Fukae Collapse

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

HOSTED BY Available online at www.sciencedirect.

com

ScienceDirect
Soils and Foundations 60 (2020) 1450–1467
www.elsevier.com/locate/sandf

Technical Paper

Fukae bridge collapse (Kobe 1995) revisited: New insights


L. Sakellariadis a, I. Anastasopoulos a,⇑, G. Gazetas b
a
Dept. of Civil, Environmental and Geomatic Engineering, ETH Zürich, Switzerland
b
National Technical University of Athens, Greece

Received 17 January 2020; received in revised form 29 July 2020; accepted 19 September 2020
Available online 6 November 2020

Abstract

The paper revisits the notorious collapse of 18 spans of the elevated Route No. 3 of Hanshin Expressway during the 1995 Kobe earth-
quake. The overturned deck was monolithically connected to 3.1 m diameter piers, which failed dramatically. In stark contrast, the mas-
sive 17–pile groups survived the earthquake and are still in use, supporting the new bridge. The scope of the study is dual. Initially, the
actual pile group–bridge–soil system is investigated employing the finite element (FE) method, accounting for material and geometric
nonlinearities. Reinforced concrete (RC) members (pier and piles) are modelled using the Concrete Damaged Plasticity (CDP) model,
while a kinematic hardening model is employed for the soil. The numerical simulation reproduces the observed shear-dominated failure
at the region of reinforcement cut-off. The analysis reveals that the pile group sustains non-negligible rocking during shaking (despite
being over-designed), leading to alternating tension and compression of the edge piles, combined with shear-moment loading. The result-
ing stiffness reduction of the cracked under tension piles leads to load redistribution towards the stiffer compressed piles, preventing plas-
tic hinging of the weaker piles (under tension). These findings are consistent with post-earthquake in-situ testing, qualitatively verifying
the numerical analysis technique. Subsequently, alternative foundation concepts are explored, starting with the pilecap of the original
configuration acting as a shallow footing, provoking nonlinear rocking response. It is shown that soil yielding acts as a ‘‘fuse”, preventing
collapse at the cost of increased settlements and limited residual foundation rotation. The benefits and limitations of such a shift towards
nonlinear soil–foundation response are further explored, studying four intermediate foundation schemes.
Ó 2020 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Kobe earthquake; Bridge collapse; Nonlinear analysis; Soil-structure interaction

1. Introduction reclaimed land. As a result, almost every type of structure


was subjected to extreme seismic shaking during the 1995
The devastating Mw 7 Great Hanshin 1995 earthquake Kobe earthquake, substantially exceeding the design con-
in Kobe, Japan, is considered an event of enormous signif- siderations of the time. The lessons learned from the Kobe
icance in the history of earthquake engineering. Located in earthquake influenced substantially the seismic practices
a very narrow stretch of land (a mere kilometer wide, but and codes not only in Japan but also worldwide. Further-
almost 30 km long) Kobe is characterized by very dense more, the reconstruction and retrofit methodologies and
infrastructure (Fig. 1). Most of its expressways are elevated techniques that were developed and applied in Kobe
on bridges, and a substantial part of the city is built on pushed forward the relevant state-of-the-art.
Despite being well-engineered (using the best methods of
analysis and construction technologies available at the
Peer review under responsibility of The Japanese Geotechnical Society. time, mostly 70–800 s), when subjected to accelerations
⇑ Corresponding author. much higher than their design levels, many structures were
E-mail addresses: lampross@ethz.ch (L. Sakellariadis), ixa@ethz.ch pushed well into their nonlinear regime and failure proved
(I. Anastasopoulos), gazetas@central.ntua.gr (G. Gazetas).

https://doi.org/10.1016/j.sandf.2020.09.005
0038-0806/Ó 2020 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Fig. 1. City of Kobe, showing the Hansin expressway Route 3 extending almost parallel to the fault projection, and the location of the collapsed bridge
piers at the area of Fukae; post-seismic photo of the collapsed bridge and exhibit of plastic hinging of one of the failed piers in the city seaside.

to be often unavoidable. This study revisits one of the most The collapse of the Fukae bridge is re-examined herein
dramatic failures of the 1995 Kobe earthquake: the col- by comparatively assessing the performance of the original
lapse of 18 spans of the elevated Hanshin Expressway foundation, which survived the earthquake and is still in
Route No. 3 in the Fukae area (Fig. 1). The deck was use (for the fully replaced bridge), to that of alternative
monolithically connected to 3.1 m diameter circular piers, design concepts, considering nonlinear soil-foundation
founded on 17–pile groups in alluvium sand and gravel. interaction (e.g., Anastasopoulos et al., 2010a, 2010b). A
The collapse has been attributed to inadequate structural single segment of the collapsed bridge is modelled employ-
design regarding mainly the addition of a prematurely ter- ing the finite element (FE) method (Figs. 3 and 4), adopting
minated third row of longitudinal reinforcement and insuf- the geometry and material properties (mass M, moment of
ficient shear capacity due to poor transverse reinforcement inertia I, pier diameter D, pile diameter d, pile length L) of
(Kawashima & Unjoh, 1997). the actual system (Kawashima & Unjoh (1997)). Emphasis
Surprisingly, the failure did not occur at the bottom of is given on the nonlinear response of each component of
the pier (location of maximum bending moment) but the entire pile-group–bridge–soil system and their
2.5 m above the pilecap, where shear cracking initiated. interactions.
In this critical section, the third row of longitudinal rein- At first, the results of the numerical simulation
forcement was terminated (Fig. 2a), mobilizing a brittle regarding the behavior of the actual system are verified
shear mode of failure instead of the desired flexural mech- against post-earthquake observations and results of pre-
anism which would have added ductility to the system. The vious relevant studies. The study then focuses on the
concept of ‘‘capacity design” was not known at that time; foundation level, investigating the ‘‘successful” geotech-
the collapse is largely due to its absence. Moreover, the role nical performance of the original pile group. The bene-
of soil–foundation–structure interaction (SFSI), which is fits of such foundation design philosophy are
still often ignored in design or considered a priori benefi- questioned through comparative assessment of the origi-
cial, was proven detrimental. Along with important site nal foundation against five foundation alternatives. Even
effects, previous studies (Gazetas et al., 2006; Mylonakis 25 years after the Kobe earthquake, the well docu-
et al., 2006) have shown that the elongation of the natural mented Fukae bridge case history not only offers a
period of the bridge due to SSI (T SSi > T fixed ) resulted in unique opportunity for validation of modern numerical
higher spectral demands S A (Fig. 2b). Nevertheless, the tools against field observations, but also offers an excel-
exact role of the piled foundation on the response and lent basis for the exploration of alternative foundation
the failure of the bridge has not yet been fully understood concepts aiming to enhance the resilience of critical
or quantified. infrastructure.

Fig. 2. (a) Reinforcement details and critical section where the 3rd row of longitudinal rebars was terminated and shear cracking initiated; lower section of
a collapsed pier exposed in the Hanshin Expressway earthquake museum; and (b) increase of spectral demand due to SSI-induced period elongation.
1451
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Fig. 3. Overview and geometric details of the examined segment of Fukae bridge.

Fig. 4. Key attributes of the FE model of a single segment of the Fukae bridge in Abaqus.

2. Finite element analysis methodology the lateral edges (multipoint kinematic constraints applied
at each level) that allow the model to mimic free-field soil
The static and dynamic response of a single segment of response under shear loading.
the Fukae bridge is analyzed employing the FE method
using ABAQUS. In total, six different foundation configu-
2.1. Soil modeling: kinematic hardening model
rations are explored, starting with the actual very stiff 17–
pile foundation with its large cap and a highly nonlinear
In contrast to the detailed reports of the structural
rocking shallow footing alternative consisting only of the
detailing of the Fukae bridge, the geotechnical data of
pilecap. Then based on the outcome of this investigation
the foundation are very limited. To that end, a simplified,
four intermediate approaches are further explored.
yet realistic, soil profile described by the equivalent
The key attributes of the FE model of the examined
undrained shear strength S u and shear wave velocity Vs is
bridge segment are shown in Fig. 4. The lateral boundaries
outlined based on the limited available SPT measurements
of the model are placed at distance 10d (d: pile diameter)
N 80 (Iwasaki et al., 1995) (Fig. 5a). The selected parameters
from the edge piles, while the vertical boundaries are placed
are based also on previous case studies of the Fukae bridge
at distance of 10d below the pile tip to reduce undesired
(Gazetas et al., 2006, Mylonakis et al., 2006). The 6-layered
boundary effects, while maintaining computational effi-
soil profile (Fig. 5a) is modelled with hexahedral (8-node)
ciency. A key feature is the tensionless interfaces with
elements, adopting a thoroughly validated kinematic hard-
appropriate friction coefficients l, which are introduced
ening model, with a modified pressure-dependent Von
between the soil and the foundation (piles and pile-cap)
Mises failure criterion and associated plastic flow rule
allowing for sliding and detachment, properly accounting
(Anastasopoulos et al., 2011). The evolution of stresses is
for such geometric nonlinearities. The 17 piles are rigidly
defined as:
connected to the pilecap to ensure fixed head conditions.
Furthermore, appropriate periodic boundaries are used at r ¼ ro þ a ð1Þ
1452
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Fig. 5. Soil modelling: (a) simplified soil profile (S u ; V s Þ, based on in-situ SPT, N 80 , measurements; and (b) Simulation of simple shear tests for cmax ¼ 0:002
and cmax ¼ 0:01 and comparison of numerical analysis (dots) to the G  c curves of Vucetic & Dobry (1993).

where ro corresponds to the stress at zero plastic strain, features of the selected constitutive model are shown
and a is the ‘‘backstress’’. The latter controls the kinematic Fig. 5b. Numerical results of shear stress, s – shear strain,
evolution of the yield surface in the stress space through a c loops are shown indicatively for two levels of shear strain
function F : (cmax ¼ 0:002 and cmax ¼ 0:01). The computed values of
F ¼ f ð r  aÞ  r o ð2Þ secant stiffness (dots) are compared to the widely used
G=Go  c (stiffness degradation) curves of Vucetic and
The plastic flow rate e_ pl follows an associated flow rule: Dobry (1991). More details on the constitutive model can
be found in Anastasopoulos et al. (2011).
_ @F

e_ pl ¼ e pl ð3Þ The adopted constitutive model has been validated
@r
against a plethora of physical model tests (centrifuge, 1 g,
_ and large-scale) referring to: (a) surface and embedded
where: e pl is the equivalent plastic strain rate.
The evolution law of the model consists of a nonlinear foundations subjected to cyclic loading and seismic shaking
kinematic hardening component, which describes the trans- (Anastasopoulos et al., 2011, 2012); (b) piles subjected to
lation of the yield surface in the stress space, and an isotro- cyclic loading (Giannakos et al., 2012); (c) bar-mat retain-
pic hardening component, which defines the size of the ing walls subjected to seismic shaking (Anastasopoulos
yield surface as a function of plastic deformation et al., 2010a, 2010b); and (d) circular tunnels subjected to
(Gerolymos & Gazetas, 2005). The evolution of the kine- seismic shaking (Bilotta et al., 2014; Tsinidis et al., 2014).
matic component of the yield stress is described as follows:
1 2.2. Modelling of structural members: Concrete damaged
a_ ¼ C ðr  aÞ_epl  ca_epl ð4Þ plasticity
ro
where: C is the initial kinematic hardening modulus To gain deeper insights on the collapse mechanism, crit-
(C ¼ ry =ey ¼ E ¼ 2ð1 þ vÞGo) and c a parameter that ical reinforced concrete (RC) structural members (pier and
determines the rate at which the kinematic hardening piles) are modelled in a fairly rigorous way. The unconfined
decreases with increasing plastic deformation. cover and the confined core of RC sections are simulated
The calibration of model parameters requires knowledge with nonlinear solid elements, employing the Concrete
of: (a) the undrained shear strength S u ; (b) the small-strain Damaged Plasticity (CDP) model, which is available in
stiffness (expressed through Go or V s ); and (c) the stiffness ABAQUS. The latter is capable of modelling the inelastic
degradation(G  c and n  c curves). The nonlinear quasi-brittle response of concrete using the concepts of
1453
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Fig. 6. Modelling of structural members: (a) stress-stain response of unconfined concrete cover and confined concrete core (left), and of the steel
reinforcement bars (right); and (b) compression and tension damage indices in function of the corresponding inelastic axial strain.

isotropic damaged elasticity combined with isotropic ten- which are functions of the corresponding inelastic strains
sile and compressive plasticity. ein (Fig. 6b):
The model is a continuum, plasticity-based, damage Eun ¼ ð1  d c ÞEc;o ð5Þ
model for concrete. Designed for monotonic, cyclic, and
dynamic loading, it considers two failure mechanisms: ten- where: Ec;o ¼ 28GPa the initial stiffness of concrete.
sile cracking and compressive crushing. The evolution of The longitudinal reinforcement is modelled by means of
the yield surface is controlled by two hardening parameters rebars (1D rods), defined as ‘‘equivalent” surface elements
related to each mechanism. Under uniaxial tension, the embedded in the concrete elements. A simple elastic-
stress–strain response follows a linear elastic relation until perfectly plastic stress strain (rs -es ) relationship (Fig. 6a)
reaching the failure stress. The subsequent formation of for steel S350 (rf ;s ¼ 350MPa; Es ¼ 210GPaÞ is used until
micro-cracks is macroscopically introduced by a softening the axial strain reaches es;u ¼ 12%. After this point, the steel
stress–strain response. Similarly, under compression the capacity drops rapidly to almost zero, corresponding to the
response is linear until initial yield; in the plastic regime initiation of buckling. The rebars are superposed on the
it is characterized by stress hardening followed by strain mesh of the concrete elements, which implies that perfect
softening beyond the ultimate stress. bonding is assumed between the reinforcement and con-
Both the uniaxial tension and compression stress–strain crete. Thus, the bahaviour of concrete is considered inde-
(rc -ec ) relationships of concrete are defined by the user as a pendently of that of the rebar. Hence, slip effects or loss
tabular function, adopting commonly used relations for of confinement cannot be captured. This limitation cer-
confined core concrete under compression (Mander et al. tainly affects post-failure behavior.
1988) and tension (Chang & Mander, 1994) and the uncon- The modelling methodology has been validated in
fined cover concrete (Chang & Mander, 1994) considering Agalianos et al. (2020) against two half-scale model tests
concrete C35. This representation indicatively shown in conducted by Prakash et al. (2010), The tests examined
Fig. 6a for the upper section of the collapsed pier. It can the response of circular RC columns subjected to axial
be seen that the compressive strength rf ;cc of the concrete force with low to moderate shear. Two column height-to-
core is higher than the strength rf ;uc of the concrete cover diameter ratios, H =D ¼ 6 and H =D ¼ 3, were considered
thanks to the confinement offered by the transverse rein- corresponding to low and moderate shear, respectively.
forcement. Furthermore, the ductility of the confined con- Both specimens had a diameter D ¼ 0.61 m, 2.1% longitu-
crete is also increased. With respect to the unloading dinal reinforcement ratio and constant axial load ratio
response, the elastic stiffness appears to be degraded. This N=N ult = 0.07. A comparison of the FE model to the exper-
degradation is different between tension and compression, imental results is shown in Fig. 7a in terms of F  d
becoming more pronounced with the increase of plastic response. The model captures reasonably well the experi-
strain. The degraded response (Fig. 6a) is characterized mental results both in terms of stiffness and ultimate
by two independent uniaxial damage indices, d t and d c , capacity.
1454
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Fig. 7. CDP modelling of RC piers and piles: (a) validation against 1:2 scaled experiments of RC columns (Agalianos et al. 2020); (b) verification against
RC section analysis (Response 2000) in terms of axial force–bending moment interaction for the lower and upper section of the collapsed pier; and (c)
shear force–bending moment interaction diagram of the upper section (2 rebar rows) compared to ASSHTO-99 guidelines, and failure modes with
superimposed contours of concrete damage.

The CDP modelling methodology is further verified ysis using the CDP model is in good agreement with the
against RC section analysis using Response 2000 (Bentz, RC section analysis for both the upper and the lower sec-
2000). The simulation is conducted adopting the material tion of the collapsed pier. Fig. 7c illustrates the compated
parameters of the actual bridge, as reported by (FE analysis) response of the upper section of the collapsed
Kawashima & Unjoh (1997). Special attention was given pier (two rebar rows) for different moment–shear ratios (by
to the reinforcement details of the collapsed pier, using dif- altering the height of the column). Good agreement is
ferent reinforcement ratio for the lower section of the pier shown with the results of RC section analysis (Response
(up to 2.5 m above the pilecap) and the upper section, cor- 2000) and the ASSHTO-99 guidelines, especially up to rea-
responding to the previously discussed 2 and 3 rows of lon- sonable height-to-diameter ratios (H  6 m). Further
gitudinal bars, respectively. Where not available (e.g. for reduction of the height of the pier leads to overestimation
the piles), reasonable assumptions were made regarding of the shear resistance.
the concrete type and the reinforcement ratio (typical min- The contours of concrete damage d c presented in
imum 1% ratio). Fig. 7c, indicate that the numerical simulation is capable
Fig. 7b and c offer an overview of the FE simulation in of reproducing the different modes of failure (flexural and
terms of axial force–bending, moment (N  M) and shear shear). These results are also assessed against post-
force–bending moment (F  M) interaction diagrams. earthquake observations and previous studies. Adachi
With respect to the flexural response (Fig. 7b), the FE anal- et al. (2003) reported the dramatically different response
1455
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

of two adjacent, almost identical piers, situated near the Winkler spring model that assumed different Moment–cur-
Fukae section. The first pier failed in shear, at the region vature ðM  cÞ loops for well and poorly reinforced con-
of longitudinal reinforcement cut-off, causing the collapse crete. In continuum–based approaches, the structural
of the supported span. On the contrary, the second pier behavior of the piles is considered in a similar simplified
experienced damage in the form of flexural micro- manner using nonlinear beam elements to simulate the
cracking at its bottom, surviving the earthquake. The only Moment–curvature ðM  cÞ response obtained from RC
difference between the two piers was the additional (3rd) section analysis. As depicted in Fig. 8a the beam elements
reinforcement row, terminating at 2.5 m distance from can be further encased in 3D continuum elements of nearly
the base. zero stiffness (dummy elements), enabling a realistic repre-
Adachi et al. (2003) also conducted a series of 1:7 scaled sentation of the pile periphery. (e.g., Giannakos et al.,
physical model tests (static and cyclic), which reproduced 2012; Sakellariadis et al. 2019).
the observed performance and revealed the inherent weak- The key limitation of such approaches is the dependency
ness of the heavier reinforced pier. The increased flexural of the adopted M  c curve to the selected level of axial
resistance of the heavier reinforced pier attracted higher load. The seismic inertia loads are transmitted to the pile
bending moment and, consequently, shear force. Given group to a large extent through axial loading, resulting in
the same shear capacity (same diameter and transverse significant variations of the axial pile forces during shaking
reinforcement), the structure failed due shear failure. On (Fig. 8b). Neglecting the interaction between axial load and
the contrary, the lightly reinforced pier reached its flexural moment capacity, (N  M), may lead to non-negligible dis-
capacity faster, and therefore the shear demand was lim- crepancies, not necessarily on the safe side (Fig. 8a). With
ited. This pier failed in bending, developing flexural cracks, few exceptions, mainly referring to static loading of piles
but did not collapse; by contrast, the shear failure of the (Gerolymos et al., 2015; Psychari et al., 2019) the sensitivity
heavier reinforced pier was brittle. A similar response is of pile group structural response to the variation of axial
observed in the numerical results of Fig. 7c, comparing loading has not yet attracted the necessary attention. To
the 8.5 m to the 12 m tall pier. The first height corresponds the best of our knowledge, no attempt has been made to
to the collapsed pier (12 m  2.5 m = 8.5 m), where a shear investigate the dynamic performance of pile groups
mode of failure is observed. The taller pier (12 m) repre- accounting for this important feature of RC behavior.
sents the case with only 2 rows of reinforcement over the
entire length of the pier, which experiences a more ductile 2.3. Seismic excitation
flexural mode of failure (as would have been the case, if
the pier had been capacity designed). It may therefore be Due to the prohibitive computational cost, limited time-
concluded that the CDP model offers a good approxima- history analyses were performed for the various foundation
tion of complex RC response within the scope of this study. configurations. In the absence of a seismic record at the
The implementation of the CDP model for the RC piles exact location, the Fukiai (000) record was chosen as the
is a key advancement of the present simulation. Although most representative. This motion was recorded just a little
the importance of the structural nonlinearities of the piles further away (about 8 km) from the collapsed bridge but at
has already been pointed out by various researchers, this a similar distance from the Noijima fault, and on similar
is typically accounted for in a simplified manner. geological and soil conditions. The nearly-fault-normal
Gerolymos & Gazetas (2005) investigated the static and component of the record was further deconvoluted at -
dynamic response of piles using a dynamic nonlinear 23 m depth using equivalent-linear analysis, and used as

Fig. 8. Pile group modelling: (a) FE modelling of RC piles using nonlinear beam elements (Mc) vs CDP modelling; and (b) Illustration of the resisting
mechanisms of the pile group under seismic loading.
1456
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

input motion. The time history and its elastic spectrum are subjected vertical push and horizontal pushover analysis.
shown in Fig. 9. The static (vertical) and the seismic (combined loading)
capacity of the existing pile group and of the shallow foot-
3. Actual pile group vs. rocking footing ing are evaluated.

This section initially investigates the performance of the 3.1. Bearing capacity analysis
original bridge–pilecap–pile group configuration. Subse-
quently, the role of the original 17-pile group is critically Fig. 10 presents the results of the FE analysis of the two
assessed by exploring an extreme modification of the foun- foundations under vertical loading in terms of load– settle-
dation: the 17 piles are removed along with the cover layer ment curves. This investigation was conducted for the com-
leaving only the pilecap acting as a shallow footing, thus pleteness of the comparison of the various alternatives
provoking highly nonlinear response. Extensive studies since the vertical (static) safety factor is typically a key
have shown that such nonlinear interaction can be used parameter in foundation design. For completeness, the
for the seismic protection of structures (e.g. e.g., Kutter bearing capacity of the two alternatives is also estimated
et al. 2003; Mergos & Kawashima 2005; Anastasopoulos applying the traditional analytical semi-empirical bearing
et al., 2010a, 2010b), suggesting a plastic design approach capacity methods which are typically used for design ignor-
tolerating full mobilization of foundation moment capacity ing pile-to-pile interaction under the large deformations
under seismic loading. Modern design guidelines, such as prevailing at failure (Kanellopoulos & Gazetas, 2019).
FEMA 750 (2009) and the second generation of Eurocode The bearing capacity of the pile group is estimated as:
8 (Pecker, 2019), progressively encourage such approaches

provided that certain performance criteria are met P ult ¼ n  P shaft þ P tip
(Performance-Based design).  
Prior to comparatively assessing the seismic perfor-
 pd 2
¼ n  p  d  L  a  Su þ  9  S u;tip  70MN ð6Þ
mance of the two foundation alternatives, they are both 4

Fig. 9. Acceleration time histories and elastic response spectra of the original and the deconvolved Fukiai_000 record used for the nonlinear time history
analyses.

Fig. 10. FE deformed mesh with superimposed soil plastic strains and load-settlement ðP  wÞ response of the actual 17-pile group and of the rocking
footing alternative under vertical loading.
1457
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

where: thermore, Eq. 6 does not consider group effects. The


n = 17 (number of piles) numerical result yields a much higher FS v  7:5. On the
d = 1 m (pile diameter) contrary the numerical result of the shallow footing yields
L = 12 m (pile length) P ult  50 MN (Fig. 10) which is in good agreement with
a = 0.6 (adhesion factor) the analytical solution offering an additional verification

S u = 105 kPa (average undrained shear strength over the of the FE modelling technique. Interestingly, for this level
pile length) of vertical loading, the shallow footing (pile cap) is
Sutip = 250 kPa (undrained shear strength at the pile tip) expected to experience a combination of uplift and bearing
Similarly, the bearing capacity of the shallow footing is capacity failure mechanisms under excessive seismic load-
estimated following Prandl΄s bearing capacity formula: ing (Gazetas et al., 2013).

P ult ¼ 5:14  Su  Acap  ð1 þ sc Þ  48:8MN ð7Þ 3.2. Pushover analysis

where: To gain deeper understanding of the behavior of the two


sc = 0.2 (shape factor for square footing) foundation schemes under combined vertical, horizontal
Acap = 121 m2 (pile cap area) and moment loading, a pushover analysis is performed in
S u = Undrained shear strength of the soil layer under- this section. The analysis is displacement controlled and
neath the cap the horizontal displacement is applied 12 m above the
Considering the acting vertical loads foundation level, corresponding to the center of mass of
(P acting ¼ P deck þ P pier þ P cap  20MNÞ, the estimated the deck. For the pushover analyses of the two foundation
capacity for the pile group using bearing capacity theory alternatives, the pier is modeled as perfectly rigid to focus
results to a global factor of safety against purely vertical on foundation response. An additional investigation is per-
loading FS v  3:5, which is a reasonable design value. On formed examining the actual bridge pier assuming fixed
the other hand, despite the extreme modification, the shal- base conditions, to determine its structural capacity.
low foundation offers an adequate vertical factor of safety The results of the FE analysis are collected in Fig. 11
FS v  2:52. and are intentionally presented with respect to shear (hor-
Comparing the numerical results to the bearing capacity izontal) force–displacement ðF  dÞ response, since this
estimates, the vertical capacity of the pile group is found to was found to be critical for the structural integrity of the
be substantially higher. This is primarily attributed to the examined pier. The analysis of the fixed-base RC pier
contribution of the pile cap (ignored in Eq. 6 above), whose shows that it fails primarily in shear, 2.5 m above its base
base and four sides contribute to foundation reaction. Fur- at the point of longitudinal reinforcement cut-off. This is

Fig. 11. FE deformed mesh with superimposed soil and concrete plastic strains and horizontal pushover ðF  dÞ response of the actual 17-pile group (PG),
compared to the shallow footing (F) alternative, and the fixed-base pier (P).
1458
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

consistent with the results of Fig. 7, but also (and most- the damage to structural members, considering their
importantly) with the post-earthquake reconnaissance by response more reliable and repairable. On the contrary,
Kawashima & Unjoh (1997) and Adachi et al. (2003), looking at the k < 1 factor of the modified shallow founda-
and as can be seen in the photos of the collapsed bridge tion, geotechnical failure is provoked prior to plastic hing-
(Fig. 1). For the evaluation of foundation capacity, it is ing of the pier, limiting the inertial load transmitted to the
convenient to use an apparent seismic safety (over- superstructure (as shown below).
strength) factor k defined as:
F foundation;ult 3.3. Seismic performance: Actual capped–17–pile group vs.
k¼ ð8Þ the rocking footing
F pier;ult
According to the results of Fig. 11, the ultimate capacity The performance of the collapsed bridge pier with the
of the pile group foundation is F pilegroup;ult  12MN while two foundation schemes (actual 17-pile group and shallow
the capacity of the shallow footing is P footing;ult  3:8MN . footing) is comparatively assessed in terms of drift d time
The shear capacity of the RC pier is in between these two histories, distinguishing between rotational dh (due to foun-
values: F pier;ult  5:2MN . Using the k factor (Eq. (8)), the dation rotation) and flexural drift dstr , which corresponds
capacity of the two foundation alternatives may be to the structural damage (Fig. 12). The foundation perfor-
expressed as a function of the structural capacity of the pier mance is also evaluated on the basis of settlement-rotation
as kpilegroup ¼ 2:2 (i.e., foundation over-strength) and ðw  hÞ response. The two foundation alternatives are also
kfooting ¼ 0:7 (i.e., foundation under-strength). compared in terms of cumulative RC damage and soil plas-
Regardless of the various simplifications and potential tification (deformed FE model with superimposed plastic
variations from the actual soil conditions (simplified Su strain contours).
profile based on limited in-situ measurements), it is evident Focusing on the actual configuration, the FE model cor-
that the design of the original foundation incorporates a rectly captures the observed failure mode (Fig. 12). The
significant over-strength factor. The pier will fail well shear-type failure at the region of longitudinal reinforce-
before any threatening load is transmitted onto the pile ment cut-off (2.5 m above the pile-cap) is consistently
cap. This is totally consistent with the condition of the reproduced. However, the FE model cannot reproduce
bridge right after the Kobe earthquake: the piers collapsed the post-failure response leading to the dramatic overturn-
while there was no visible evidence of foundation rotation, ing collapse. The pile group experiences limited rocking
settlement, or damage (Fig. 1). Although in this case the and almost the entire drift d is due to structural deflection
over-strength factor was rather high, overall this aspect dstr . The maximum flexural drift dstr;max ¼ 0:27m, is primar-
of design of the original bridge complies to the provisions ily concentrated at the upper section forming a plastic zone
(‘‘capacity” design) of modern seismic codes, which guide 2.5 m above the pile cap.

Fig. 12. Seismic performance of the bridge subjected to the Fukiai_000 record in terms of deformed model with plastic strain contours, drift d time
histories, and settlement-rotation ðw  hÞ response. Comparison of: (a) actual 17-pile group; to (b) shallow footing.

1459
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

This critical section is further investigated plotting the limited to the unconfined cover, with d c not exceeding 0.2;
contours of concrete compressive damage d c and the the M  F load path of the critical section remains well
moment–shear ðM  F Þ load path. Each point of Fig. 13 within the failure envelope throughout the earthquake.
corresponds to the M  F response at the critical section The price to pay is the increased settlement w ¼ 19cm
at each time step throughout the dynamic time history and residual rotational drift dh;res ¼ 4:8cm. In view of
analysis. The results show that the M  F load path reaches the similarity of piers and soil conditions, and given the
the failure envelope at various time increments during 35 m spans, these residual deformations would not cause
shaking, resulting in a large zone of elements around the any appreciable damage; the dramatic collapse is
critical section that reach their ultimate compressive capac- avoided.
ity (d c contours). The strength degradation of the critical Looking back at the drift time histories of the pile group
section after failure can also be visualized in the M  F founded bridge (Fig. 12), limited foundation rocking can
load path: after reaching the failure envelope, the M  F be observed. The moment of maximum drift dmax is picked,
points tend to concentrate around the origin, as they can- to further investigate the loading of the edge piles. The
not exceed the ‘‘degraded” failure envelope (shown results are shown in Fig. 14, both with respect to the axial
schematically in the graph) that corresponds to the residual load–bending moment ðN  MÞ interaction of the critical
capacity of the RC pier. With d c reaching 0.9 and the flex- pile section (Fig. 14a), and in terms of axial force ðN Þ
ural drift of the upper 8.5 m dstr;upper ¼ 0:24m, collapse is and bending moment ðMÞ distribution with depth
certain. The drift ratio of the upper 8.5 m, dr;upper ¼ 2:8% (Fig. 14b). Seismic loading subjects the edge piles to alter-
corresponds to the collapse damage state according to nating compression and tension, with the latter signifi-
Ghobarah (2001). Moreover, the M  F load path reaches cantly reducing their moment capacity. Interestingly, the
the failure envelope at a point corresponding to shear- failure envelope is not reached (Fig. 14a). This can be
dominated failure, which is consistent with the observed explained using the bending moment distributions of
brittle collapse. Fig. 14b. The additional kinematic loading is evident, judg-
In stark contrast, the bridge with the (perhaps extre- ing from the shape of the diagrams at the interfaces
mely) under-designed shallow foundation survives with between the various soil layers. However, focusing on the
minor damage. The maximum total drift is similar to pile head it becomes evident that the compressed piles (4)
the previous case, but is now mainly due to foundation undertake almost the entire loading, while the bending
rocking (Fig. 12). Thanks to foundation compliance, moments at the pile head of the piles under tension (1) is
the structural deflection is limited to dstr;max ¼ 0:04m, almost zero. Interestingly, this is not related to structural
which corresponds to minor damage of the RC pier. This failure, as the N  M load path is not reaching the failure
is confirmed in Fig. 13, which shows that the damage is envelope.

Fig. 13. Comparative assessment of the structural performance of the pier founded on the 17-pile group (left) and the shallow footing (right) in terms of
concrete damage dc (top) and moment–shear ðM  F Þ load path compared to the failure envelope of the critical section (the degraded failure envelope is
shown schematically).
1460
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Fig. 14. Loading of the edge piles at the moment of maximum drift dmax (Fukiai_000 record): (a) axial force-bending moment ðN  MÞ load path
compared to the failure envelope of the RC edge piles; and (b) distribution with depth of axial force ðN Þ and bending moment ðMÞ of the edge piles.

Despite not reaching the failure envelope, the piles structural and acceptable geotechnical performance. This
under tension (pile 1) sustain significant stiffness reduction section investigates 4 intermediate concepts, ranging from
due to micro-cracking, which is nicely captured by the smaller capped-9-pile group and embedded footing, to an
Concrete Damaged Plasticity (CDP) model. This stiffness unconnected piled raft foundation, and to a shallow foot-
reduction leads to load redistribution towards the stiffer ing supported by an unconnected confining ‘‘ring”. The
foundation components (i.e., the compressed piles, 1), thus increase of foundation capacity compared to the shallow
preventing structural failure of the ‘‘weaker” piles under footing is expected to partially or completely cancel the
tension (4). At the end of shaking, a significant number ‘‘rocking isolation” effect, unavoidably increasing pier dis-
of elements close to the pile-cap have exhausted their ten- tress. The four concepts are briefly described below with
sile strength, which is consistent with the post-seismically examples of similar research studies and applications in
observed cracks, and with the reduced stiffness of concrete practice:
specimens in subsequent field and laboratory tests. (i) Reduced piled foundation
The results of the original bridge–pile group configura- A smaller pile group foundation is studied, removing the
tion are in good agreement with post-earthquake inspec- two edge rows of piles in the transverse direction. This
tions, providing increased confidence on the numerical modification results to a 9-pile group consisting of 2 rows
analysis technique. Better understanding is gained with of 4 piles and an additional pile in the center. The reduced
respect to nonlinear pile–soil interaction as the structural pile group is adequate in terms of static (due to purely ver-
nonlinearities of the piles play an important role, which tical loading) safety factor and settlement, offering at the
is typically ignored. Based on the performance of the same time reduced moment capacity in the transverse direc-
bridge it may be safely concluded that foundation seismic tion of seismic loading. A similar concept has been
design should not be treated separately from the super- explored by Sakellariadis et al. (2019) referring to widening
structure, since geotechnical ‘‘success” does not prevent of existing bridges.
structural failure. Furthermore, the study so far provides (ii) Unconnected piled raft
further evidence on the beneficial role of nonlinear soil– The approach of an unconnected piled raft has been
foundation interaction under excessive seismic loading. explored experimentally as a remediation measure for set-
The under-designed shallow footing protects the bridge tlement and rotation of rocking footings (Deng & Kutter,
pier from excessive structural deflection, at the cost of 2012; Loli et al., 2015). Such concepts have also been
increased settlement (and residual rotation). The latter applied in some major bridges, such as the Rion–Antirion
may limit the applicability of such an approach in systems bridge in Greece, where unconnected piles where used in a
where settlements are not tolerable. To that end, it is essen- similar manner as part of a hybrid foundation solution
tial to investigate further mitigation measures or intermedi- (Pecker, 2003). The original configuration is slightly modi-
ate approaches that may offer a compromise between fied, removing the cover layer and the kinematic con-
geotechnical and structural performance. straints between the piles and the cap. An additional thin
soil layer is introduced between the pile cap and the ‘‘im-
4. Intermediate foundation concepts proved soil” targeting low interface friction angle, thus
provoking sliding. Depending on the material of this layer,
Following the previous fundamental conclusions, an such an approach can limit the maximum allowable inertia
attempt is made to make ends meet between successful load, while bearing capacity failure and excessive settle-
1461
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

ments are prevented thanks to the axial pile stiffnesses. This the additional confining stresses and the kinematic con-
layer is added to the FE model as a frictional interface of straint it imposes to the underlying soil, preventing it’s
low friction coefficient l ¼ 0:25, which corresponds to an ‘‘squeezing” out during seismic loading, and thus reducing
interface between precast concrete and silty soil. settlements.
(iii) Embedded footing The four foundation alternatives are initially subjected
The embedded footing is very similar to the previously to vertical and horizontal pushover loading in order to
investigated shallow footing. The main difference is the determine the vertical safety factor FS v and the apparent
cover layer, which is not removed, offering additional con- seismic over-strength factor k (Eq. (8)). Table 1 offers a
finement compared to the shallow foundation. From a summary of the FE results for both types of loading.
practical point of view, this approach is more typical than Fig. 15 compares the mobilized failure mechanisms (FE
a shallow footing, since a small excavation is always neces- deformed mesh with superimposed plastic strain contours)
sary for a number of reasons. The nonlinear seismic beha- and the load settlement ðP  wÞ response of the four foun-
viour of embedded foundations in cohesive soil was dation alternatives subjected to purely vertical loading. The
investigated recently by Zafeirakos and Gerolymos, 2014 mobilised failure mechanisms vary substantially, resulting
who proposed a seismic design methodology focusing on in a wide range of bearing capacities and static (due to
bridge systems. Similarly for non-cohesive soils, Taeseri purely vertical loading) factors of safety ðFS v ¼ 2:7  7Þ.
et al. (2019) investigated the entire range of nonlinear rock- The results of the horizontal pushover analyses are pre-
ing response of embedded footings proposing stiffness- sented in Fig. 16 in terms of mobilized failure mechanisms
degradation charts which can be implemented for design and horizontal force–displacementðF  dÞ response, also
purposes. compared to the capacity of the pier. All alternatives
(iv) Rocking footing with confining ‘‘ring” reduce significantly the over-strength factor
The last alternative is the previously examined rocking k ¼ 0:93  1:3; compared to the original pile group
footing, combined with an unconnected confining ‘’ring’’. ðk ¼ 2:2Þ. In 2 out of 4 cases (reduced pile group, embed-
The latter is a heavy hollow rectangular RC slab, placed ded footing), the pier remains the most vulnerable compo-
next to the footing with a small clearance. Being nent ðk > 1Þ. The resistance of the unconnected piled-raft is
unconnected to the footing, its contribution is limited to almost equal to that of the pier, while the shallow footing
with confining ring is the only case of under-strength
ðk < 1Þ. Interestingly, the four concepts do not only cover
Table 1
Vertical safety factor FS v and over-strength factor k of the studied a wide range of k, but also vary in stiffness.
foundation alternatives. The seismic performance (Fukiai_000 record) of the
Foundation Configuration FS v k alternative foundation configurations is comparatively
assessed in Fig. 17, in terms of deck drift time histories (dis-
Capped 17-Pile group (original) 8 2.20
Capped 9-Pile group 5 1.30 tinguishing between dh and dstr ) and settlement–rotation
Unconnected piled raft 7 1.01 ðw-hÞ response. FE deformed mesh with superimposed con-
Embedded footing 3.5 1.15 tours of concrete cumulative damage and soil plastification
Shallow footing with confining ring 2.7 0.93 are also plotted. The results are summarized in Table 2
Shallow footing (rocking) 2.5 0.70

Fig. 15. Four intermediate foundation concepts subjected to purely vertical loading: mobilized failure mechanisms (FE deformed mesh with superimposed
soil plastic strains) (left) and load–settlement ðP  wÞ response (right).

1462
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Fig. 16. Four intermediate foundation concepts subjected to horizontal pushover loading: mobilized failure mechanisms (FE deformed mesh with
superimposed soil plastic strains) (left) and horizontal force–displacement (F  d) response (right), compared to the fixed-base RC pier.

with respect to the maximum values of deck drift and resid- zone is avoided: d c ¼ 0:9 only at the pier edges. With this
ual values of foundation rotation and settlement. foundation configuration, although the damage might have
Starting with the smaller 9-pile group (Fig. 17), the been irreparable, collapse would probably be avoided.
structural performance is improved compared to the origi- Although the foundation experiences significant uplifting
nal 17-pile group. Thanks to the increased foundation com- (as revealed by the w  h plot of Fig. 17b), the residual set-
pliance and the mobilized nonlinear energy dissipation tlement is limited to w ¼ 2:2cm; the residual rotation is
mechanisms, the rocking drift is increased (dh;max ¼ negligible.
0.11 m compared to 0.04 m of the original), leading to The seismic performance of the remaining two configura-
20% reduction of the flexural drift (dstr;max = 0.23 m as tions is qualitatively similar to that of the previously ana-
opposed to 0.27 m). As for the original foundation, the lyzed shallow footing. The embedded footing results in
piles survive the earthquake without any structural dam- over-strength factor k > 1, therefore the pier remains the
age. Still though, the pier sustains severe damage, as most vulnerable system component. However, the seismic
revealed by the M  F load path and the contours of of performance of the pier founded on an embedded footing
concrete damage d c of Fig. 18. As for the original 17-pile seems to be the most ‘‘balanced”, satisfying both structural
group, after reaching the failure envelope, the M  F and geotechnical performance criteria. The flexural drift is
points are bounded by the ‘‘degraded” failure envelope almost double compared to the rocking footing
(shown schematically), which is indicative of a high proba- (dstr;max = 0.08 m compared to 0.04 m), but the pier survives
bility of collapse. Nonetheless, the reduction of dstr;max is the earthquake with repairable damage (Fig. 17c). This is
promising in terms of the potential application of such pile confirmed in Fig. 18c, where it can be seen that the M  F
group layouts (limited pile rows in the critical direction of load path (of the critical section) marginally reaches the fail-
seismic loading). Especially for retrofit of upgraded/ ure envelope, without any signs of strength degradation.
widened bridges, installing the new piles (if needed for ade- The limited amount of damage is revealed more clearly by
quate FS v or for settlement reduction) close to the founda- the contours of concrete damage, with d c  0.3. Although
tion center of rotation would not increase k, thus allowing the foundation moment capacity is 15% higher than that
beneficial nonlinear pile group response. of the pier ðk ¼ 1:15Þ, the energy dissipation due to nonlin-
Proceeding to the unconnected piled raft concept, the ear foundation response reduces structural damage signifi-
flexural drift is reduced by 60% compared to the original cantly. At the same time, the settlements are reduced by
configuration (Fig. 17b, Table 2). Thanks to the reduced 35% compared to the shallow footing.
k ¼ 1.01 (the foundation moment capacity is almost equal The last foundation configuration incorporates a hybrid
to that of the pier), the critical section does not fail com- concept, adding a ‘‘confining ring” to the previously
pletely. This can be seen in the M  F load path (of the crit- explored rocking (shallow) footing. The role of such addi-
ical section) and the contours of concrete damage d c of tional component is dual: it increases the confining pres-
Fig. 18. In contrast to the previous case, no strength degra- sures around the foundation, also offering kinematic
dation is observed and the formation of a broad failure restraint to the shearing of soil in the vicinity of the foot-

1463
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Fig. 17. Seismic performance of the bridge founded on the four intermediate foundation configurations (Fukiai_000 record). FE deformed mesh with
superimposed concrete damage and soil plastic strains (left), deck drift d (distinguishing bewteen the rocking dh and the flexural dStr component), and
settlement–rotation ðw  hÞ response: (a) smaller capped 9-pile group; (b) unconnected piled raft; (c) embedded footing; and (d) shallow footing with
confining ring.

Table 2
Seismic performance of the bridge for all foundation concepts (Fukiai_000 record).
Foundation Configuration dstr;max ðmÞ dh;max ðmÞ hres ðradÞ wðcmÞ
17-Pile group (original) 0.27 0.04 0.0009 0.5
9-Pile group 0.23 0.11 0.003 1.3
Unconnected piled raft 0.11 0.24 0.0001 2.2
Embedded footing 0.08 0.24 0.003 12.2
Footing with confining ring 0.05 0.25 0.004 14.5
Shallow footing (rocking) 0.04 0.25 0.004 18.8

1464
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Fig. 18. Comparative assessment of the structural performance of the pier founded on the four intermediate foundation alternatives in terms of concrete
damage dc and moment–shear ðM  F Þ load paths compared to the failure envelope of the critical section (the degraded failure envelope is shown
schematically).

ing, reducing its tendency to be squeezed out during cyclic tion employing a confining ring would have prevented
loading. This dual role is proven beneficial (Fig. 17), pre- structural failure at the cost of further increased settle-
venting structural damage of the pier, equally sufficient to ments. The unconnected piled raft improved substantially
the shallow footing (dstr;max ¼ 0.05 m as opposed to the overall performance, but could be much better exploited
0.04 m), and reducing by 25% the foundation settlement if the sliding (shear) capacity of the footing was lower than
(w ¼14.5 cm instead of 18.8 cm). As shown in Fig. 18, that of the pier. Finally, the reduced pile group would not
M  F load path (of the critical section) remains within have prevented collapse, but still led to improved overall
the failure envelope throughout the seismic motion. It is performance, constituting a promising alternative if
worth mentioning that the dimensions of the confining ring designed properly along with structural design.
were selected indicatively. The performance could be In all cases examined almost the entire flexural distor-
improved further, if the dimensions and dead load of the tion was concentrated at the upper 8.5 m of the pier.
confining ring were optimized. Table 3 characterizes the performance of the studied alter-
Summarizing, the performance of the examined founda- natives, using the damage states of Ghobarah (2001) in
tion alternatives lies between that of the original 17-pile function of the drift ratio of the upper section. For the sin-
group and the rocking shallow footing. It is worth pointing gle record used in this study (Fukiai_000), the structural
out that the main scope of this study was to explore how dif- damage ranges from repairable to collapse. The presented
ferent foundation concepts would have potentially altered results should not be generalized; the damage state of the
the dramatic collapse of the Fukae bridge. To that end, it bridge for each foundation configuration is sensitive not
is crucial to account for the poorly reinforced RC pier only to the intensity but also to the frequency characteris-
and the level of seismic excitation that substantially tics of the input motion. To fully explore the advantages
exceeded the design considerations, when evaluating the and limitations of each scheme, further parametric studies
previous results. Under the specific assumptions, the accounting for various earthquake scenarios are essential.
embedded footing was proven the most efficient solution However, it may be concluded that the foundation concept
preventing structural damage at the cost of tolerable settle- significantly affects the seismic performance and can poten-
ments. Similarly, the shallow footing and the hybrid solu- tially enhance the resilience of critical infrastructure.

Table 3
Comparison of all foundation alternatives in terms of damage state, based on the drift ratio.
Foundation configuration dstr;max ðmÞ dr;max Damage State
17-Pile group (original) 0.24 2.8% Collapse
9-Pile group 0.19 2.2% Severe
Unconnected piled raft 0.072 0.85% Irreparable
Embedded footing 0.043 0.5% Repairable 8.5 m
Footing with confining ring 0.038 0.45% Repairable
Shallow footing (rocking) 0.04 0.47% Repairable

1465
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

5. Synopsis and conclusions  The unconnected piled raft alternative reduces both the
structural distress and the settlement. To fully exploit
The collapse of the Fukae bridge is a dramatic exam- this approach, the superstructure should be tolerant to
ple of a system prone to brittle structural failure sub- potentially large displacements due to sliding.
jected to seismic loading substantially exceeding its  Partial or full embedment of the footing can be an effec-
design. Its failure was accompanied by ‘‘successful” foun- tive measure for settlement reduction, but a large
dation performance. To gain deeper insights on this con- embedment may partially cancel the isolation effect of
tradictory response, advanced numerical tools were a rocking foundation.
implemented regarding RC members (pier and piles)  Hybrid approaches adding confinement and kinematic
and soil nonlinear behavior. The numerical simulation constraints (using a confining ring) around the rocking
successfully reproduced the shear-dominated mode of foundation can be quite efficient. Such approaches
failure at the region of longitudinal reinforcement cut- would require optimization of dimensions and confine-
off. The analysis also confirmed that, despite being highly ment to maximize benefits.
overdesigned, the pile group foundation experienced lim-
ited but non-negligible swaying and rocking during shak- All cases examined adopted the structural details of the
ing, as a result of which the piles were subjected to collapsed pier. The performance of each concept could be
tension and combined shear-moment loading. The result- further improved following a holistic design approach,
ing stiffness reduction of the cracked under tension piles treating the structural members and the foundation as a
leads to load redistribution towards the stiffer compressed system. The weakness of the current design trend is sum-
piles, preventing plastic hinging of the weaker piles (un- marized in the paradox of the overturned collapsed bridge
der tension). These findings are consistent with post- and the almost intact foundation. Evidently, over-sized
earthquake in-situ testing. foundations can lead to seismically inadequate systems,
The successful comparison of the analysis to post- while permitting or provoking nonlinear soil–foundation
earthquake observations increases further the confidence response can be quite beneficial. The penalty of residual
to the numerical analysis technique and the implemented settlement (and rotation) should be considered and miti-
concrete damaged plasticity (DCP) model. This allowed gated, especially for deformation intolerant structures. To
deeper understanding of complex pile group–soil–structure that end, performance–based design considering the entire
interaction mechanisms, accounting for structural nonlin- soil-foundation-structure system (and not each component
earities (in addition to soil). The fact that the ‘‘injured” individually) is essential towards enhancing the safety and
piles were left in place and remained in use for the new resilience of critical infrastructure.
bridge, despite their cracking and stiffness reduction,
inspired the extreme modification of the original founda- Acknowledgement
tion concept, removing all 17 piles and the cover layer. In
this way, highly nonlinear SFSI was provoked, which The financial support for this paper has been provided
was found to prevent collapse at the cost of increased set- by the Swiss Federal Roads Office (FEDRO) within the
tlement and limited residual foundation rotation. The project AGB2017/001 (Development of reliable methods
favorable response of the pier founded on just the pile for optimized retrofit design of bridge pile groups) and is
cap offers support to the rocking foundation concept greatly appreciated. The authors would like to acknowl-
(e.g., Kutter et al. 2003; Mergos & Kawashima 2005; edge the contribution of Dr. Evangelia Garini in the decon-
Anastasopoulos et al., 2010a, 2010b). volution analysis, as well as Antonia Psychari and
To deal with the shortcomings of the rocking footing, but Athanasios Agalianos in the calibration of CDP model
still gain the benefits of nonlinear SFSI, four additional parameters. The support of Dr. Hiroyuki Kimata and
intermediate foundation concepts were investigated. The Dr. Yukio Adachi in providing valuable data is also grate-
examined concepts were based on the minimum modifica- fully acknowledged. The paper is dedicated to the memory
tion of the original configuration; no optimization was fol- of Dr. Takashi Tazoh, whose contribution in understand-
lowed. In this sense, accounting also for the poor ing and documenting case histories from the Kobe 1995
structural details of the pier and the (realistic but) single seis- earthquake is invaluable.
mic record used, it is not possible to conclude on a clearly
superior approach. References
Based on the numerical results of the specific case exam-
ined, the following conclusions can be drawn: ABAQUS 6.11. (2014). Standard user’s manual. Dassault Systèmes
Simulia Corp., Providence, RI, USA.
Adachi, Y., Ishizaki, H., Ikehata, S., Ikeda, S., 2003. Verification of the
 Reducing the number of pile rows in the critical direc- Failure of Reinforced Concrete Piers under the Near-field Earthquake.
tion of seismic loading is promising in improving the Hanshin Expressway Public Corporation Publications, Japan.
seismic performance, but not sufficient to prevent severe Agalianos, A., Sieber, M., Anastopoulos, I., 2020. Cost-effective analysis
damage of the examined pier. technique for the design of bridges against strike-slip faulting.

1466
L. Sakellariadis et al. Soils and Foundations 60 (2020) 1450–1467

Earthquake Engineering & Structural Dynamics 49, 1137–1157. Ghobarah, A., 2001. Performance-based design in earthquake engineer-
https://doi.org/10.1002/eqe.3282. ing: State of development. Eng. Struct. 23 (8), 878–884. https://doi.
Anastasopoulos, I., Gazetas, G., Loli, M., Apostolou, M., Gerolymos, N., org/10.1016/S0141-0296(01)00036-0.
2010a. Soil failure can be used for seismic protection of structures. Kanellopoulos, K., Gazetas, G., 2019. Vertical static and dynamic pile-to-
Bulletin of Earthquake Engineering, 309–326. pile interaction in non-linear soil. Géotechnique, 1–16.
Anastasopoulos, I., Georgarakos, T., Georgiannou, V., Drosos, V., Kawashima, K., Unjoh, S., 1997. The damage of highway bridges in the
Kourkoulis, R., 2010b. Seismic performance of bar-mat reinforced-soil 1995 Hyogo-ken nanbu earthquake and its impact on Japanese seismic
retaining wall: Shaking table testing versus numerical analysis with design. Journal of Earthquake Engineering 1 (03), 505–541.
modified kinematic hardening constitutive model. Soil Dynamics and Kutter BL, Martin G, Hutchinson TC, Harden C, Gajan S, Phalen JD
Earthquake Engineering 30 (10), 1089–1105. (2003) Status report on study of modeling of nonlinear cyclic load-
Anastasopoulos, I., Gelagoti, F., Kourkoulis, R., Gazetas, G., 2011. deformation behavior of shallow foundations. In: PEER workshop,
Simplified constitutive model for simulation of cyclic response of University of California, Davis, March 2003
shallow foundations: validation against laboratory tests. Journal of Loli, M., Knappett, J.A., Brown, M.J., Anastasopoulos, I., Gazetas, G.,
Geotechnical and Geoenvironmental Engineering 137 (12), 1154–1168. 2015. Centrifuge testing of a bridge pier on a rocking isolated
Anastasopoulos, I., Loli, M., Gelagoti, F., Kourkoulis, R., Gazetas, G., foundation supported on unconnected piles. Proceedings of 6th
2012. ’Nonlinear soil–foundation interaction: numerical analysis’. ICEGE, paper, 362..
Proc. 2nd Int. Conf. on Performance-Based Design in Earthquake Mander, J.B., Priestley, M.J.N., Park, R., 1988. Theoretical Stress-Strain
Geotechnical Engineering. Model for Confined Concrete. Journal of Structural Engineering 114
Bentz, E.C., 2000. Sectional Analysis of Reinforced ConcreteMembers. (8), 1804–1826.
Department of Civil Engineering. University of Toronto, Toronto, Mergos, P.E., Kawashima, K., 2005. Rocking isolation of a typical bridge
ON, Canada, p. 184. pier on spread foundation. Journal of Earthquake Engineering 9 (sup
Bilotta, E., Lanzano, G., Madabhushi, S.G., Silvestri, F., 2014. A 2), 395–414.
numerical Round Robin on tunnels under seismic actions. Acta Mylonakis, G., Syngros, C., Gazetas, G., Tazoh, T., 2006. The role of soil
Geotechnica 9 (4), 563–579. in the collapse of 18 piers of Hanshin Expressway in the Kobe
Chang, G., and Mander, J. B. (1994). ‘‘Seismic energy based fatigue earthquake. Earthquake engineering & structural dynamics 35 (5),
damage analysis of bridge columns: Part I—Evaluation of seismic 547–575.
capacity.” NCEER Technical Report No. 94-0006, Buffalo, N.Y. Pecker, A., (2003) A seismic foundation design process, lessons learned
Deng, L., Kutter, B.L., 2012. Characterization of rocking shallow from two major projects: the Vasco de Gama and the Rion Antirion
foundations using centrifuge model tests. Earthquake Engineering & bridges. In ACI International Conference on Seismic Bridge Design
Structural Dynamics 41 (5), 1043–1060. and Retrofit. La Jolla, California.
FEMA. 2009. Recommended seismic provisions for new buildings Pecker, A. (2019). Development of the second generation of Eurocode 8–
andother structures. FEMA 750. Washington, DC: Building Seismic Part 5: A move towards performance-based design. In Proceedings of
Safety Council, National Institute of Building. the 7th International Conference on Earthquake Geotechnical Engi-
Iwasaki, T., Fujino, Y., Iemura, H., Ikeda, S., Kameda, H., Katayama, T., neering, ICEGE, Rome, Italy (p. 273). CRC Press.
Kawashima, K., Onishi, Y., Saeki, S., Toki, K., 1995. Report on Prakash, S., Belarbi, A., You, Y.M., 2010. Seismic performance of circular
Highway bridge damage caused by the Hyogo-ken Nanbu earthquake RC columns subjected to axial force, bending, and torsion with low
of 1995. Committee on Highway Bridge Damage caused by the and moderate shear. Engineering Structures 32 (1), 46–59.
Hyogo-Ken Nambu Earthquake, Japan. Psychari, A., Agalianos, A., Sakellariadis, L., Anastasopoulos, I., 2019.
Gazetas, G., Anastasopoulos, I., Gerolymos, N., Mylonakis, G., Syngros, On the effect of RC moment-axial load interaction in the moment
C., 2006. The Collapse of the Hanshin Expressway (Fukae) Bridge, capacity of a pile group. In Proceedings of the 2nd International
Kobe 1995: Soil-Foundation-Structure Interaction, Reconstruction. Conference on Natural Hazards and Infrastructure.
Seismic Isolation. Entwicklungen in der Bodenmechanik, Bodendy- Sakellariadis, L., Marin, A., Anastasopoulos, I., 2019. Widening of
namik und Geotechnik, pp. 93–120. Existing Motorway Bridges: Pile Group Retrofit versus Nonlinear Pile-
Gazetas, G., Anastasopoulos, I., Adamidis, O., Kontoroupi, T., 2013. Soil Response. Journal of Geotechnical and Geoenvironmental Engi-
Nonlinear rocking stiffness of foundations. Soil Dynamics and neering 145 (12), 04019107.
Earthquake Engineering 47, 83–91. Taeseri, D., Laue, J., Anastasopoulos, I., 2019. Non-linear rocking
Gerolymos, N., Papakyriakopoulos, O., Brinkgreve, R.B.J., 2015. stiffness of embedded foundations in sand. Géotechnique 69 (9), 767–
Macroelement modeling of piles in cohesive soil subjected to combined 782.
lateral and axial loading. Analytical Methods in Petroleum Upstream Tsinidis, G., Pitilakis, K., Trikalioti, A.D., 2014. Numerical simulation of
Applications 373. round robin numerical test on tunnels using a simplified kinematic
Gerolymos, N., Gazetas, G., 2005. Phenomenological model applied to hardening model. Acta Geotechnica 9 (4), 641–659.
inelastic response of soil-pile interaction systems. Soils and Founda- Vucetic, M., Dobry, R., 1991. Effect of soil plasticity on cyclic response.
tions 45 (4), 119–132. Journal of geotechnical engineering 117 (1), 89–107.
Giannakos, S., Gerolymos, N., Gazetas, G., 2012. Cyclic lateral response Zafeirakos, A., Gerolymos, N., 2014. Towards a seismic capacity design of
of piles in dry sand: Finite element modeling and validation. caisson foundations supporting bridge piers. Soil Dynamics and
Computers and Geotechnics 44, 116–131. Earthquake Engineering 67, 179–197.

1467

You might also like