J. Materials, Conc. Struct., Pavements, JSCE, No.

571/V-36, 251-267, 1997 August

FAILURE ANALYSIS OF UNDERGROUND RC

FRAME SUBJECTED TO SEISMIC ACTIONS

Xuehui AN1 and Koichi MAEKAWA2

1Member of JSCE, Dr. of Eng., Seismic Eng. Dept., Tokyo Electric Power Services Co., Ltd. (Higashi-Ueno
3-3-3, Taito-ku, Tokyo 110, Japan)
2 Member of JSCE, Dr. of Eng., Professor, Department of Civil Engineering, The University of Tokyo,
(Hongo 7-3-1, Bunkyo-ku, Tokyo 113, Japan)

This paper presents failure analysis of underground RC frames. The Hanshin Great Earthquake in 1995
brought serious damage to RC frames for subway stations. For studying the collapse mechanism of
underground RC, seismic response of a subway station is simulated in using FEM program WCOMD-SJ of
two-dimension based on the path dependent RC smeared crack model, soil foundation and interfacial
models. The shear failure of intermediate vertical columns is found to be the major cause of the structural
collapse. Further, parametric study on reinforcement ratio and foundation properties is performed for
investigating seismic resistant performance for underground RC culverts.

Key Words: FEM, dynamic analysis, underground structure, RC, soil-structure interaction

1. INTRODUCTION employed, and the constitutive law of reinforcing

bars is composed. The concrete model consists of
Hanshin Great Earthquake on January 17, 1995 tension stiffening, compression and shear transfer
brought about disastrous collapse of reinforced models. These models are given as the relationship
concrete structures including some underground between average stress and average strain in
subway stations in Kobe city. As catastrophic failure reinforced concrete. The crack spacing, or density,
of underground structures was firstly experienced, it and diameter of reinforcing bars have negligible
is much required to clarify the collapse mechanism of effect on the spatially average stress-strain relation
these underground frames and to update the latest defined on RC control volume, as shown in
seismic design if necessary. Fig.1a3)4. The continuum damage model of concrete
In this study, underground RC frame with double encompasses the reduction of compressive capacity
decks for a subway station is targeted to study the of cracked concrete in relation to the strain normal to
failure mode and its mechanism, and to examine the the crack5.
computational tool for further study of earthquake A path-dependent constitutive model for soil is
resistant design. The observed damage reported in indispensable for dealing with kinematic interaction
the literature1)2) is compared with analytical results of RC-soil entire system under strong seismic loads,
for verification on which the enhanced seismic as shown in Fig. lb6. Here, Ohsaki's model's defines
resistant design will be based in future. the formula for envelope to express the nonlinear
relation of the shear stress-strain for soil as well as
internal loop with Masing's rule. In addition,
2. COMPUTATION TOOL
separation and sliding between soil and structure are
In this paper, two dimensional FEM program taken into account along the interfacial zone as
WCOMD-SJ is used for analyzing coupled reinforced shown in Fig. 1c.
concrete and soil structures under seismic excitation. The full path-dependent constitutive models were
Here, the constitutive models for reinforced concrete, integrated in the scheme of Newmark step-by-step
soil and the interface between soil and RC are direct integration of both time and strain histories. In .
installed. this way, the dynamic nonlinear response of RC-soil
In the model for reinforced concrete, the multi- structure can be computed within the versatile
directional smeared crack model of concrete is computational scheme.

251
 3. COMPUTATION TARGET
One of the most typical damages experienced in
Kobe in 1995 is the failure of intermediate columns
in underground RC box culvert. Reported are 264 RC
columns in the subway line. Among them, 139
columns were damaged and 118 of them were fatally
collapsed. In this study, one typical section of a
station is chosen for the collapse simulation and
evaluation of seismic resistant performance. The
scenario of failure and the final collapse of the
structures are of great interest to the authors.

(1) Failure observation

Diagonal shear cracks were clearly observed in
the intermediate columns of the station. Heavy
damage was identified at the upper deck columns
located at the middle of the underground frame
spans, and subsidence occurred to the top slab as the
column could not carry the dead load after shear
failure occurred (Fig. 2a). The maximum of
subsidence reached 5cm by observation. But for the
lower deck columns, the damage was not so serious
as to finally fail. Only few diagonal shear cracks in
Multi-system LOCAL RESPONSE E7 MEAN RESPONSE the column were seen (Fig. 2b). The typical failure of
this RC structure is the collapse of intermediate
Fig. la Constitutive model of reinforced concrete3f columns accompanying shear cracks and the damage
concentrated into the upstairs columns of the double
deck frames.

(2) Layout and structural details

Fig. 3a gives the shape and dimension of the
station section consisting of RC membersl2. There
are two floors in this RC box type culvert and in each
floor one RC column was placed. For this section, the
spacing of columns is Sm. The RC underground
culvert has an outer dimension of 16.68m width and
13. 25m height. The wall thickness is 0. 8m in the first
floor and 1.Om in the second floor. The thickness of
the top slab is 1.Om and 1.3m for the bottom slab. The
Fig. 1b Computed shear stress-strain relation for soi16f
thickness of middle slab is 0.6m. The intermediate
column has a cross section of 0. 7m x 1.4m for the
upper level and 0.6m x 1.Sm for lower level with
average reinforcement ratio 6.0%. The columns are
idealized being completely fixed to the slabs both in
reality and in computation. The station is inside the
earth with Sm overlay of soil. The underground
system for dynamic simulation is composed of both
the 2D RC frame and the surrounding soil.
Fig. 3b shows reinforcement arrangement of the
station section concerned. The intermediate columns,
which suffered from serious damage in the
earthquake, has the features of reinforcement
arrangement as heavily reinforced in the longitudinal
direction and few web reinforcement placed. The
Fig. 1c Normal and shear relations for RC/soil interface model6f
longitudinal reinforcement ratio is 5. 1% for the upper
252
 (Space of columns=5m)
Fig. 3a Dimension of the target RC frame.

Fig. 2a Collapse of a column at the upper level.

(Steel bar number/m)

Flg. 3b Reinforcement arrangement of the target RC frame.

seismic actions are not the most expected inputs on

the foundation rock but mere records measured on
the ground surfaces as a result of magnification by
Fig. 2b Shear cracks of a column at the lower level. the effect of propagation path of soil foundation.
Since the authors' capability is so limited as to
column and 5.7% for the lower column. The web produce the most reliable seismicity at the
construction site, magnified ground surface waves
reinforcement ratio is only 0.075%-0. 15%. This are used as a substitute of the base rock accelerogram.
amount can not substantially serve as shear It means difficulty to investigate the real cause of
reinforcement. Fig. 5 and Table 1 show the details of disaster which actually occurred. But, it may be
the surrounding soil layers. In the computation, the
possible to investigate the intrinsic structural
acceleration seismic wave was applied at the base of
performances under severe seismic actions and to
layer 6 as the stiffness of layer 7 is assumed large examine applicability of coupled structural failure
enough to be regarded as a rigid engineering analysis and soil foundation. Then, this paper is
boundary for computation. directed in a great deal to structural aspects and check
of computational capability. As a future stage, joint
(3) Input ground acceleration waves works with seismic field of study should follow this
Earthquake motions that are used for input were structural oriented investigation here.
recorded at Kobe meteorological observatory and For seismic performance check of existing
other sites as shown in Fig. 4 and Fig. 29 in the later structures, most expected seismic actions and
section. Both horizontal and vertical acceleration material properties already developed should be
records are used for simulating dynamic response of defined since the purpose of verification is directed to
soil-RC box system having longer dimension along actually attained performances.
an axis of subway line. It must be noted that used
253
Fig. 4a Horizontal acceleration of earthquake wave used. Fig. 4b Vertical acceleration of earthquake wave used.

Fig. 5 Soil profile of surrounding foundation.

Table 1 Characteristic variation of surrounding soils.

254
 Total Node: 2705
Total Element: 888

Fig. 6 Central part of the finite element mesh of soil-RC used in FEM analysis.

(4) FEM mesh for RC frame and foundation (1) Inelasticity of the whole RC structure8}
The finite element discretization is shown in In order to indicate the damage level of the entire
Fig. S,6. Higher order isoparametric elements are RC structure in terms of leakage resistance against
used to analyze the target section. In flexure, only one ground water after earthquake, crack width oriented
layer of higher order elements is necessary and inelastic output in time domain is needed. The first
sufficient but in shear, several layers are required strain invariant denoted by (Ii) is closely associated
since shear strain develops nonlinearly over the with the crack occurrence and expansion of the in-
thickness of members unlike flexural normal plane element (volumetric change of the element)8.
strains8. Then, three layers are placed at the The mean strain invariant denoted by (I) can be
intermediate columns and mesh sensitivity and defined as the spatial average value of (Ii) for all RC
convergence were checked in advance. As the elements. The value is equal to zero in the case of
thickness of RC outer frame and the column are elastic shear behavior as no volumetric change and
different, the stiffness changes sharply near the joint no residual deformation exist under pure elastic shear
plane. In order to account for the incontinuous deformation. Hence the mean strain invariant (I),
deformation rooted in the joint area, RC joint called as inelastic strain, can be adopted to represent
elements are placed between the column and slab. the magnitude of the damage of reinforced concrete.
The two extreme sides of this whole analysis domain The value of (I) can be calculated as follows.
have the mixed artificial boundary elements to
simulate the far field of soil layers8.
I=I1(x,y)dxdy/A, li=e1+e2/2 (1)

4. COLLAPSE SIMULATION OF where, Ei and E2are the 2-D principal strains at (x,y)
SUBWAY STATION and A is the total area of the RC in-plane elements.
The computation of coupled underground RC and Fig. 7 shows the inelastic strain (I) of the target
the surrounding soil system under Kobe earthquake RC frames in time domain. This index is used to
wave was conducted by WCOMD-SJ3. The qualitatively present how much damage to the RC
following results give the behavior of the system structure and how much deformation resided after the
under seismic load and the induced forces to the dynamic action. In this figure, we can see great
intermediate column.
255
Fig. 7 Inelastic strain representing damage of RC in time domain.

Fig9a Crack pattern of structure just before failure.

Fig. 8a Deformationprofile of RC-soilsystemjust before failure.

Fig. 9b Cracks developed at the failure.

(3) Crack pattern of the RC frame

In order to confirm the location of failure and the
failure mode, the crack pattern of the structure just
before the failure is shown in Fig. 9a, and Fig. 9b
indicates the cracks which are propagating during the
failure process. From Fig. 9a it can be seen that the
large cracks are concentrated into the upper column,
on the contrary, the lower column and the other parts
of the frame have just small cracks. All cracks shown
in Fig. 9b occur in the last step of failure. We can see
Fig. 8b Deformation profile of station at failure. very clearly that only large shear cracks developed in
the upper column, crossed the section of this column
increase in the inelastic strain at 7. 32 sec and the and caused the failure. These crack patterns with the
structure reached failure. observed cracks shown in Fig. 2 look fairly realistic.

(2) Dynamic response of the underground RC (4) Internal stresses in the intermediate column
Fig. 8a gives the magnified deformation profile of Computational results and observation show that
the RC-soil system at the maximum response just the major collapse may occur at the intermediate
before the failure of RC in shear. The maximum column. It is important to discuss the induced forces
deformation at the top column is about 0.6% by and ductility of the internal columns. Fig. 10 shows
average shear. And Fig. 8b shows the deformational the internal nominal stresses (axial compression and
shear divided by the cross-sectional area of members)
profile of the station at the failure. It can be seen that
the deformation is concentrated into the upper and the ductility of the column for both upper and
intermediate column, which was the actual failure inwer nartc
location of the entire RC structural system.
256
 Fig10a Shear stress-displacement relationship for columns. Fig. 10b Compression stress variation of the columns.

Fig. 10a shows the relation between the nominal The aim of dynamic analysis of the RC
shear stress (shear force normalized by the cross- underground structure is to study the rational seismic
sectional area of a member) and the relative resistant design method based on the knowledge of
displacement between top and bottom of the column. the collapse mechanism investigated. As most of the
The relative displacement is also normalized by the damage of underground RC is rooted in the diagonal
height of the column. The column fails with the shear failure of the intermediate column9, the shear
maximum normalized displacement 0. 7% and behavior of the RC columns will be checked with the
maximum shear stress 18kgf/cm2. The lower column shear capacity equation of the JSCE code10 as,
undertakes similar nominal stress with slightly
V=VCd+Vsd (2)
smaller shear displacement and no failure takes
place. where, V is the shear capacity of the RC member; Vsd
Fig. 10b shows the variation of nominal is the shear force carried by concrete and longitudinal
compressive stress in the columns. It can be seen that reinforcement; Vsd is the shear force carried by the
the compressive stress varies between 20kgf/cm2 and web reinforcement. Vsd and Vsd can be evaluated by
70 kgf/cm2 due to the up and down motion of the the following formulas.
earthquake. The maximum compressive stress is less
than 100kgf/cm2 and far away from the compression dOC(pIl3(f(di
capacity of the RC columns.
VsdocPw (3)
(5) Collapse mechanism study
According to FEM simulation of the failure where, pl is the reinforcement ratio of longitudinal
mechanism, it is considered that the RC column bars; pw is the reinforcement ratio of web steel; f is
would lose axial load carrying capacity after the the compressive strength of concrete and d is the
occurrence of the localized diagonal shear cracks, effective depth of RC member.
and sudden failure of the outer frame would be From these equations, the shear capacity of the
followed. Some cracks would be introduced at the upper column in underground RC can be calculated.
lower column and the corner of the RC box but with If no axial force is considered the shear strength of
lower damage level. this column is 14.6kgf/cm. As the compression force
257
 ordinary concrete strength, there are two ways to
increase the shear capacity, one is to increase the
amount of longitudinal reinforcing bars, another one
is to enhance the web reinforcement (See Fig. 11).
In the first case, the bending capacity, defined as
My,will be simultaneously increased. As the bending
capacity is corresponding to the yielding of main
reinforcing bars, the shear force Vy at the yielding
moment becomes very large, as calculated by Eq. (5).
The shear capacity will also be increased according to
the JSCE code as,

Vlo<(ρ1)5 (6)
Fig. 11 Interaction of shear capacity and ductility.
According to Eq. (5) and Eq. (6), the yielding
shear force increases more than the shear capacity as
varied from 10kgf/cm2 to 100kgf/cm in the dynamic
the amount of main reinforcement rises in general.
response, the shear capacity reaches 17.2kgf/cm2 as
RC members with this kind of reinforcement
the maximum value in considering the axial
arrangement are very brittle, and fail suddenly
compression force. The FEM computation above
without much ductility. The failure of the
shows that shear stress in the upper column can reach
intermediate columns of the subway stations is
18kgf/cm2, a little bit higher than the shear strength.
categorized into this case.
Another aspect to evaluate the seismic resistant
On the other hand, if the shear capacity is
capacity of the RC member is the ductility level. In
increased by enhancing the web reinforcement, the
order to avoid the sudden failure, the RC member
shear capacity V can be larger than the yielding shear
should be designed to fail after yielding of
force V, Then the ductility of the RC member will be
longitudinal reinforcement. In the dynamic
elevated. RC members having higher shear capacity
simulation of the underground RC, it was found that
the intermediate column failed before or just after generally have higher seismic resistant performance.
yielding of longitudinal bars. This was also pointed
out by some 3-D FEM analysis9. So it is necessary to 5. ENHANCEMENT OF RC SEISMIC
discuss the ductility of the column in RC frames. PERFORMANCE
The ductility level of RC member can be
estimated by, As discussed above, the shear capacity level of the
existing column in the underground RC is found to be
(4) lower and results in the small ductility. In order to
enhance seismic resistance of RC structures in
where, N is a factor which influences the ductility of general, both shear capacity and ductility are
the RC member. If N is less than unity, the member effective.
behaves with brittle mode of failure. Vyis the shear But, the capacity of intermediate columns of
force corresponding to the bending capacity denoted underground box sections hardly influence on the
by My (Fig. 11), and if defined as yielding shear force overall shear deformation of RC and the induced
herein. Then, Vy,can be calculated as, sectional forces, because the structural deformation is
much associated with that of interacting soil
(5) foundation. Since the chief required performance of
where, H is the height of the column. the column is to sustain vertical forces no matter how
In the case of upper column of the underground large the shear deformation of the box section is
RC concerned, the bending capacity is more than 410 induced. For underground RC structures, the member
ton-m when yielding of longitudinal bars takes place. ductility is much to be focused in design of newly
The shear force at this time is about 20kgf/cm2. Then, constructed structures and retrofitting of existing RC.
ductility factor N is 0.73, which is much smaller than Then, the ratio of shear capacity to the flexure will be
unity. Certainly, the intermediate column ductility is of great interest to us.
less accompanying diagonal shear cracking.
For evaluating the seismic resistance, both the (1) Increase in web reinforcement ratio in the
shear capacity and the ductility should be considered. intermediate column
From Eq. (2), it seems that for RC member with Computation is based on the same RC frame as
that in the previous chapter. The web amount in the
258
 Fig. 12 Inelastic strain representing damage in time domain.
Fig. 14a Crack pattern of structure just before failure.

Fig. 13 Deformation profile of RC-soil system at failure. Fig. 14b Cracks developed at the failure.

column is increased up to 0. 76% (D16 with spacing c) Crack pattern of the RC frame
of 7.5cm, volumetric ratio is 1.66%), and the shear For identifying the failure location and the failure
capacity can reach 33. 3kgf/cm2 according to the shear mode when the web of the column is much reinforced
equation of JSCE code (no consideration of axial in shear, Fig. 14a shows the crack pattern of the
compressive force). As no change is made in structure just before the failure. In Fig. 14a large
longitudinal reinforcement arrangement, the yielding cracks are seen in several places such as the upper
shear force Vyis still 20kgf/cm. The shear to flexure column, the lower column, the middle slab-column
capacity ratio N is 1.67. In this case, the columns are joint and the corners of the frame. Fig. 14b shows the
expected to have sufficient shear resistance against cracks which developed at the failure. All the cracks
seismic actions. shown in Fig. 14b are introduced in the last step of
a) Inelasticity of the whole RC structure computation. We can see just distinct shear cracks
Fig. 12 shows the inelastic strain (I) of shear within the upper slab, near the left corner, crossing
enhanced case with additional web reinforcement in the section of this column. So the shear is brought
time domain. In this figure, we can see that the about at the upper slab near the corner.
structure fails at 7.32 second. d) Internal stresses in the intermediate column
b) Dynamic response of the RC frame The computational results of the shear enhanced
Fig. 13 shows the deformational profile of the case indicate that the collapse takes place not at the
station at the failure. It can be seen that the intermediate column but at the upper slab near the
deformation is concentrated at the corner of the RC left corner. Discussion of induced forces in the
outer frame, and the failure took place at the upper internal column may be advisable for further
slab. Since the number of finite element layer is just clarifying failure section in the slab.
one, failure possibility will be again checked in terms Fig. 15 shows the relation between the nominal
of shear forces developed and empirical formula by shear stress and the relative shear displacement of the
the JSCE code. There is no localization of upper and lower columns. The relative displacement
deformation in the columns, since the intermediate is normalized by the height of the column. For the
column is strongly reinforced in shear. upper column, when the failure occurs, the shear
stress in the column is small. The maximum
259
 Fig. 16 shear stress variation at the failure section.

Fig. 17 Inelastic strain representing damage in time domain.

Fig. 15 Shear stress-displacement relationship for columns.

(2) Reduction of main reinforcement for
improving ductility
The computation was conducted based on the
normalized displacement is 0.5% and the maximum
shear stress is 18kgf/cm2. same RC frame in which web reinforcement ratio in
the column is increased up to 0.76%, which is the
4n the other hand, the lower column experiences
same as that of section (1). But the longitudinal
shear stress of 12kgf/cm2 and no failure takes place.
reinforcement ratio reaches 3.67%, trying to enhance
There exists less difference of Fig. 10a and Fig. 15,
the ductility of the column. The shear capacity comes
but quite much difference in failure mode. The shear
up to 32. lkgf/cm' according to the JSCE code. The
capacity is much increased in this case, but the
main reinforcement ratio is reduced so as to get the
ductility response remains unchanged as the main
reinforcement is kept constant and absolutely large. yield shear stress as 14.4kgf/cm2 equivalent to yield
of main steel. The shear/flexure capacity ratio N can
e) Internal stresses in the slab corner
reach 2.23 in this case. Thus, much ductility is
Fig. 16 shows the variation of shear stress in the
section of the corner. It can be seen that the shear granted with sufficient shear capacity.
stress reaches 18kgf/cm2 just before failure. The a) Inelasticity of the whole RC structure
Fig. 1.7 shows the inelastic strain (1) of the
estimated shear capacity of this slab in use of JSCE
enhanced shear and reduced flexure case in time
code10 is 15.2kgf/cm. So, the slab is supposed to fail
domain. Herein, we can see that the structure does not
in shear.
fail until the seismic load leaves.
f) Collapse mechanism of the subway station
In the shear enhanced case, according to FEM b) Internal stresses in the intermediate column
simulation, it is considered that the intermediate RC It can be seen more clearly by comparing the
induced forces in the internal column and the failure
column gets strong enough to resist the shear and will
section in the slab with the web enhanced case
not fail under the earthquake wave. But, the overall
discussed in the previous section.
structural performance is not improved much as the
main reinforcement ratio is kept unchanged. Then, Fig. 18 shows the relation between the nominal
shear stress and the relative displacement. The
the partial strengthening merely results in the shift of
relative displacement is normalized by the height of
the weakest section.
260
 Fig. l9 Shear stress variation at the corner section.

Table 2 Parametric study on member ductility

Fig. 18 Shear stress-displacement relationship for columns.

the column. For the upper column, the maximum 6. SOIL STRUCTURE AND SEISMIC
normalized displacement can reach 1.2% and the ACTIONS
maximum shear stress does 18kgf/cm2. On the other The dynamic system discussed above includes the
hand, the lower column undertakes shear stress as underground RC and soil foundation. The change of
12kgf/cm2. Through the comparison of this figure reinforcement ratio in the RC frame was proved to be
with Fig. lla and Fig. 15, the improved ductility very effective from a view point of the damage
accompanying the flexural nonlinearity is clearly control. In this section, interacting aspects with soil
identified. foundation and the wave property will be studied for
c) Internal stresses in the slab corner further understanding of the behavior of whole
In order to compare this high ductility case with dynamic system.
previous one, the variation of shear stress in the
section of comer, where the failure took place in the (1) Soil profile
enhanced shear with just additional web In the original case for Kamisawa station, the soil
reinforcement case is shown in Fig. 19. The foundation property varies in the vertical direction.
maximum shear stress is reduced down to 15kgf/cm The shear modulus of soil is specified larger as the
by reducing the flexure capacity but elevating the depth increased. Fig. 20 shows the soil profile used
ductility of the structure. as original. The soil around the upper deck is softer
than the one around the lower deck. As the softer soil
(3) Summary of the parametric computation may cause larger induced shear deformation in the
The parametric computation of three cases is RC culvert, larger shear deformation is supposed to
summarized in Table 2. These trial computations be produced in the upper column under the
show that both the shear capacity and the ductility are earthquake. This may be the reason why the column
important factors for seismic resistant performance. of the upper floor failed but the column of the lower
The RC member with low shear strength will fail in floor had just few shear cracks.
shear mode under seismic load, and the RC structure The effect of soil profile can be checked by
with higher shear strength and ductility can survive intentionally changing the soil profile in the
during the earthquake. It can be found from the computation. If the failure mode and position would
computational experience that the ductile structure be affected by the change of soil profile, soil structure
possesses higher seismic resistance. would be found to have much to do with the failure
261
 Fig. 22a Deformation profile of RC-soil system just before
Fig. 20 Assumed shear modulus profile for soil foundation. failure.

Fig. 22b Deformation profile of station at failure.

Fig. 21 Inelastic strain representing damage in time domain.

mechanism. For this purpose, the computation is It means the failure position changes from upper to
performed with the different soil profile from the lower as the soil profile be changed as uniform.
standard, that is, the foundation around the RC frame The crack pattern shown in Fig. 23a supports the
is taken as the same material property of the shear above stated discussion on the failure location. There
modulus of 1318 kgf/cm2, which is similar to that of are many cracks in the two columns, unlike the case
layer 3 in the original case. Other factors are kept the in Fig. 9. The damage is much heavier than that in the
same as the original one. original case (Fig. 9a). Fig. 23b shows cracks
The computational results of this modified soil developing in the last step. The diagonal shear failure
profile are shown in Fig. 21-Fig. 24. The inelastic and cracking are sharply detected.
strain of the RC box in time domain is shown in Fig. 24 shows the internal shear stresses and the
Fig. 21. By comparing Fig. 21 with the original case shear deformation of the column for both upper and
(Fig. 7), much difference of the response is identical. lower parts of the frames. The great shear
The structure failed at 5.2 sec with abrupt increase in deformation is produced in the lower column which
the structural damage. As the mean shear modulus of finally failed. The shear stress is also higher in the
the original case is 1603 kgf/cm2, the premature lower column than that in the upper one.
failure would be associated with soft foundation. All these computational results make it clear that
The next point of interest is the location of failure. the soil profile will affect the damage occurring in the
Fig. 22 shows the deformation profile of the RC RC underground structure. It also proves that the soft
frame under the dynamic excitation. In Fig. 22a, foundation around the upper deck of the frame is one
which shows the deformation profile just before of the main reasons why upstairs columns are
failure, larger deformation at the lower column is damaged in reality. Regarding the magnitude of shear
seen than the original case (Fig. 8a). In order to focus deformation in soil-RC coupled systems,
on the failure position more clearly, the deformation combination of wave characteristics, structural and
soil stiffness is a point of design. As has being well
profile of the station at failure is shown in Fig. 22b.
The deformation is concentrated in the lower column. pointed out, the entire system has to be modeled in
seismic design of underground structures.

262
 Fig. 23a Crack pattern of the RC frame just before failure.

Fig. 24 Shear stress-displacement relationship for columns.

Fig. 23b Cracks produced at the failure.

(2) Effect of vertical seismic action

One of the characteristics of the seismic motion
used in the above discussion is that the vertical
component of the ground motion is rather high. In
fact, the maximum vertical acceleration reached
about 40% of the maximum horizontal one. In
general, the vertical component of seismicity is
ignored in some cases owing to its small contribution
to the structural safety and dynamic response.
Fig. 25 Inelastic strain representing damage in time domain.
However, influenced structural behavior was
reportedt l even when the vertical ground motion is
perfectly cut off in the dynamic computation. So it is
meaningful to discuss the influence of the vertical
ground motion, especially for the case where failure
does not occur in the column. .
In section 5. 1 the case where the column is
heavily reinforced in shear was discussed and the
failure took place in the upper slab. Here, the
computation will be done for this case while the
vertical component of the ground motion is
completely cut off. Fig. 26 Deformation profile of station at failure.
Computed results under no vertical seismicity are
summarized in Fig. 25-Fig. 28. Fig. 25 shows the Fig. 26. It can be seen that the deformation is
inelastic strain in time. The structure failed at 6.96 localized in the left side of the middle slab. It means
second. the failure position shifts from the upper slab to the
In order to identify the failure position more middle slab owing to avoidance of the vertical
clearly, the deformation profile at failure is shown in component of the ground motion.

263.
 Fig. 27 Shear stress-displacement relationship for columns.

It is needed to check the internal stress condition Fig. 28 Shear stress variation in time domain.

carefully to confirm the mode of failure. Fig. 27

on the underground structures, two kinds of wave are
shows the internal shear stresses and the shear
used for the dynamic computation (Fig. 29).
deformation of the column. It can be seen through
The first one is an artificial seismic wave
comparison with Fig. 15 that the shear strength
induced to the column is hardly affected but merely produced for Koutouen area, which is based on the
earthquake record on solid engineering foundation'2.
the induced axial compression is influenced.
The period of this wave is close to the Kobe wave
The shear stress variations at section A-A and B-
which the authors used while the maximum
B in the upper and middle slabs are shown in Fig. 28.
horizontal acceleration is a little smaller (Fig. 29a).
Section A-A is the point where the slab failed in shear
The second one is the seismic wave recorded at
mode in case where only shear reinforcement is
Amagasaki city. It shows medium acceleration level
enhanced(Fig. 16). Provided that the up-down motion
and long period of motion (Fig. 29b). The maximum
is cut in computation, the shear stress in section A-A
acceleration in both horizontal and vertical directions
does not reach the failure level(over 15 kgf/cm2), but
is similar. Fig. 3o shows the inelastic strain
the shear stress in section B-B of the middle slab
representing the damage level in dynamic response. It
reaches shear capacity (over 9kgf/cm2). So, the
can be seen that the target underground RC failed
structure would fail in the middle slab as a result of
under the artificial Koutouenn wave but no failure
erasing the vertical ground motion in computation.
was brought under the Amakasaki wave.
Even though the vertical ground motion is not a
The deformation profile of RC-soil system when
primary cause of the failure, the combination of
failure occurs is shown in Fig. 31. This deformation
horizontal and vertical wave may change the dynamic
response of the structure including the location of the profile is very close to that under the seismic load of
failure position. The effect of this combination is very Kobe wave (Fig. 8). The deformation is also
complex and needs to be further studied to concentrated into the upper column, where the failure
understand the core of mechanism. took place in reality.
The crack pattern at failure is shown in Fig. 32.
(3) Seismic action characteristics The zoom of upper column shows many diagonal
As stated above, the Kobe wave recorded at Kobe shear cracks occurring at this time while only few
meteorological observatory includes very high cracks are introduced at the other parts of the RC
horizontal acceleration with short period. In order to frame. This situation is also similar to the original
study the effect of characteristics of the seismic wave case under the Kobe wave.

264
 Fig. 30 Inelastic strain representing the damage in time domain.
Fig. 29a Artificial acceleration profile for Koutouenn areal2.

Fig. 31 Deformation profile of RC-soil system at failure under

Koutouen wave.

Fig. 29b Acceleration records at Amagasaki city.

Fig. 32 Crack pattern at the failure under Koutouenn wave.

265
 Fig. 33 Shear stress-displacement relationship for columns.

The internal shear stresses in the intermediate In group II, the soil profile and the characteristics
columns are shown in Fig. 33 for these two cases. of the seismic wave have been changed for some trial
Under the seismic load of Koutouen, the shear stress computation, in order to understand the effect of
in the upper column comes up to nearly 20 kgf/cm2, soil-structure interaction in the dynamic system of
and failed at this time, which is similar to the original underground RC structures.
case under the Kobe wave (Fig. 10a). In the case of According to the results of these analyses, we
Amagasaki wave, the shear stress was less than 15 have,
kgf/cm, and there is no failure in both of the (1) Nonlinear mechanics of reinforced concrete
columns. coupled with soil nonlinearity can be a tool for
Within the limited numbers of input waves, the simulating the collapse of underground structure
seismicity with higher acceleration may cause under seismic action and evaluating its performances.
collapse of the middle columns. But, it is not the (2) The collapse of subway station is due to the low
general case because all the wave characteristics and shear capacity and ductility of the intermediate
mechanical properties of both RC and foundation columns.
would be influential on the safety. It can be just (3) Well known and accepted strategy that an
concluded that overall seismic performance of increase in shear capacity enhances ductility and
underground RC must be estimated in consideration seismic resistance is confirmed to be effective for
of soil and RC nonlinearity under dynamics in time design of underground RC, too.
and space. (4) The soil profile affects the damage level and
failure location. Softer foundation may cause larger
shear damage.
6. CONCLUSION
(5) Combination of horizontal and vertical
In this study, seven cases of dynamic analyses for component of seismic action is a factor which affects
an underground reinforced concrete box-section the location of failure of existing underground RC.
damaged in Kobe city were conducted. The group I (6) Seismic waves from different sources give similar
(Table 2), including three cases, was performed for dynamic response if the wave has similar period and
the collapse study with seismic resistant view point. maximum acceleration. Further discussion is needed
on the characteristics of earthquake wave.

266
ACKNOWLEDGMENT: The authors appreciate 6) Shawky, A. and Maekawa, K.: Nonlinear dynamic analysis
for underground reinforced concrete structures, East Asian-
technical advises by Dr. A. Shawky, Cairo Pacific Conference on Structural Engineering and
University, and kind offer of the seismic acceleration Construction, EASEC-5, Australia, July, 1995.
diagram by Dr. Hajime Ouchi, Ohbayashi 7) Ohsaki, Y.: Some notes on Masing's law and nonlinear
Corporation. response of soil deposits, Journal of the Faculty of Eng. (B),
The University of Tokyo, Vol. XXXV, No. 4, 1980.
8) Shawky, A. and Maekawa, K.: Computational approach to
REFERENCES
path-dependent nonlinear RC/soil Interaction, J. Material,
1) Samata, S., Nagamitsu N., Yamamoto, K. and Mori, S.: A
Concrete. Struct., Pavements, JSCE, No. 532/V-30, pp. 197-
study on a failure mechanism of subway station analyzed by a
207, Feb, 1996.
non-linear seismic deformation method, Proceedings of
9) Tajiri, M., Samata, S., Matsuda, T. and Ouchi, H.: A study
Technical Conference on the Great Hanshin- Awaji
on the damage of underground railway structure during the
Earthquake, JSCE, Tokyo, pp. 231-238, 1996.
Great Hanshin Earthquake, Proceedings of Technical
2) Tajiri, M., Samata, S., Siba, Y., Sakashita, K. and Watanabe,
Conference on the Great Hanshin- Awaji Earthquake, JSCE,
K.: An analytical study on failure mechanism of a subway
Tokyo, pp. 255-262, 1996.
station damaged by the 1995 Hyogoken-Nanbe earthquake, 10) JSCE, Standard specification for design and construction of
Proceedings of Technical Conference on the Great Hanshin-
concrete structure, pert I(Design), 1st ed, Tokyo, 1986.
Awaji Earthquake, JSCE, Tokyo, pp. 239-246, 1996.
11) Shawky, A. and Maekawa, K.: Collapse mechanism of
3) Okamura, H. and Maekawa, K.: Nonlinear Analysis and
underground RC structures during Hanshin Great
Constitutive Models of Reinforced Concrete, Gihodo,
Earthquake, Cairo first International Conference on
Tokyo, 1991.
Concrete Structures, 1996.
4) Okamura, H. and Maekawa, K.: Reinforced concrete design
12) Ouchi, H., Ejiri, J., Matsumoto, N. and Matsuoka, Y.: A
and size effect in structural nonlinearity, invited, Proceeding
study on the Shinkansen viaduct damage during the Great
of JCI International Workshop: Size Effect in Concrete Hanshin Earthquake and seismic performance of retrofitted
Structures, Sendai, Japan, pp. 1-20, 1993. structure, Proceedings of Technical Conference on the Great
5) Collins, M.P. and Mitchell, D.: Prestressed Concrete
Hanshin-Awaji Earthquake, JSCE, Tokyo, pp. 305-312,
Structures, PRENTICE HALL, Englewood Cliffs, New
1996.
Jersey 07632, pp. 345, 1991.

(Received August 5, 1996)

地 震 作 用 を受 け る 地 中鉄 筋 コ ン ク リ ー トの 破 壊 解 析

Xuehui AN・ 前 川 宏 一

本研 究 は地 中鉄 筋 コ ンク リ ー ト構 造 の破 壊 解析 を提 示 し,経 路 依 存 性 を考慮 した材 料 構 成 モ デ ル に基 づ

き,地 中 構 造 の破 壊 機 構 に つ い て論 ず る もの で あ る. 考 慮 した 非線 形性 は鉄 筋 コ ンク リー ト,地 盤 材料 お よ
び地 盤構 造 間の応 力 伝 達機構 であ る. 主 と して鉛 直 荷重 を支 え る部材 のせ ん断 破壊 に着 目 し,破 壊 に及 ぼ す
地 盤 の非 線 形性, 地 震 入 力 の違 い 及 び鉛 直 成 分 の有 無 につ い て検 討 を行 っ た. 保有 す る耐 震性能 は地 盤 と構
造 自体 の非線 形 性 を共 に考慮 しな けれ ば な らな い こ とが 確認 され た.

267