Journal of Japan Association for Earthquake Engineering, Vol.4, No.

3 (Special Issue), 2004

DAMAGE ANALYSIS AND SEISMIC DESIGN OF

RAILWAY STRUCTURES FOR HYOGOKEN-NANBU
(KOBE) EARTHQUAKE

Akihiko NISHIMURA

Member of the JSCE, Dr. Eng., Vice-president, Representative Director,

JR Souken Engineering Co., Ltd., 2-8-38 Hikari-cho, Kokubunji-shi, Tokyo 185-0034, Japan,
nisimura@jrseg.co.jp

ABSTRACT: The devastation wrought on railway structures by the 1995

Hyogoken-Nanbu earthquake has shown that there is an urgent need for developing a
new seismic design of railway structures. After the earthquake, many groups set to
investigate the reasons for the damage using static and dynamic analyses. This paper
presents causes of seismically induced damage to bridges elucidated by some of these
analyses of the earthquake, a contemporary seismic design method for railway structures,
and issues related to the new seismic design method.

Key Words: Damage analysis, seismic design, earthquake disaster.

INTRODUCTION

The 1995 Hyogoken-Nanbu (Kobe) earthquake in Japan caused severe damage to railway structures,
including completely collapsing. This extensive damage emphasized the need to develop new
procedures and specifications to assess existing structures and to improve seismic designs for new
structures. This disaster has thus become fixed in memory for the same reason as the 1923 Great
Kanto earthquake. Majority of Japanese civil engineers were over confident that such serious
structural damage would not be widespread in the event of a large earthquake because of progress in
seismic design.
In the pre-Kobe earthquake, seismic design has been based exclusively on the seismic coefficient
method, which was firstly adopted for civil structures in the early Showa era (1930’s). Although the
methodology has been modified after every major earthquake disaster, basic philosophy has remained
unchanged. The Niigata earthquake (1964) caused railway bridges to collapse, but there was no heavy
damage or collapse to be experienced until the event of the Hyogoken-Nanbu earthquake. The
devastation wrought on civil structures by this earthquake indicates to civil engineers that the
conventional seismic design method is still inadequate. This leads to conclude that this seismic
coefficient method could be used to design bridges that withstand only levels of the past large-scale
earthquakes. In this article, the lessons on seismic design learned from this latest earthquake and a new
seismic design methodology are described.

- 184 -
 CAUSES OF DAMAGE TO VIADUCTS IN THE HYOGOKEN-NANBU EARTHQUAKE

Outline of Damage

The Hyogoken-Nanbu earthquake caused considerable damage to railway constructions such as

elevated bridges, embankments, and other civil structures. Its impact affected many of railway, the
Sanyo Shinkansen line, Tokaido Main line, Hanshin-Kobe line, Itami line, Hanshin Main line, and
others. This section focuses on the damage of concrete bridges on the Sanyo Shinkansen line. The
damage extended from the epicenter in northeast towards as far as the cities of Akashi and Takatsuki,
as shown in Fig. 1.The damage also caused to concrete bridges in the section between Osaka and
Himeji, particularly concentrated on several kilometers between Shin-Osaka side of the portal of
Rokko tunnel and starting point, and between Nagasaka tunnel and Nishiakashi. And, it also spread
over elevated bridge between Kyoto and Shin-Osaka, nearby Takatsuki on the Tokaido Shinkansen
line. The forms of damage to concrete bridges can be summarized as follows. Floor beams and/or
slabs of reinforced concrete viaducts and abutments collapsed after supporting columns failed. About
1,200 columns were collapsed and further 3,400 were damaged in total in the region. Concrete
viaducts and bridges collapsed completely at many locations, as shown in Fig. 1. All the viaducts were
rigid-frame structures, and some of them collapsed because their columns failed after shear cracks
developed either at the top or at the bottom of columns due to huge horizontal seismic loads.

Fig. 1 Location of major damaged railroad

Causes of Damage

Consideration of Yield Strength and Deformation Capacity

Damage is categorized as pattern “S” and “M” that indicate the causes of failure, “shear” and
“flexural”, respectively, as shown in Fig. 2. It is also classified into 4 levels A to D (with D:
non-damaged). Shear failure is generally brittle in concrete therefore the structure is completely
destroyed if it suffers seismically induced shear failure.
The strength and deformation capacity of viaducts on the Sanyo Shinkansen line were verified
according to "Standards for Design of Railway Constructions (Concrete Constructions)". Horizontal
seismic coefficients were calculated by static linear analysis corresponding to the strength of flexural
yield, flexural ultimate, and shear of each framework member. In this analysis, the strength of each
member was determined by the amount of reinforcement and so on specified in the verification of
failed structures. Strength coefficients of materials were obtained from the measurements taken just
after the earthquake. Patterns of damage found in the site investigations, and analysis results are
shown in Fig. 2. The analysis was carried out in the sections along with and perpendicular to

- 185 -
longitudinal bridge axis. Each framework member was classified into damage categories, as follows:
1) Shear failure: kh(My) > kh(Vy)
2) Shear failure beyond flexural yield: kh(My) < kh(Vy) < kh(Mu)
3) Flexural failure: kh(Mu) < kh(Vy)
where, kh(My) is the horizontal seismic coefficient at flexural yield strength, kh(Mu) is the horizontal
seismic coefficient at flexural ultimate strength, and kh(Vy)is the horizontal seismic coefficient at shear
yield strength.
Damage level
C B A

Shear (S)

Flexural (M)

Fig. 2 Patterns and levels of damage

8
7
6
5
Number

4 S: Shear failure
3 MS: Flexural yielding prior to
2 shear failure
1 M: Flexural failure
0
S MS M
Damage patterns in analysis

Fig. 3 Relationship between observed damage classifications and analytical damage patterns

All the viaducts that completely collapsed in the earthquake (categorized as SA) are thus assumed
to have suffered shear failure at level A. Fig. 3 shows that analytical damage patterns and observed
damage classifications of completely collapsed viaducts are well corresponded together.

Study by Seismic Response Analysis

Fig. 4 shows a geological cross section at location where a shinkansen viaduct collapsed. Most
damage occurred on diluvium that is newly formed with generally low strength. Thus, characteristic of
subsurface layer is considered as one of causes of seismically induced damage.
Two collapse cases of Hansui and Shimokema viaducts, which are located 7 km apart and two
non-collapse cases of No. 2 Danjo viaduct near Hansui viaduct and No. 2 Noma viaduct on a diluvium
formation were analyzed. They are all 2-layer and 3-span rigid-frame structures as schematically
shown in Fig. 5. All of them have a 1.2 m-diameter and 8 m-deep cast-in-place pile foundation except
No. 2 Noma viaduct has a spread foundation, as shown in Fig. 6 with their boring logs.
An outline of the analysis schematically shown in Fig. 7 is described as follows. Firstly, the
response of subsurface layers to seismic waves propagating from the hypocenter was calculated. Then,

- 186 -
it was applied to analytical models that take account of soil-structure interactions between ground,
foundations, and structures. This yielded a dynamic response of the structure. Stiffness of structural
members was fixed according to the load-displacement relationship for a concrete structure.

Fig. 4 Geological section from the Sanyo Shinkansen Shin-Osaka station to Shin-Kobe station

Fig. 5 General drawing of elevated bridge

Fig. 6 Foundations and soil boring logs

- 187 -
 The EW component of acceleration measured at GL-83 beneath Kobe Port Island was reduced
with the distance from the epicenter to each target viaduct. This modified seismic motion was then
applied as an input seismic wave to the structure.
Seismic response of surface layers is computed by using one-dimensional effective stress analysis
program. Soil properties were determined according to the “Japanese Design Code for Railway
Structure Foundations”. Table 1 shows maximum values of seismic wave on the ground surface at
each viaduct. It is observed that acceleration, velocity, and displacement varied with the differences in
subsurface layers beneath viaducts. The analytical results are shown in Fig. 8 as ratios of bending
moment and shear force strength to sectional forces of seismic response analysis on the vertical axis of
diagram. Ratios less than unity represent collapse.

Fig. 7 Schematic diagram of the analysis

Table 1 Maximum values of surface seismic motion

Peak Peak Peak

Position Peak displacement
acceleration velocity displacement
Shimokema 235(280) 42.2(27.4) 19.0(11.1)
Value in parenthesis
Hansui 256(309) 34.7(30.2) 15.5(12.2) is seismic base layer
No.2 Danjo 277(309) 29.8(30.2) 12.7(12.2) seismic motion used
in analysis
No.2 Noma 309 30.2 12.2

Ranges of value represent deviations of sectional force strength by different proposed formulas.
The figure shows clearly that collapsed viaducts have smaller value for shear. Viaducts failed due to
shear forces varied with different characteristics of subsurface layers.

Analytical Prediction of Damage Origin

All of the structures investigated were designed to withstand a horizontal seismic coefficient of 0.2
under the Standard. Therefore, many of them were devastated by a recorded horizontal seismic force
of larger than 0.2. Further, where the beam of shinkansen rigid-frame viaducts collapsed, greater
damage occurred that one factor of safety of the bending moment is smaller compared to that of shear

- 188 -
strength.

Fig. 8 Ratios of sectional forces for viaducts during an earthquake

Fig. 9 Collapse process of a rigid-frame viaduct

Fig. 9 schematically illustrates a mechanism of damage to a rigid-frame viaduct. Shear cracks that
developed on the top or bottom of middle beam advanced rapidly under a strong earthquake motion.
Ultimately, spalling concretes fell out. Consequently, the structure could no longer support its own
weight then upper structure slipped down perpendicular to the longitudinal bridge axis.

SEISMIC DESIGN FOR RAILWAY STRUCTURES

The facts mentioned above indicate that following procedures are important to seismic design for
railway structures.
1) Taking inland earthquake into account
2) Evaluating safety of members by considering failure modes of structures
3) Necessary to use dynamic analysis methods and considering dynamic behavior of surface
ground in response analysis of structures.
Seismic design of a railway structure should therefore be carried out according to the following
procedures. Firstly, damage degree of the structure (seismic performance) should be identified in
respect to damage control. Secondly, surface ground responses are analyzed by inputting a design
earthquake motion in the base ground. Thirdly, responses of the structure are analyzed with the input
of obtained surface ground response. Finally, seismic performance of the structure can be checked
based on the obtained structural responses.

Setting of Design Earthquake Motion

Setting of Earthquake Motions for Bedrock

There are two types of design earthquake motion, L1 and L2 to be determined for bedrock. L1
earthquake motion has a recurrence probability of a few times during service life of a structure. It has

- 189 -
approximately the same level as the acceleration spectrum corresponding to high quality ground,
which used to be adopted in allowable stress designs. The maximum response value of acceleration is
250 gal corresponding to a damping coefficient of 5%. L2 earthquake motion that is caused by a
near-land-large-scale interplate earthquake or an inland earthquake near structure with high intensity
has lower occurrence probability. It is classified into 3 following types.
1) Spectrum I: acceleration spectrum corresponding to near-land interplate earthquakes of
magnitude 8.0 and epicenter distance of 30 to 40 kilometers.
2) Spectrum II: acceleration spectrum based on statistic analysis of the earthquake data recorded
in the past inland earthquakes caused by active faults.
3) Spectrum III: also representing the motions caused by active inland faults, but based on the
analysis of active faults when such a model of active fault is available.

Setting of Design Earthquake Motions on Ground Surface

Table 2 presents 8 types of soil profile used in the design. Depending upon each soil profile, design
acceleration response spectra on the ground surface are determined corresponding to L1 earthquake
motion, Spectrum I and Spectrum II of L2 earthquake motion. Fig. 10 is an example for Spectrum II.

Table 2 Soil profile types

Soil Profile Types Period (sec) Soil Profile Names/Generic Descriptions

G0 - Hard rock
G1 - Bedrock
G3 - 0.25 Diluvium
G4 0.25 – 0.50 Dense soil
G5 0.50 – 0.75 Dense soil to soft
G6 1.00 – 1.50 Very soft soil
G7 1.50 - Extremely soft soil

Fig. 10 Design response spectra of acceleration on ground surface for Spectrum II

Seismic Performance of Structure

Setting of Seismic Performance Level for Structures

Corresponding to presumed levels of repair and reinforcement of structures that may be required after an
intense earthquake, seismic performance can be categorized into 3 levels as briefly described in Fig. 11.
These performance levels are mainly defined by the degree of ease to recovery structures after an

- 190 -
earthquake. Therefore, the relationship between the earthquake motion level and seismic performance
was established as follows.
For L1 earthquake, structural seismic performance SPI should be satisfied by all the structures.
For L2 earthquake, SPII should be satisfied by more important structures, and SPIII by others.

Structural Seismic Damage Levels of Member

Performance Levels
Damage Level 1: No damage
Seismic Performance I (SPI) Damage Level 2: Damage that may require repair
Capability of maintaining the original depending on situation
functions without any repair and no Damage Level 3: Damage requiring repair
excessive displacement occurring during an Damage Level 4: Damage requiring repair, and
earthquake replacement of members depending on situation

Seismic Performance II (SPII) Stability Levels of Foundation

Capability of making quick recovery of the
original functions with repairs after an Stability Level 1: No damage (loading smaller than
earthquake bearing capacity)
Stability Level 2: Damage requiring repair depending
Seismic Performance III (SPIII) on situation
Capability of keeping the overall structure in Stability Level 3: Damage requiring repair, and
place without collapse during an earthquake correction of structure depending on situation
Fig. 11 Relationship among seismic performance levels, damage levels of member and stability levels
of foundation (bridges and viaducts)

Reinforcing bars yield Maintaining maximum load

in axial direction Maintaining yield load
C
B D
Lateral load

Skeleton curve
Envelop curve for analysis
A of test result
Cracks occur
Deformation

Fig. 12 Relationship of lateral load-deformation relation for reinforced concrete members with a
general level of compressive axial force

Damage Levels of Members

It is considered appropriate to determinate damage levels of a member by considering the relation
among member properties, damage state, and repair method. Fig. 12 shows a load-deformation
relationship of a member in the case of flexural failure occurring first under an exerting compressive
axial force. Considering its characteristics, damage level of the member is defined corresponding to
deformation range as follows.
1) Damage Level 1: before point B
2) Damage Level 2: from point B to C
3) Damage Level 3: from point C to D
4) Damage Level 4: after point D
Once a relationship between damage level and deformation is established, the amount of
deformation that may be directly calculated from a response analysis becomes a relevant index for
checking damage levels. If nonlinear behavior of a member is evaluated with a mechanical model of
bar, generally, rotation angle or curvature for the section of plastic hinge is taken as an index for
checking member. The relationship between them is shown in Table 3.

- 191 -
 Table 3 Relationship between damage level of member and rotation angle

Limit Values of Rotation Angle

Damage level 1 υyd : Yield rotation angle of member
υmd : Rotation angle of member corresponding to the maximum deformation resulting from
Damage level 2
the peak lateral loading
υnd : Rotation angle of member corresponding to the maximum deformation being able to
Damage level 3
resist the yield lateral loading
Damage level 4 υud : Rotation angle of member for limiting the excessive deformation in axial direction

Level 1 Level 2 Level 3

Py : Yield bearing capacity
Lateral load, P

B C Pm : Maximum bearing capacity

Pm
A δy : Yield displacement
Py
δm : Displacement corresponding to
maximum loading
δu : Ultimate displacement

δy δm δn
Displacement, δ
Fig. 13 Load-displacement curve with stability levels of a foundation

Stability Levels of Foundation

In order to ensure seismic performance for an overall structure, stability level of foundation should be
determined with two aspects: damage level with respect to the stability of foundation itself, and
damage level of members constituting it.
For evaluating these items, two indexes that are response ductility ratio defined as a ratio of
seismic response displacement to yield displacement obtained from the load-displacement curve of
foundation, and residual displacement should be used. Using displacement indices in the
load-displacement curve of foundation with stability levels generally illustrated in Fig. 13, the stability
levels of foundation can be determined as follows.
1) Stability Level 1: In principle, load acting on the foundation should be less than its yield
bearing capacity and no excessive displacement occurs. Stress resultant of members
composing the foundation should not exceed yield strength.
2) Stability Level 2: Bearing capacity should be maintained sufficiently even though either
subgrade supporting foundation or members composing the foundation or both of them are
deformed plastically. There is neither displacement detrimental to maintenance of structural
functions, nor residual displacement to be allowable after an earthquake.
3) Stability Level 3: Bearing capacity should be maintained sufficiently enough to protect the
structure from collapse because of damage of bearing subgrade or structural members.
Besides the values of stability level are set corresponding to the types of foundation.

Limit Values
Based on the consideration explained above, the parts of a rigid frame viaduct where damage may
occur, are illustrated in Fig. 14, and an example of the relationship among the limit values of
structure’s seismic performance levels, member’s damage levels and foundation’s stability levels is
shown in Table 4.

- 192 -
 Fig. 14 Illustration of damaged parts of a rigid frame viaduct

Table 4 An example of the relationship among the limit values of structure’s seismic performance
levels, member’s damage levels and foundation’s stability levels (rigid frame viaduct)

Structure SPI SPII SPIII

Damage Superstructure girder and underground beam 1 2 3
level of Other beams 1 3 4
member Columns 1 3 3
Stability level of foundation 1 2 3

SAFETY (SEISMIC PERFORMANCE) CHECKING OF STRUCTURES

Static nonlinear (i.e. “pushover”) analysis is stipulated to apply in the checking process. Its procedures
are: i) modeling overall structure (from superstructure to foundations) to a frame structure, and
subgrade supporting foundation to a spring system; ii) setting strengths and deformation behaviors for
structural members and subgrade reactions based on mentioned above; iii) calculating structural
displacements by increasing seismic load incrementally and plotting the relationship between seismic
load and displacement. By indicating various critical steps in the load-displacement curve, the failure
of overall structure can be grasped. Such critical steps include the steps where structural capacity
reaches the limit values of yield, maximum and ultimate. The ultimate displacement can be determined
by comparing calculated displacement with the limit value listed in Table 3. The ultimate displacement
for overall structure is determined from the displacement when the capacity of a certain member
belonging to superstructure or foundation reaches the limit value of ultimate state. Therefore, a
structure is judged safe if its ultimate displacement is larger than the response displacement calculated
by dynamic analysis method, that is, the designed seismic performance satisfies the seismic
performance objective.
Furthermore, the judgment of each member’s damage level and foundation’s stability level should
be conducted by checking the deformation state of the step in the pushover analysis, whose
displacement is as same as that calculated by the dynamic analysis method. The main contents about
this checking are described as follows.

Checking Damage Levels of Members

In checking damage level of a concrete member, failure mode should be judged at first. If shear stress
calculated is smaller than shear strength when flexural strength is reached, the failure mode is defined
namely as flexural failure mode, inversely as shear failure mode. In the codes, it is stipulated that real
strength of reinforcing bar should be used in the judgment of failure mode.
Damage level of a member with flexural failure mode can be judged with the deformation
calculated from a static nonlinear analysis. However, in the case of shear failure mode, judgment can

- 193 -
only be done according to the strength. That means, deformation behavior of a member with shear
failure mode should be set to linearity in the overall structural model for static nonlinear analysis.

Checking Stability Levels of Foundation

In the approach, following items are stipulated for checking the stability level of foundation.
1) Response ductility ratio of foundations;
2) Damage levels of the members composing foundations;
3) Residual displacement of foundations.
This residual displacement is taken as a main index for checking Seismic Performance II.
Therefore, allowable value of the residual displacement should be limited within a small range so that
operational function of train could be quickly recovered.
All the items above are checked based on results obtained from the static nonlinear analysis.

CONCLUSIONS

This paper presents a seismic damage analysis of railway structures due to the 1995 Hyogoken-Nanbu
earthquake in Japan and a new seismic design established based on the lessons learned from the
analysis results, with basic principles and some important advances for the design.
Adequacy of the methodology should be confirmed through precise analysis of real damage
examples from the past earthquakes. The current seismic design methodology will be improved and
become more and more perfect along with achievements from modern researches in the near future.
Moreover, the methodology becomes rather complicated because of the consideration of
non-linearity of both structure and subgrade. In order to avoid meaningless complication, approaches
for seismic design are essentially used to express damage level of structures. Therefore, damage state
of a designed structure during an intense earthquake could be anticipated.
At last, it is noticed that the precision of input parameters concerning structure and subgrade and
the computing accuracy should be appropriate to the execution of computer. Even though the level of
design method is promoted, a design using incorrect input data cannot be considered as a good one.

REFERENCES

Nishimura, A. (2002). “Seismic design for railway structures, earthquake resistant design codes in
Japan.” JSCE, January 2002.
Nishimura, A. (1999). “Earthquake resistant performance and design method for foundation.” RTRI
Report, Vol. 13, No. 3, March 1999.

(Submitted: March 31, 2004)

(Accepted: July 2, 2004)
Copyright JAEE

- 194 -