver, Canada, Paper No. 2298.
Sextos, A. and Kappos, A. (2008) “Seismic response of bridges under asynchronous excitation
and comparison with EC8 design rules”, Bulletin of Earthquake Engineering Engineering,
7, 519-545.
Sextos, A., Mackie, K., Stojadinovic, B and Taskari, O. (2008) “Simplified P-y relationships
for modelling embankment-abutment systems of typical California bridges”, 14th Word
Conference on Earthquake Engineering, Beijing, China, CD-ROM Volume.
Shinozuka M, Saxena V and Deodatis, G. (2000) “Effect of Spatial Variation of Ground
Motion on Highway Structures”, MCEER-00-0013, MCEER, NY.
Sokol, M. and Flesch, R. (2005) “Assessment of Soil Stiffness Properties by Dynamic Tests
on Bridges”, Journal of Bridge Engineering, 10(1): 77-86.
Spencer Jr., B. F., Elnashai, A. S., Park, K., and Kwon, O. (2006) “Hybrid Test Using UI-
SimCor, Three-Site Experiment”, Final report to NEESit for Phase I project of hybrid
simulation framework development, University of Illinois at Urbana-Champain.
Stylianidis, K. and Sextos, A. (2009) “Back analysis of Thessaloniki Byzantine Land Walls as
a means to assess its Seismic History”, International Journal of Architectural Heritage,
3(4): 1-23.
18 A. Sextos – Earthquake Engineering by the Beach Workshop, July 2-4 2009, Capri, Italy.

Tzanetos N., Elnashai A. S., Hamdan F. H., Antoniou S. (2000) “Inelastic Dynamic Response
of RC Bridges Subjected to Spatial Non-Synchronous Earthquake Motion”, Advances in
Structural Engineering, 3(3): 191-214.
Zerva, A. (1990) “Response of multi-span beams to spatially incoherent seismic ground
motions”, Earthquake Engineering and Structural Dynamics, 19(6): 819-832.
Zerva, A. (2009) Spatial Variation of Seismic Ground Motions: Modeling and Engineering
Applications (Advances in Engineering Series). Ed. Haym Beranoya. CRC Press, Taylor
& Francis Group, FL.
Zhang, J., and Makris, N. (2002) “Kinematic response functions and dynamic stiffness of
bridge embankments”, Earthquake Engineering and Structural Dynamics, 31: 1933–1966.