Seismic assessment of a major bridge using pushover analysis

P. Mergos1, A. G. Sextos1, A. J. Kappos1

Summary
This study focuses on the earthquake-resistant design and assessment of the Krystallopigi bridge,
which is currently under construction as part of the EGNATIA highway in northern Greece. This long
and curved, twelve span bridge structure is designed according to current seismic codes and then
assessed for motions up to twice the design earthquake intensity. The behavior is found to be
satisfactory, yet dependent to a significant degree on geometry, earthquake and modeling assumptions.

Introduction
Although elastic analysis provides an overview of the expected dynamic response of a bridge, it is
clear that it cannot predict the failure mechanisms or the redistribution of forces that follow the plastic
hinge development and the potential progressive collapse of the structure. Non-linear pushover analysis
on the other hand, is a widely used assessment tool that allows for the evaluation of the structural
behavior in the inelastic range and the identification of the failure mechanisms, while it highlights the
critical points of structural weaknesses. Although a substantial amount of work has been done on
pushover analysis of buildings, corresponding work on bridges has been much more limited; the main
reason for this should be the fact that fundamental mode analysis (a key characteristic of the standard
pushover method) is often not appropriate for describing the behavior of bridges [1].
The present study focuses on the assessment of the expected non-linear behavior of bridges
designed according to modern seismic code provisions. An effort was made to (a) identify the actual
dissipation mechanisms of a real structure with respect to the behavior factor assumed (b) investigate
whether modern code capacity design concepts ensure failure hierarchy and prevent structural collapse
(c) focus on the effect of bridge irregularity (i.e. curvature) on the structural response in the inelastic
range (d) investigate the role played by the soil-foundation-superstructure interaction in terms of action
effects and (e) study the sensitivity of the aforementioned issues on the modeling assumptions made
during the analysis stage. Along these lines, the long, curved and irregular Krystallopigi bridge is
selected as the focus of the present study.

Overview of the bridge studied

The Krystallopigi bridge is a twelve span structure of 638m total length (Fig. 1) that crosses a
valley, as a part of the 680 km EGNATIA highway in northern Greece. The curvature radius is equal to
488m while its deck width is 13m. The slope and the pier height vary along the length. According to the
initial design, the deck is a prestressed at its top flange concrete box girder section; concrete grade is
B45 (characteristic cylinder strength fck=35 MPa) and prestressing steel grade 1570/1770 (fy=1570
MPa). Piers are in reinforced concrete, concrete grade is B35 (fck=27.5MPa), steel grade st500s
(fy=500 MPa). For abutments and foundations B25 (fck=20 MPa) and st500s are used.
The structure is supported on piers (M1-M11 in Fig. 1) of height that varies between 11 and 27m,
For the end piers 1, 2, 3, 9, 10, 11 a bearing type pier-to-deck connection is adopted (see

1
Dept of Civil Engineering, Aristotle University Thessaloniki, 54124 Greece
Fig. 2), while the interior piers are monolithically connected to the deck. It is noted that for practical
reasons (i.e. anchorage of the prestressing cables) the initial 0.50x0.20m pier section is widened to
0.70x0.20m at the pier top range. The piers are supported on groups of piles of length and configuration
that differs between support points due to the change of the soil profile along the bridge axis.
Finite element analysis was used for the assessment of the non-linear response of the bridge,
involving the discretisation of the structure in 220 non-prismatic 3D beam elements (Fig. 3). For the
piers connected to the deck through bearings, the movement along the longitudinal axis as well as the
rotation around both the longitudinal and transverse axis is unrestrained. On the contrary, the existence
of shear keys results to the prevention of transverse displacements and the movement and rotation along
and about the vertical axis. Apart from the model described above (Model A), alternative models with
cracked concrete sections (Model B) and flexible foundations (Model C) were also studied. In Model B
the effective pier stiffness eff was calculated from the initial slope of the moment-curvature
relationship of the section at the location of the plastic hinge, as prescribed by Eurocode 8 [2] and the
Greek Code [3] for bridge design. The stiffness properties of the coupled soil-foundation pier, on the
other hand, were calculated using the computer code ASNG [4]. The structure is then designed in shear
and bending at all critical regions using a commercial software package.
A full design of the actual bridge (for usual, as well as seismic, actions) was carried out by the
Greek consultancy firm DENCO (Athens). In the present study, the seismic design of the bridge was
repeated using an independent model and software, but still in conformity with the current Greek codes
[3, 5]. First, the structure was analyzed using multi-modal response spectrum analysis with the widely
used F.E. program SAP2000 [6]. Then, the piers were designed in bending using a greek commercial
program (EDOS) for the analysis of arbitrarily shaped reinforced-concrete cross sections in biaxial
bending; typical interaction curves (flexural strength vs. axial load) are shown in Figures 4 and 5.
Finally, the hoop and cross-tie reinforcement was evaluated according to the bridge code [3] for
adequate shear resistance, confinement and prevention of buckling of the longitudinal bars.
The seismic behavior of the designed bridge was then assessed using non-linear static (pushover)
analysis. For the definition of target displacements of the structure the response spectrum of [5] was
used. Soil conditions were taken to correspond to category B of the Greek seismic code, which can be
considered equivalent to subsoil class B of Eurocode 8. For the considered Zone III of the Greek Code
a peak ground acceleration of 0.24g is specified, while a behavior factor of 3.0 was adopted. The
inelastic behavior of the critical cross-sections of the piers was evaluated using the program RCCOLA-
90 [7]. Then, parametric analyses were carried out in order to estimate the influence of various
parameters of the static non-linear analysis to the results of the seismic assessment. The structure was
assessed in both directions using the model that was considered as the most adequate.

Assessment of the effect of analysis assumptions on final design

The results of the present study, described in more detail elsewhere [8], show the influence of
several assumptions on the seismic design of the bridge. The assumptions made regarding stiffness
reduction due to cracking, as studied in Model B, are a typical example in this respect. Considering an
effective pier stiffness based on the initial slope of the moment-curvature relationship resulted in 30 to
50% of the uncracked pier section stiffness, and to a substantial increase in the fundamental period of
the structure (i.e. from 1.54sec to 2.12sec in the longitudinal, and from 1.02sec to 1.39sec in the
transverse direction), which in turn, resulted to an increase by about 50% in the calculated deck
displacements in both directions. On the other hand, the calculated flexural reinforcement requirements
at the locations of potential plastic hinge development were reduced by up to 30% when cracked pier
stiffnesses were considered.
 Figure 1: Overview of the Krystallopigi Bridge

Figure 2: Non-monolithic pier-to-deck Figure 3: Layout of the 3D Finite

connection (end piers) element model

-300000
-250000 =45
(kN)

-200000
Force

-150000

-100000
Axial

a =45 b
[kN]

-50000
0
50000 Moment
(kNm)
100000 [kNm]
a 150000
0 40000 80000 120000 160000 200000 240000
b

Figure 4: M-N interaction curves (in the Figure 5: M-N interaction curves (in 2D
3D space) for the hollow piers space) along different axes of the pier
 250

Fixed
Flexible

200
200

[mm]

150
Displacement

150
(mm)

100
100

50
Lateral

50

0
0

1 1 2 3 4 5 6 7 8 9 10 11 2
Pier
Pier

Figure 6: Effect of pier support pier support conditions on the lateral displacement

0,35

0,3
0.30
Design earthquake

0,25
0.25

Collapse earthquake

V/W

0,2
0.20
V/W

0,15
0.15

0,1
0.10
Design shear

0.05
0,05
0
0
0,00 0,01
0.01 0,02
0.02 0,03
0.03
/
/

Figure 7: V- curve in the transverse direction

The effect of accounting for the flexibility of the foundation soil (Model C) was found to be rather
minor for the bridge studied; the period of the structure was just slightly increased and transverse
displacements (from modal analysis) were about 10% larger than in the fixed base model (Figure 6),
while a 15% reduction in bending moments at the pier base was also observed. This conclusion can not
be generalized, since only the (static) stiffness of the soil-foundation-pier system was investigated, and
foundation flexibility was dominated by the overall high stiffness of the soil conditions in the site of the
bridge, leading to a response that does not significantly diverge from the assumption of a completely
fixed pier system.
In addition to pier cracking and foundation compliance, the assumption regarding the overall
structural damping is also crucial for design based on modal analysis (as requested by the codes). For
bridges with a prestressed deck (a very common case for long-span structures), the adoption of a 5%
damping for the entire structure does not account for the reduced ability of prestressed concrete to
absorb energy, and may lead to erroneous reduction of seismic forces by about 15%. This was the
reason that for all analyses 3.5% of critical damping was assumed, which is the average of 2%
(prestressed concrete) and 5% (reinforced concrete).

Assessment of seismic performance using pushover analysis

Unlike the case of buildings, wherein the pushover curve is generally defined in terms of base
shear vs. top displacement (in the direction under consideration), in bridges the variability in the
pushover curve is significant, depending on which pier it is referred to (particularly when piers are of
unequal height, as in the bridge studied). A typical pushover curve, calculated by applying the modal
load pattern of the 1st mode in the transverse direction of the bridge is shown in Fig. 7, referring to the
central pier M6 (see Fig. 1); a similar curve was derived for the longitudinal direction. It is seen that the
overall response of the bridge is indeed very satisfactory since neither local nor global failure are
predicted even under forces that far exceed the design level (target displacement corresponding to
Collapse Earthquake in Fig. 7). The sequence of plastic hinge formation (along with the force (V/W) -
displacement (/) curves), is also presented in Figure 8 for both the longitudinal and transverse
direction of the bridge.
It was also deemed interesting to identify the potential effect of the assumption made with respect
to the pattern adopted for the loading of the structure; this aspect is still an open issue in the case of
pushover analysis of bridges. For the analyses presented above, the assessment was based on a loading
model based on the first and second mode of vibration for both the longitudinal and transverse
direction, resulting to a force at each node that is the product of the modal displacement, the square of
the circular frequency of the particular mode and the corresponding nodal concentrated mass, and acts
in the direction of the modal displacement. For comparison purposes, the alternative model of uniform
loading in each one of the three directions of the generalised coordinate system was also adopted; this is
usually required in pushover analysis of buildings, with a view to identifying critical combinations of
shear and flexure (V/M). According to this model, the force acting at each node is proportional to the
nodal mass in the direction of the control displacement.
The results indicate that the adoption of a particular loading model for the excitation of the
structure plays indeed an important role with respect to the inelastic response of the bridge. In
particular, when the uniform loading pattern is applied, the overall strength of the system appears to be
higher (i.e. yielding occurs at a higher level of base shear). This trend is illustrated in Figure 9 and it is
more clear in the transverse direction. The more pronounced effect in the transverse direction should be
attributed to the fact that due the shape of the first transverse mode, the largest displacement
corresponds to the middle pier (M6), hence the modal force is higher at that particular pier compared to
the rest, whereas in the uniform pattern, forces at all piers are about the same (since masses are similar).
As a result, for the same target displacement, higher forces are developed in the latter case. It is
interesting to note that a similar overestimation of strength is found in pushover analysis of buildings,
but in this case the key reason is the distribution of the lateral loading along the height of the structure
(which results in higher overturning moment at the base in the case of modal loading).
Due to the relatively tall piers of the structure, studying the effect of geometric non-linearity (P-
effects) was also of particular interest. The observation was that ignoring such effects, leads to
overestimating the strength at failure by 15%, although it was apparent that P- effects did not dominate
the overall response, an observation that is in agreement with other studies [9].
For the Krystallopigi bridge a high displacement ductility was found in both principal directions
(i.e. 6.1 in the longitudinal and 4.6 in the transverse direction) due to the significant available curvature
ductility at the critical locations of potential plastic hinging; these local ductilities resulted from the
application of provisions that are related to the prevention of longitudinal bar buckling and the capacity
design against shear of the middle piers [3]. Moreover, a significant overstrength of the bridge was
found (compare design base shear to yield force in Fig. 7) due not only to the partial safety factors
applied during design, but also to the use of a minimum amount of reinforcement even when the
strength demand is less (min code requirements), and to force redistribution after yielding. It has to be
noted that the observed overstrength was higher in the transverse direction, since the longitudinal
direction was the critical one for the design of pier reinforcement. Similar conclusions regarding the
good inelastic response were drawn for the Greveniotikos bridge, which is also a long ravine bridge,
part of the EGNATIA highway [10].
Finally, regarding bridge curvature (in plan), the satisfactory overall seismic response of the
Krystallopigi bridge does not justify the provision of the Greek Bridge Code [3], which considers
bridge curvature as a structural irregularity and recommends quasi-elastic design (behavior factor
q=1.0). Nevertheless, this observation cannot be generalized without extensive parametric (pushover
and time history) analyses for an appropriate number of curved bridge structures.
Conclusions
The seismic design and assessment of the Krystallopigi bridge through non-linear static analysis
indicates that the application of modern seismic codes resulted in a bridge able to resist earthquakes that
exceeded the design level even by a factor of two. Displacement ductility factors of about 5 or more
were estimated for both directions of the bridge and calculated overstrength varied from 35 to 70%.
However, a number of choices regarding the application of particular provisions (pier stiffness,
foundation compliance, damping ratio) clearly involve engineering judgment and hint to the need for
better clarifying these issues within the codes. As a result, it appears appropriate to verify the above
observations through well defined parametric analysis schemes and to investigate further aspects of the
problem such as dynamic soil-structure interaction, the effect of multi-support excitation, the
importance of curvature in plan and height, the identification of a critical axis for the assessment of the
seismic response of the structure, and finally the consistency between assessments based on non-linear
static and non-linear dynamic analysis; the writers are currently working in this direction.

Acknowledgement
The writers would like to thank EGNATIA ODOS S.A. for providing all available data and
information with respect to the design of the Krystallopigi bridge.
 0.35

0.30

0.25

0.20
V/W

0.15

0.10
Plastic hinge development
0.05 Transverse direction
Longitudinal direction
0.00
0.00 0.01 0.02 0.03 0.04 0.05
/
Figure 8: Shear Force-Displacement combinations corresponding
to the development of plastic hinges in both directions

0.35
Transverse direction
0.30

0.25

0.20
V/W

Longitudinal direction
0.15

0.10

0.05 Modal loading pattern

Uniform loading pattern
0.00
0.00 0.01 0.02 0.03 0.04 0.05
/

Figure 9: Effect of loading pattern on the pier stiffness and strength

 References
1 Fischinger, M. and Isakovic, T. (2002): Applicability of the pushover based procedures for
bridges, Proceedings of the Third National Seismic Conference and Workshop on Bridges and
Highways, Portland, Oregon, 2002, pp. 335-344.
2 CEN. Eurocode 8 (1994): Design provisions for earthquake resistance of structures. Part 2:
Bridges, ENV 1998-2, CEN, Brussels.
3 Ministry of Public Works (1999): Guidelines for Seismic Design of Bridges, E39/99, Athens.
4 Sextos, A., Pitilakis, K. and Kappos, A. (2003): A global approach for dealing with spatial
variability, site effects and soil-structure-interaction for non-linear bridges: a. verification study,
Earthquake Engineering and Structural Dynamics, Vol. 4, pp. 607-629.
5 Ministry of Public Works (2000): Greek Seismic Code (E.A.K. 2000), Athens.
6 Computers and Structures Inc. (1999): SAP2000: Three dimensional static and dynamic finite
element analysis and design of structures, Berkeley, California.
7 Kappos AJ. (1993): RCCOLA-90: A Microcomputer Program for the Analysis of the inelastic
Response of Reinforced Concrete Sections, Dept. of Civil Engineering, Aristotle University of
Thessaloniki, Greece.
8 Mergos, P. (2002): Earthquake resistant design and assessment of a curved in plan 12-span ravine
bridge, MSc Thesis, Aristotle University of Thessaloniki, Greece (in Greek).
9 Kappos, A.J. and Sextos, A. (2001): Effect of foundation compliance on the lateral load response
of R/C bridges, Journal of Bridge Engineering, Vol. 6 (2), pp. 120-130.
10 Abeysinghe, R. et al. (2002): Pushover analysis of inelastic seismic behavior of Greveniotikos
Bridge, Journal of Bridge Engineering, Vol. 7 (2), pp. 117-126.