Journal of Advanced Concrete Technology Vol. 7, No.
3, 439-454, October 2009 / Copyright © 2009 Japan Concrete Institute
Scientific paper
Nonlinear Seismic Response and Damage of Reinforced Concrete Ducts
in Liquefiable Soils
Mohammad Reza Okhovat1, Feng Shang2 and Koichi Maekawa3
Received 19 May 2009, accepted 31 July 2009
Abstract
Highly inelastic nonlinear interaction of liquefied soil and underground RC ducts is computationally investigated in
view of the structural damage. The required ductility is expected to be insensitive to the risk of liquefaction for normally
deposited layers of soil, and the lesser ductility is acceptable for the case of highly liquefiable foundation similar to the
seismic isolation. Although the structural nonlinearity has a fewer effect on the uplift of the underground ducts, the
amount of main reinforcement may control the structural damage with the same efficiency for both drained and
undrained soil deposits.
1. Introduction
Underground facilities as indispensable parts of urban
infrastructures are used for various applications ranging
from small pipelines to large RC ducts and tunnels. In
the past decades, less seismic damage to large under-
ground structures was reported compared to on-ground
ones although damages to small pipelines were ob-
served as early as 1964 in Niigata and Alaska earth-
quakes (Hall and O’Rourke 1991). These facts made
structural engineers deem that underground RC struc-
tures might be rather safe until Hyogoken-Nanbu earth-
quake in 1995. This earthquake caused severe damages
in some RC underground structures which failed in the
mode of shear accompanying deteriorated axial load
carrying mechanism (An et al. 1997). Afterward, this
catastrophic damage was observed again in the follow-
ing earthquakes of 1999 Chi-Chi, Taiwan (Wang et al.
2001) and 1999 Duzce, Turkey earthquakes (O’Rourke
et al. 2001). Thus, earthquake-induced damages to un-
derground structures and the corresponding earthquake
resistant designs have received remarkable attention.
Although the seismic performance of underground
structures has been extensively investigated (for exam-
ple An et al. 1997; Huo et al. 2005; Hashash et al. 2001),
there have been limited studies regarding the liquefac-
tion-related seismic performance. Kimura et al. (1995)
conducted centrifuge tests to examine the effect of vari-
ous countermeasures against liquefaction with a mock-
up of underground structures. Koseki et al. (1997a and
1
Ph.D. candidate, Department of Civil Engineering,
University of Tokyo, Tokyo, Japan.
E-mail:okhovat@concrete.t.u-tokyo.ac.jp
2
Ph.D. candidate, State Key Laboratory of Hydro
Science and Engineering, Tsinghua University, Beijing,
P. R. China.
3
Professor, Department of Civil Engineering, University
of Tokyo, Tokyo, Japan.
439
1997b) investigated dynamic behaviors of several types
of underground structures in liquefiable sand subjected
to earthquake excitations. Tamari and Towhata (2003)
performed a series of shaking table tests on an alumi-
num fixed base structure model embedded in liquefiable
soil to propose a solution for determining dynamic soil
pressures on a flexible underground structure during
liquefaction. In most of these researches, main focus has
been addressed to nonlinear behaviors of soil foundation
where a simple linear response of built-in structures was
assumed. In contrast, this paper aims at deeper views
towards the inelastic performances of underground RC
structures subjected to seismic excitations in considera-
tion of extremely high nonlinearity of liquefied soil
foundations.
Seismic actions for practical designs of underground
ducts are currently characterized by the forced dis-
placement and/or mean strains imposed to the target
structure through fictitious springs of soil foundation.
First, free-field ground deformations under a specified
design seismic event(s) are estimated and second, the
underground RC is designed to accommodate these re-
quired deformations. This approach is satisfactory if the
underground facility is located in a stiffer soil matrix
(Hashash et al. 2001).
In this paper, inelasticity and damage of RC ducts like
subway tunnels, having interaction with liquefiable soils
of extremely high nonlinearity, are targeted in use of the
nonlinear dynamic finite element analysis. For investi-
gating how much the structural nonlinearity would in-
fluence on the entire response, both linear and nonlinear
analyses are used to numerically reproduce the under-
ground structures. The mechanical properties and di-
mensions of structural models are systematically
changed to examine their effects on the entire structural
performances under ground. All analyses are conducted
in both drained and undrained states of pore water. Liq-
uefaction may greatly increase in the deformation of soil
around structures, but at the same time, the stiffness and
440 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009
internal stress of soil are dramatically reduced as well. A
question is raised, what is the resultant of the both
kinematics in RC damages?
2. Nonlinear constitutive models
2.1 Constitutive model for reinforced concrete
A reinforced concrete material model has been con-
structed by combining constitutive laws for cracked
concrete and those for reinforcement. The fixed multi-
directional smeared crack constitutive equations
(Maekawa et al. 2003) are used as summarized in Fig. 1.
Crack spacing and diameters of reinforcing bars are
implicitly taken into account in smeared and joint inter-
face elements no matter how large they are.
The constitutive equations of structural concrete sat-
isfy uniqueness for compression, tension and shear
transfer along crack planes. The bond between concrete
and reinforcing bars is taken into account in the form of
tension stiffening model, and the space-averaged stress-
strain relation of reinforcement is assumed to represent
the localized plasticity of steel around concrete cracks.
The hysteresis rule of reinforcement is formulated based
upon the Kato’s model (1979) for a bare bar under re-
versed cyclic loads. This RC in-plane constitutive mod-
eling has been verified by member-based and structural-
oriented experiments. Herein, the authors skip the de-
Fig. 1 Composition of constitutive models of reinforced concrete (Maekawa et al. 2003).
tails of the RC modeling by referring to Maekawa et al.
(2003).
2.2 Constitutive model for soil
A nonlinear path-dependent constitutive model of soil is
essential to simulate the entire RC-soil system. Here, the
multi-yield surface plasticity concept (Towhata and
Ishihara 1985; Towhata 2008; Maekawa et al. 2003) is
applied to formulate the shear stress - shear strain rela-
tion of the soil following Masing’s rule (1926).
The basic idea of this integral scheme is actualized to
sum up component stresses which may represent micro-
scopic events. First, the total stress applied on soil parti-
cle assembly, denoted by σij, can be decomposed into
the deviatoric shear stresses (sij) and the mean confining
stress (p) as,
σ ij = s ij + pδ ij (1)
where δij is Kronecker’s delta symbol.
Soil is idealized as an assembly of finite numbers of
elasto-perfectly plastic components, which are concep-
tually connected in parallel. As each component is given
different strengths, all components subsequently begin
to yield at different total shear strains, which results in a
gradual increase of entire nonlinearity. The nonlinear
behavior appears naturally as a combined response of all
 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009 441

components. Hence, the authors propose the total shear For simulation of the pore water pressure and related
stress carried by soil particles being expressed with re- softening of soil stiffness in shear, the volumetric
gard to an integral of each component stress as, nonlinearity of soil skeleton has to be taken into account.
The authors simply divide the dilatancy into two com-
n
sij = ∑ sijm (ε kl , ε mpkl , Go m , F m ) ponents according to the microscopic events. One is the
m =1
consolidation (negative dilation) as unrecoverable plas-
ticity denoted by εvc. The other is the positive dilatancy
dsijm = 2Gom deeij
m
= 2Gom (deij − de mpij ) associated with alternate shear stress due to the overrid-
(2)
sijm ing of soil particles, which is denoted by εvd as,
de mpij = df ,
2F m p = 3K 0 (ε 0 − ε v ) , ε v = ε vc + ε vd (4)
s klm dekl s klm dε kl where K0 is the initial volumetric bulk stiffness of soil
df = =
Fm F m
particles assembly and can be calculated by assuming
the initial elastic Poisson’s ratio denoted by ν (=0.2) as,
where εkl and εpkl are total and plastic strain tensors of
(k,l) component, Gom and Fm are the initial shear stiff- 2(1 +ν ) (5)
K0 = G0
ness and the yield strength of the m-th component, (eij, 3(1 − 2ν )
empij, emeij) are deviatoric tensors of total strain, those of
plastic and elastic strains of the m-th component, re- The volume reduction of pores among soil particles
spectively. These component parameters can be will cause increasing pore pressure under undrained
uniquely decided from the shear stress strain relation of states, which may lead to liquefaction. According to
soil under the referential constant confinement (Maki et experiments of sandy soils, the following formulae are
al. 2005). adopted as,
ε vc = ε v ,lim {1 − exp(− 2( J 2 p + J 2 p ,ini ) )}− ε vc,ini (6)
In general, the volumetric components may fluctuate
and affect the shear strength of the soil skeleton. In real-
ity, the shear strength of soil may decay when increasing
pore water pressure leads to reduced confining stress of 1
J 2 p = ∫ dJ 2 p , dJ 2 p ≡ skl ⋅ dε kl (7)
soil particle skeleton. The multi-yield surface plastic 2
envelope may inflate or contract according to the con-
finement stress as shown in Fig. 2. It can be formulated which is represented by the accumulated shear strain
by summing up the linear relation of the shear strength invariant of the soil skeleton denoted by J2p (Maekawa
and the confinement stress as, and An 2000; Maki et al. 2005), and εv,lim is the intrinsic
volumetric compacting strain corresponding to the
F m = χ Finim minimum void ratio as,
(3)
χ=
(c − I1′ tan φ )
ε vc,ini = ε v ,lim {1 − exp(−2 J 2 p,ini )}
Su (8)
(σ + σ 2′ + σ 3′ )
I 1′ = 1
′ (
ε v ,lim = 0.1 log10 I1′0.6 + 1.0 )
3
If the relative density of soil is assumed to be Dr, the
where Su is the specific shear strength corresponding to following relation can be used to inversely decide J2p,ini,
a certain confinement (98kN), Fmini is the specified yield which is a constant corresponding to the initial com-
strength of the m-th component corresponding to Su, χ is pactness of soil particles as,
the confinement index, (c,φ) are the cohesive stress and
ε vc , ini
= {1 − exp( −2 J 2 p , ini )}
the frictional angel, respectively.
Dr (%) = (9)
ε v , lim
The shear provoked dilation, which is path-independent
and defined by the updated shear strain intensity J2s, is
empirically formulated as,

(a J 2 s )2
ε vd = η
1 + (a J 2 s )2
1 (10)
J 2s = eij eij , a = 25.0
2
0.015 (ε vc + ε v ,ini )
η=
Fig. 2 Confinement dependent soil model under drained ε v ,lim
cyclic shear loadings.
442 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009
Within this scheme, the liquefaction induced nonlin-
earity and cyclic dilatancy evolution can be consistently
computed. Figure 3 shows the pure shear stress-strain
relation and the corresponding pore pressure of perfect
undrained soil. Here, the soil characterisitics are repre-
sented solely by the initial stiffness and the relative den-
sity. The pore water media is assumed to be perfect elas-
tic body with no shear stiffness and Biot’s two-phase
theory (1962) was coupled with the effective stress
model of soil skeleton as stated above. Mean shear stiff-
ness decay and the following cyclic mobility are repro-
duced as shown in Fig. 3.
The general 3D constitutive model of soil is directly
applied to the 2D plane strain field to consider the high
confinement along the tunnel axis. Here, out-of-plane
strain is computationally forced to zero and the corre-
sponding out-of-plane confining stress is computed as a
variable together with the in-plane stresses, since the
stress states are fully three-dimensional.
The multi-yield surface plasticity model has two main
advantages. First, it can easily simulate the shear cyclic
responses by means of the simplified algorithm with
rather few material constants. Second, the path-
dependency of soil can be represented only by the plas-
tic strains (empij) of all constituent components. As the
multi-component scheme has the great similarity to the
contact density model of crack shear transfer (Li et al.
1989) and the multi-directional crack modeling of rein-
forced concrete (Maekawa et al. 2003), higher stability
of computing soil-RC structural interaction is made pos-
sible.
The overall experimental verification of the interac-
tion analysis was reported (JSCE 2002) in the case of
drained soil with RC ducts. The model has been used to
simulate the static and seismic behaviors of nonlinear
soil-structure systems (An et al. 1997; Nam et al. 2006).
Regarding the liquefiable soil-RC interaction, the appli-
cability of the models used in this paper was examined
and verified by Maki et al. (2005).
2.3 Constitutive model for joint interface
It is obvious that the interface property has an influence
on the entire soil-structure responses as well. As soil and
structure have different nonlinearity, complete contact at
the interfaces does not hold. The initial interface of the
Fig. 3 Confinement dependent soil model under undrained cyclic shear loading.
soil and the structure is naturally in mostly complete
contact under static loads (self-weight and service
loads) with lateral earth pressures that soil exerts on the
structure. Under reversed cyclic shear due to the seismic
excitations, however, rocking of a structure may cause
gap-opening and shear sliding along the soil-structure
contact especially when cohesive soil is assumed
(Maekawa et al. 2003).
In this paper, the contact interface model assuming a
bilinear relation for the opening/closure mode is em-
ployed. The normal stress of the joint is assumed to be
zero in the case of separation. The contact stiffness in
closure mode is assigned to be so large as to numeri-
cally ensure no overlapping as shown in Fig. 4(a). For
the shear sliding mode, the shear stress-slip relation is
assumed to be linear till the frictional limit as shown in
Fig. 4(b). A couple of contact planes are allowed to
slide if the magnitude of the applied shear stress ex-
ceeds the frictional limit, which follows the Coulomb’s
law (Maekawa et al. 2008). The initial state of the soil-
structure interface must be simulated to represent the
static earth pressure. This is achieved by performing an
analysis that considers the body force of the soil mass
alone before applying dynamic actions.
The current joint interface model can be applied to
the states of soil liquefaction as well as the drained
static and dynamic conditions and pre-liquefied soils.
However, due to the dramatically reduced shear stiffness
of liquefied soil similar to liquid, a quasi-hydrostatic
pressure consequently develops inside the soil founda-
tion after liquefaction and it allows less separation to
occur between soil and RC joint interfaces. Then, the
joint interface modeling occupies minor roles on the
structural damage after liquefaction.
3. Simulation of liquefiable soil and RC
ducts
3.1 Model properties
To investigate the seismic damages of underground RC
ducts, a typical subway tunnel section is modeled whose
wall and slab dimensions are shown in Fig. 5(a). The
center column to mainly support the dead weight of soil
overlay has a rectangular cross section of 0.60 × 0.80 m
and is idealized as firmly fixed to the upper and bottom
 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009 443

re-contacted
Normal stress: σ Shear stress: τ

O pen and closed again

closed μσ
δ
ω
Joint open Joint closed Joint −μσ
opened

(a) Normal response (b) Shear response

Fig. 4 Normal and shear response of linear elastic interface model. (Maekawa et al. 2008).

slabs. The clear distance between two adjacent columns
is 3.0 m. The tunnel which is regarded as the standard
case is stiffened with 45° haunches at the corners and
has a longitudinal reinforcement ratio of 1.1% for side
walls and slabs, 1.6% for the column, and web rein-
forcement ratio of 0.2% for all elements. The soil de-
posit is assumed to be loose sand with a thickness of 15
m which is put on a 5-meter-thick layer of the non-
liquefiable soil which lies on the bedrock as shown in
Fig. 5(a). The details of material property for rein-
forcement, concrete, joint interface and non-liquefiable
clayey layer are shown in Table 1.
By assuming the plane strain condition, the finite
element mesh used in the analysis is composed of eight-
node isoparametric 2D elements. The RC-soil interfacial
elements are placed in between the soil and the RC ele-
ments. Since the angle of internal friction of the model
sand is 30°, the friction angle of the interface is obtained
by using the formulae tan-1[(2/3) tanφ], which is about
21°. Totally, 7303 nodes and 2352 finite elements are
arranged in the dynamic model. The north-south com-
ponent of the rock base acceleration measured at 1995
Kobe earthquake, which is scale-adjusted to 0.3g based
on the measurement at Kobe meteorological observatory,
is used as the input bed rock motion. It shows the high
horizontal ground acceleration with a short period as
shown in Fig. 6.
3.2 Boundary conditions
The boundary between the soil deposit and the bedrock
(see Table 1) is simply assumed to be fixed and would
act as the bottom boundary of the analysis domain at
which the earthquake motion is imposed. The ground
surface is assumed to be flat and free of loads. The un-
derground water level is assumed to be at the level of
ground surface so that the entire soil is saturated.
The soil deposit at the far fields should be assumed as
the boundary of free shaking. In the shaking table tests
of soil-structures, a laminar shear box may be used to
simulate the quasi-far-field boundary (Towhata 2008).
Here, quasi-far-field elements with a length of 10 m are
placed at each extreme side of the analysis domain as
illustrated in Fig. 5(b). Both stiffness and unit weight of
far-field elements are increased 100 times fictitiously in
order to computationally realize the free-excitation
which is not substantially affected by the motion of the
target domain of high nonlinear interaction with the
structures. It should be noted that these fictitious pa-
rameters of these two boundary columns result in the
same free natural frequencies and dynamic properties as
those of the pure soil layers without underground struc-
tures. As the far-field mode of seismic motion is lami-
nated simple shear, the horizontal length of these
boundaries is selected approximately half of the domain
height so that the bending deformational mode would
not appear. In addition, confinement independent soil
elements are used in the quasi-far-field zone in order to
prevent the edge collapse in analysis. Consequently, this
boundary simply makes the horizontal displacements at
both right and left sides equal to each other, which is
similar to the case of laminar shear boxes widely used in
the shaking table tests.

Table 1 Material properties used.

Non-liquefiable clayey layer* Reinforced Concrete Interface**
Initial stiffness 105 MPa Concrete strength 24 MPa Normal modulus 108 kPa/m
Poisson ratio 0.20 Unit weight 24 kN/m3 Shear modulus 103 kPa/m
3
Dry unit weight 16 kN/m Poisson ratio 0.18 Friction angle 21º
5
Friction angle 40 º Steel Young modulus 2.0 x10 MPa Cohesion 0
Cohesion 100 kPa Steel yield stress 240 MPa
Relative density 75 %
*) These properties are used for hard base bed of foundation for all parametric studies to numerically avoid liquefaction.
**) Simple frictional model used.
444 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009

y
x

(a) Soil-structure layout

(b) Finite element discretization

Fig. 5 Entire soil-structure system properties.

Fig. 6 Input earthquake motion and its response spectrum.

Several trial analyses were carried out to determine
the size of the analysis domain. Finally, a relatively
large analyzed domain (200 m) was specified so as to
make the wave reflection negligible.
3.3 Analysis procedure
As the seismic analysis of the soil-structure system re-
quires an initial stress field of the static equilibrium be-
fore dynamic earthquake loads (Liu and Song 2005;
Maekawa et al. 2003), the static analysis of drained
foundation was firstly performed. In general, repro-
duced static soil stress fields might depend on the proc-
ess of construction works. But, it was confirmed by the
preliminary analyses that the initial earth pressure on the
RC ducts has minor influence on the damage of struc-
tures when high nonlinearity is induced to both soil and
structure under large ground motions.
For investigating the effect of soil liquefaction on the
structural damages, several models with and without the
ducts are analyzed in both drained and undrained states
of pore water with initial shear stiffness (G0) varying
from 12 MPa to 182 MPa as shown in Table 2. The
structure is assumed to be located inside the soil at a
depth of 4 m without any change in the mesh property
of the remaining soil elements. Then, the standard case
of the RC duct is analyzed by using linear elastic ele-
ments with a Young modulus of 30,000 MPa, Poisson’s
ratio of 0.20 and the unit weight of 24 kN/m3, keeping
the sandy soil property constant (G0= 35 MPa, Dr=32 %
and dry unit weight=14 kN/m3). Finally, a parametric
study is performed to investigate the effect of RC
nonlinear properties. Wholly, 34 different cases are
studied with different reinforcement ratios, different
surrounding soil stiffness and two different yield
strengths of steel. All these cases are analyzed in both
drained and saturated soil conditions.
It should be pointed out that the authors intentionally
implement fully undrained conditions for saturated soil
 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009 445

Table 2 Material property for liquefiable sandy layer. 4. Nonlinear interaction and simulation
NSPT G0* (MPa) K0 (MPa) Dr** (%)
4.1 Liquefaction induced damage to under-
1 12 16 25 ground RC ducts
3 29 39 28
5 44 59 32 Figure 7 shows the shear response versus the Standard
Penetration Test N-value (denoted by SPT) of soil layer
10 76 101 40
for both cases of plain soil and with RC ducts in drained
15 105 140 48 and undrained conditions. The values of maximum
20 132 176 56 shear deformation are also shown in Table 3 for further
25 158 211 65 discussion. Here, the shear deformation of the duct is
30 182 243 75 defined as the drift angle between the upper and bottom
slabs. It can be observed that current design approach
*) G0 is defined according to the empirical formulae as
based on the free-field ground displacement can predict
G0 = 12(N-value) 0.8.
the deformational demand on the underground RC ducts
**) The value of Dr represents how tightly the soil is
which are embedded in somehow stiffer soil mediums.
deposited, and it gives the risk of liquefaction.
In soft layers of soil which consists of loosely deposited
sand, the required ductility of the duct in structural de-
signs becomes smaller than that of the single soil me-
elements which may bring about earlier liquefaction dium. This is attributed to the reduced stiffness of the
than reality, because the aim of this study is to investi- soil foundation compared to that of the RC duct (Fig.
gate the magnitude of damage to RC ducts by liquefi- 7(a)). More or less, the required ductility of the struc-
able soils. As a matter of fact, this assumption is not far ture rises according to the reduction of soil stiffness
from the facts that the required time for drainage of a provided that no liquefaction takes place.
several-meter-thick sand layer is 10-30 minutes, which Besides, liquefaction may significantly bring about
is much longer than the duration time of earthquake increased soil deformation which indicates large soil
loading (Towhata 2008). strains associated with greatly reduced shear stiffness.
Accordingly, the ductility demand on the RC duct under
liquefaction becomes larger than those of the drained

Rebar yielded Elastic

Concrete cracked

Principal tensile strain state of tunnel at maximum shear deformation

(a) Drained condition

Rebar yielded Elastic

Concrete cracked

Principal tensile strain state of tunnel at maximum shear deformation

(b) Undrained condition
Fig. 7 Effect of soil stiffness on the maximum response of averaged shear deformation.
446 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009

Table 3 Maximum responses of averaged shear liquefied soil. This trend is hardly taken into account in
deformation. current simplified design procedures with linear soil-
spring even though its stiffness is made degenerated.
Maximum Shear Deformation The pore pressure ratio (the excessive pore pressure
SPT normalized by the initial effective overburden pressure)
N- Drained Condition Undrained Condition above and below the duct are shown in Fig. 8. It can be
value seen that the liquefaction below the duct is compara-
Free with Tun- Free with Tun-
Field nel Field nel tively minor. This is consistent with the site observation
1 1.20 0.55 1.98 0.31 of the past earthquakes. As a matter of fact, the exces-
sive uplift flotation of the underground lightweight
3 0.39 0.30 1.23 0.41 structures would cause larger shear deformation of sur-
5 0.27 0.25 0.86 0.41 rounding soil, which may occasionally cause the de-
10 0.15 0.18 0.45 0.33 crease in pore pressure. Under larger shear deformation,
normal sand tends to dilate, which shall lead to the low-
15 0.11 0.14 0.36 0.28 ering of pore pressure (Liu and Song 2005), that is to
20 0.07 0.11 0.23 0.22 say, the cyclic mobility.
25 0.06 0.10 0.10 0.16 Figure 9 shows the comparison of acceleration re-
sponses at the ground surface. It can be observed that
30 0.05 0.09 0.05 0.12 the acceleration in liquefiable soil is much smaller than
that of the drained dry soil. The earthquake energy
would be damped out by dramatically reduced shear
case if N-value is more than 5. It means that the magni- stiffness of soil skeleton and accompanying larger de-
fied soil deformation by liquefaction and the associated formation which occur at the time of liquefaction. The
reduced stiffness of the foundation may not be balanced acceleration response is not affected by the presence of
but the former factor gets extensive in view of the RC the structure.
structural damage. The current designs tend to require
further enhanced ductility in liquefiable foundation, 4.2 Effect of RC nonlinearity
which causes denser arrangement of web reinforcement The horizontal and vertical dynamic displacements of
and tends to accompany some difficulty of construction. the linear elastic duct and the highly nonlinear RC are
Anyhow, this trend is consistent with the analytical re- compared as shown in Fig. 10 and Fig. 11, respectively.
sults as discussed above. The duct exhibits some settlement during the ground
On the contrary, if the soil is much loosely deposited motions in the drained condition, while liquefied soil
(N-value less than 5), we have dramatically decreased pushes the underground duct upward significantly as
deformations compared to the case of drained founda- observed in the past earthquakes and some laboratory
tion as shown in Fig. 7(b). The practical point of impor- tests (Towhata 2008). Then, some countermeasures have
tance is that the maximum shear deformation finally been discussed to reduce the uplift of underground
declines according to the greatly reduced stiffness of the structures in liquefiable soils.

Transient zone Fully liquefied Transient zone Fully liquefied

(a) 1.5 m above the top slab (b) 2.5 m below the bottom slab
Fig. 8 Excessive pore pressures in soil foundation.
 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009 447

(a) Dry sandy ground (b) Saturated sandy ground

(c) With underground structure

Fig. 9 Acceleration response of the ground surface.

(a) Drained condition

(b) Undrained condition

Fig. 10 Horizontal displacement responses of the linear and nonlinear ducts.

The overall rigid body motion of the structure is
hardly affected no matter how nonlinearly the under-
ground ducts may behave. On the contrary, the averaged
shear deformation of the linear duct is much smaller
than that of the nonlinear case as demonstrated in Fig.
12. RC nonlinearities cause the duct to accommodate
more shear deformation. Besides, as the linear duct re-
mains stiff during the earthquake motion, the adjacent
unsaturated dry soil movement is restricted by the struc-
ture, and the shear stiffness degradation of the soil is
limited (Huo et al. 2005). In addition, the nominal shear
stress of the center column, which has much to do with
the axial load carrying capacity, demonstrates higher
magnitude for the case of nonlinearity rather than the
linear case of the structure as shown in Fig. 13. It can be
said that nonlinear interaction analysis is indispensable
for safety design of underground structures.
448 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009

(a) Drained condition (b) Undrained condition

(c) Uplift mode of deformation at soil liquefaction (5 times magnified displacement of analysis)

(d) Experimental uplift mode of deformation by Towhata (2008).

Fig. 11 Vertical displacement responses of the linear and nonlinear ducts.

(a) Drained condition – Linear structure (b) Drained condition – Nonlinear structure

(c) Undrained condition – Linear structure (d) Undrained condition – Nonlinear structure
Fig. 12 Shear deformation responses of the linear and nonlinear ducts.
 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009 449

(a) Drained condition – Linear structure (b) Drained condition – Nonlinear structure

(c) Undrained condition – Linear structure (d) Undrained condition – Nonlinear structure
Fig. 13 Nominal shear stress response of the center column for linear and nonlinear cases.

(a) Drained condition (b) Undrained condition

Fig. 14 Horizontal displacement response of the duct with increased yield strength of rebars.

(a) Drained condition (b) Undrained condition

Fig. 15 Vertical displacement response of the duct with increased yield strength of rebars.

4.3 Yield strength of reinforcement
Cracking of concrete and plasticity of steel reinforce-
ment are two major sources of RC nonlinearities. In-
creased yield strength of reinforcement would reduce
nonlinearity of RC elements without varying initial
stiffness. Figure 14 and Fig. 15 show the horizontal and
vertical displacements of the duct respectively accord-
ing to the higher yield strength of 400 MPa. By compar-
ing these with the case of 240 MPa (Figs. 10 and 11), it
can be understood that the rigid body motion of the
structure is not much affected by the steel nonlinearity.
However, the high strength reinforcement would lead to
slightly reduced shear response of the duct as shown in
Fig. 16. By minimizing the nonlinearity of reinforce-
ment, the duct response would come closer to the linear
one (Fig. 12(a, c)) although the duct still suffers from
larger shear deformations especially in the dry soil.
Figure 17 illustrates the crack pattern of the structure
450 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009

(a) Drained condition (b) Undrained condition

Fig. 16 Shear deformation response of the duct with increased yield strength of rebars.

(a) Drained condition (b) Undrained condition

Fig. 17 Crack pattern of the duct after the earthquake motion.

with increased yield strength at the end of dynamic Table 4 Variation of reinforcement ratio for different
analysis. It can be observed that large cracks are con- cases.
centrated at the corners and both ends of the column.
Since most of the major cracks occur at the early stages Reinforcement ratio Standard Case B Case C
when the acceleration is large and the soil was not yet Slab & Wall 1.10% 2.20% 0.55%
fully liquefied, the crack pattern in the dry and saturated Long.
soil is almost similar. There are small additional cracks reinforcement Column 1.60% 3.20% 0.80%
at the internal side of the floor slab for the case of lique- Web
faction, because of the “squeezing” pressure of liquefied All elements 0.20% 0.20% 0.20%
reinforcement
soil beneath the structure to push it upward. This
mechanism is discussed in (Koseki et al. 1997a; Ha-
shash et al. 2001; Liu and Song 2006) as the main cause tility of the column compression members can be sim-
for the uplift of underground structures. ply estimated (JSCE 2002) by,

4.4 Effect of Reinforcement ratio Vu ⋅ H

The main reinforcement ratio is a key parameter to con-
trol the performance of structures practically at the de-
sign stage. Thus, parametric analyses with different re-
inforcement ratios (Table 4) are performed to investi-
gate the responses and damages of the structure associ-
ated with the ductility and the shear capacity. The hori-
zontal and vertical displacement responses are not influ-
enced by the amount of longitudinal reinforcement as
shown in Fig. 18 and Fig. 19.
By comparing Fig. 20 with Fig. 12(b, d), it can be
said that the imposed structural shear deformation from
soil foundation is reduced to about 70% of the standard
case by placing two time large amount of reinforcement,
and also increases about 20% by placing the half
amount of reinforcement. In view of the structural dam-
age control, the effectiveness of the reinforcement ratio
is found to be comparable no matter how much likely
the soil liquefaction occurs. On the other hand, the duc-
φ= (11)
Mu
where Vu is the shear capacity of the RC member, H is
the shear span (half of the column height in the case of
RC duct), Mu is the flexural capacity. When the ductility
index φ is greater than 1.3~1.5, practically sufficient
inelastic deformation is realized. If it is less than unity,
shear failure may take place before yielding of main
reinforcement and brittle failure will be initiated. As the
main reinforcement ratio is almost proportional to Mu, it
must be noted that the increased main reinforcement
may decline the RC ductility as well. In considering the
sensitivity of reinforcement ratio on the structural duc-
tility (output) and imposed deformation (input) from
soil foundation, generally, heavy reinforcement is
thought not to be effective.
Figure 21 shows the effect of soil stiffness on the
maximum shear deformation response of the duct in
 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009 451

(a) Drained condition (b) Undrained condition

Fig. 18 Horizontal displacement response of the heavily reinforced duct.

(a) Drained condition (b) Undrained condition

Fig. 19 Vertical displacement response of the heavily reinforced duct.

(a) Drained condition – Case B (b) Drained condition – Case C

(c) Undrained condition – Case B (d) Undrained condition – Case C

Fig. 20 Shear deformation responses of the ducts with different reinforcement ratios.

Case B and Case C (Table 4). By comparing Fig. 21(a,
b) with Fig. 7(a, b), it can be understood that the re-
duced amount of main reinforcement may greatly in-
crease the ductility demand of the duct especially in the
softer layers of soil deposit which structural properties
become more effective and predominant. Liquefaction,
however, deteriorates the surrounding soil stiffness and
thus the deformational demand on the RC duct conse-
quently decreases as shown in Fig. 21(c, d). The maxi-
mum shear deformation of the duct in the case C in-
creases when the soil is extremely soft (N-value = 1)
and full saturated as shown in Fig. 21(d), which is in
contrast with Figs. 21(c) and 7(b). It is due to high plas-
ticity of yielded reinforcement having low amount of
reinforcement in the case C before soil is fully liquefied.
However, it is still smaller than that of the same duct
under the drained condition.
If this highly nonlinear interaction would not be con-
452 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009

(a) Drained condition – Case B (b) Drained condition – Case C

(c) Undrained condition – Case B (d) Undrained condition – Case C

Fig. 21 Effect of duct reinforcement ratio on the maximum response of averaged shear deformation: Case B – Heavily
reinforced, Case C – Lightly reinforced

Fig. 22 Hachinohe record and its response spectrum.

sidered for the risk of liquefaction, over-reinforcement
of web steel, which currently matters in practice, is not
unavoidable.
4.5 Effect of Input Excitation
In order to investigate the characteristics of input seis-
mic excitations, another earthquake record is tried. The
north-south component recorded at Hachinohe City dur-
ing the Tokachi-oki earthquake of 1968, which is known
to have its elongated period, is selected and scaled to
0.3g as shown in Fig. 22. Then, the same procedure
which was applied in section 4.1 is followed with the
Hachinohe record. The results are shown in Fig. 23,
which can be compared by Fig. 7. Besides, in order to
make the discussion more clear, the relative shear de-
formation is employed, which is defined as the maxi-
mum shear deformation of the structure normalized by
the maximum distortion of free-field ground response
and can be expressed (Wang 1993) as,
γ structure
R= (12)
γ free−field
where γ is the maximum angular drift of the rectangular
RC duct. The relative shear deformation of the duct in
drained and undrained conditions is given in Table 5
and can be compared for different input motions. By
contrast, the conclusion made in section 4.1 is found to
be valid for different kinds of earthquake properties as
long as the surrounding soil fully liquefies.
5. Conclusions
Strong coupling effect of highly nonlinear liquefiable
soil and inelastic underground RC was investigated, and
consequent damage induced to reinforced concrete was
 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009 453

(a) Drained condition (b) Undrained condition

Fig. 23 Effect of duct reinforcement ratio on the maximum response of averaged shear deformation.

focused in consideration of rising pore water pressure

Table 5 The relative shear deformation of the duct in
and the drastic reduction of shear stiffness of soil parti-
different cases.
cle skeleton. The parametric study brings about the fol-
lowing conclusions in practice. Relative shear deformation (%)
1) In case of the drained state of soil foundation with- SPT
out liquefaction, the smaller stiffness of soil layer is, N- Kobe Record Hachinohe Record
the larger ductility of the underground RC duct is value
demanded in structural design. Consequently, large Drained Undrained Drained Undrained
amount of web reinforcement is being specified in 1 0.46 0.16 0.69 0.19
practical design.
3 0.77 0.33 0.93 0.26
2) In case of the liquefiable soil, however, the structural
damage is found to be reduced owing to the signifi- 5 0.94 0.48 0.78 0.38
cantly reduced shear stiffness of soil foundation, 10 1.20 0.74 1.07 0.64
even though the larger soil deformation is realized.
15 1.29 0.77 1.48 0.78
Thus, the ductility demand to the structure is sharply
lighten due to the effect of soil seismic isolation un- 20 1.53 0.98 1.71 0.86
der very loosely deposited foundation. If this high 25 1.67 1.56 1.43 1.12
nonlinearity would not be considered in practice, so
30 1.85 2.54 1.52 1.55
heavy reinforcement is inevitably forced which
causes difficulty of concreting works. Of course, it
should be pointed out that some countermeasures
such as sheet piling and controlling the weight of case of unsaturated dry soil.
tunnels should be considered to reduce the rigid
body motion of underground structures in liquefiable References
soils as being conducted in practice. The shear de- An, X., Shawky, A. A. and Maekawa, K. (1997). “The
formation and the uplift are in relation of trade-off as collapse mechanism of a subway station during the
simulated in this study. Great Hanshin earthquake.” Cement and Concrete
3) The uplift rigid body motion caused by liquefaction Composite, 19, 241-257.
is hardly affected by the nonlinearity of underground Biot, M. A. (1962). “Mechanics of deformation and
RC. In other words, the uplift stability cannot be acoustic propagation in porous media.” Journal of
controlled well by the structural design except for Applied Physics, 33(4), 1482-1498.
self-weight. In fact, lots of underground ducts were Hall, W. J. and O’Rourke, T. D. (1991). “Seismic
reported to drift upwards without accompanying any behavior and vulnerability of pipelines.” Proc. Third
structural damage. These past experiences are con- US Conf. Lifeline Earthquake Engineering, 761-773.
firmed by the nonlinear coupling analysis as well. Hashash, Y. M. A., Hook, J. J., Schmidt, B. and Yao, J.
4) Reinforcement with higher yield strength may re- (2001). “Seismic design and analysis of underground
duce the deformation and the damage to RC ducts. structures.” Tunneling Underground Space
But its effectiveness is practically negligibly small. Technology, 16(4), 247-293.
5) Increased main reinforcement may reduce the mean Huo, H., Bobet, A., Fernandez, G. and Ramirez, J.
deformation and material damages to the under- (2005). “Load transfer mechanisms between
ground ducts with similar magnitude for both underground structure and surrounding ground:
drained and undrained states of soil foundation. If Evaluation of the failure of the Daikai station.”
the soil is loosely deposited with high risk of lique- Journal of Geoenvironmental Engineering, 131(12),
faction, the structural damage may be less than the 1522-1533.
454 M. R. Okhovat, F. Shang and K. Maekawa / Journal of Advanced Concrete Technology Vol. 7, No. 3, 439-454, 2009
Japan Society of Civil Engineers (2002).
“Recommended code and manual for seismic
performance verification of out-door important
structures of nuclear power plants.” Concrete Library
International.
Kato, B. (1979). “Mechanical properties of steel under
load cycles idealizing seismic action.” CEB Bulletin
D’Information, 131, 7-27.
Kimura, T., Takemura, J., Hiro-oka, A. and Okamura, M.
(1995). “Countermeasures against liquefaction of
sand deposits with structures.” Proc. First
International Conference on Earthquake
Geotechnical Engineering, 1203-1224, Tokyo, Japan.
Koseki, J., Matsuo, O., Koga, Y. (1997a). “Uplift
behavior of underground structures caused by
liquefaction of surrounding soil during earthquake.”
Soils and Foundations, 37(1), 97-108.
Koseki, J., Matsuo, O., Ninomiya, Y., Yoshida, T.
(1997b). “Uplift of sewer manholes during the 1993
Kushiro-Oki earthquake.” Soils and Foundations,
37(1), 109-121.
Li, B., Maekawa, K. and Okamura, H. (1989). “Contact
density model for stress transfer across crack in
concrete.” Journal of Faculty of Engineering,
University of Tokyo (B), 40(1), 9-52.
Liu, H. and Song, E. (2005). “Seismic Response of
Large Underground Structures in Liquefiable Soils
Subjected to Horizontal and Vertical Earthquake
Excitations.” Computers Geotechnics, 32, 223-244.
Liu, H., Song, E. (2006). “Working mechanism of cutoff
walls in reducing uplift of large underground
structures induced by soil liquefaction.” Computers
Geotechnics, 33, 209-221.
Maekawa, K., Irawan, P. and Okamura, H. (1997).
“Path-dependent three-dimensional constitutive laws
of reinforced concrete: Formulation and experimental
verifications.” Structural Engineering and Mechanics,
5(6), 743-54.
Maekawa, K. and An, X. (2000). “Shear failure and
ductility of RC columns after yielding of main
reinforcement.” Engineering Fracture Mechanics, 65,
335-368.
Maekawa, K., Pimanmas, A. and Okamura, H. (2003).
“Nonlinear Mechanics of Reinforced Concrete.” Spon
Press, London.
Maekawa, K., Fukuura, N. and Soltani, M. (2008).
“Path-dependent high cycle fatigue modeling of joint
interfaces in structural concrete.” Journal of
Advanced Concrete Technology, 6(1), 227-242.
Maki, T. and Mutsuyoshi, K. (2004). “Seismic behavior
of reinforced concrete pile under ground.” Journal of
Advanced Concrete Technology, 2(1), 37-47.
Maki, T., Maekawa, K. and Mutsuyoshi, H. (2005). “RC
pile-soil interaction analysis using a 3D-finite
element method with fiber theory-based beam
elements.” Earthquake Engineering and Structural
Dynamics, 99, 1-26.
Masing, G. (1926). “Eigenspannungen and Verfestigung
Beim Messing.” Proc. of Second International
Congress of Applied Mechanics, 332, Zurich.
Nam S.H., Song H.W., Byun K.J. and Maekawa K.
(2006). “Seismic analysis of underground reinforced
concrete structures considering elasto-plastic
interface element with thickness.” Engineering
Structures, 28, 1122-1131.
O’Rourke, T. D., Goh, S. H., Menkiti, C. O. and Mair, R.
J. (2001). “Highway tunnel performance during the
1999 Duzce earthquake.” Proc. 15th International
Conference on Soil Mechanics and Geotechnical
Engineering.
Tamari, Y. and Towhata, I. (2003). “Seismic soil-
structure interaction of cross sections of flexible
underground structures subjected to soil
liquefaction.” Soils and Foundations, 43(2), 69-87.
Towhata, I. and Ishihara, K. (1985). “Modeling soil
behaviors under principal stress axes rotation.” 5th Int.
Conf. on Numerical Method in Geomechanics,
Nagoya, 523-530.
Towhata, I. (2008). “Geotechnical Earthquake
Engineering.” Springer, Germany.
Wang, J. N. (1993). “Seismic design of tunnels: a state-
of-the-art approach.” Monograph, monograph 7,
Parsons, Brinckerhoff, Quade and Douglas Inc, New
York.
Wang, W. L. , Wang, T. T. , Su, J. J., Lin, C. H., Seng, C.
R. and Huang, T. H. (2001). “Assessment of damage
in mountain tunnels due to the Taiwan Chi-Chi
earthquake.” Tunneling Underground Space
Technology, 16(3), 133-150.